AU2019337807B2

AU2019337807B2 - Aircraft engine maintenance system and method

Info

Publication number: AU2019337807B2
Application number: AU2019337807A
Authority: AU
Inventors: Adam DURANT; Lisa LENTATI; Sam Richardson; Antony Rix
Original assignee: Satavia Ltd
Current assignee: Satavia Ltd
Priority date: 2018-09-11
Filing date: 2019-09-11
Publication date: 2022-12-22
Anticipated expiration: 2039-09-11
Also published as: WO2020053778A1; AU2019337807A1; GB201814784D0; GB2577065A

Abstract

There is provided an aircraft engine maintenance system comprising at least one input interface that receives, when in operation, data related to a given aircraft engine from a sensor arrangement that is connected to the given aircraft engine and supplementary data from a database arrangement; the supplementary data describes operational and maintenance information that characterizes the given aircraft engine. The aircraft engine maintenance system includes a data processing arrangement that receives, when in operation, the data related to the given aircraft engine as an input data and the supplementary data and uses the input data and the supplementary data to generate normalized data by ingesting the input data and the supplementary data. Ingesting the input data and the supplementary data includes normalizing relative weightings of the input data and the supplementary data to generate the normalized data. The data processing arrangement uses the normalized data to determine an operating status of the given aircraft engine by applying the normalized data to a statistical model of the given aircraft engine. There is thereby generated a data output representative of the operating status of the given aircraft engine, wherein the data output includes at least one of: (i) a maintenance schedule required for the given aircraft engine; and (ii) a measure of wear or an adjustment optimization of the given aircraft engine.

Description

AIRCRAFT ENGINE MAINTENANCE SYSTEM AND METHOD

TECHNICAL FIELD

The present disclosure relates generally to aircraft engine maintenance systems that employ sensors to monitor aircraft engine operation to generate corresponding sensor data, and to apply the sensor data to mathematical models to generate a data output representative of a "state-of-health", namely an operating status, of aircraft engines to semi-automate or fully-automate recommendations of maintenance of the aircraft engines. Moreover, the present disclosure also relates to methods for (of) operating aforesaid aircraft engine maintenance systems. Furthermore, the present disclosure also relates generally to systems that predict a "state-of-health" of aircraft from sensed atmospheric contamination, wherein such state-of-health includes, for example, aircraft engine health, but not limited thereto; moreover, the present disclosure relates to methods for (of) predicting a "state-of-health" of an aircraft from sensed atmospheric contamination, for example a "state-of-health" of an aircraft, as well as its at least one engine and its equipment. Furthermore, the present disclosure is concerned with computer program products comprising a non- transitory computer readable storage medium having computer-readable instructions stored thereon, the computer-readable instructions being executable by a computerized device comprising processing hardware to execute the aforesaid methods.

BACKGROUND

Aircraft engines are key components of propulsion systems of aircraft that generate mechanical power. A high cost is associated with running contemporary aircraft engines which results in a need to optimise their operation, management and upkeep to ensure that the engines deliver a high degree of reliability. Such contemporary aircraft engines consume considerable quantities of fuel (kerosene) and emit substantial particulate pollutants and C0₂ when in operation; appropriate aircraft engine maintenance is, for example, desirable to reduce emissions of such particulate pollutants and C0₂.

Many approaches have been proposed to maintain aircraft engines, for example when to schedule a maintenance for a given aircraft engine within a manufacturer's recommended "temporal window", whether or not to have spare aircraft engines in reserve or which engineers to have available at a maintenance warehouse to service the given aircraft engine.

Analytics products are known that optimise recommendations for maintaining a given aircraft; these products require a wide variety of data to monitor and model a performance of one or more aircraft engines of the given aircraft. For example, it will be appreciated that a functioning and a performance of the given aircraft and its one or more engines depend on multiple settings, exposure to contaminants, previous maintenance operations, wear and other condition degradations. The functioning and performance of the one or more engines are susceptible to being measured by using multiple sensors coupled to the one or more engines, from which a maintenance scheduling operation is potentially recommended. Furthermore, such a wide variety of data often originates from multiple locations and sources, for example, from scheduling and sensor data that resides with a carrier, from maintenance records held by a Mandatory Occurrence Report (MOR) and from environmental data from a service provider. On account of a need to analyse the wide variety of data from the multiple sources in respect of multiple indicators, it is found to be technically challenging to aggregate the data from multiple sources which are mutually independent and to determine indicators that are relevant to a specific type of aircraft or engine. In practice, contemporarily, decisions are taken and recommendations are often made that are based only on a limited set of related data that has been analysed; such a limited set of data leads to inaccuracies in predicting when a maintenance operation or a scheduling operation is required. Furthermore, these data are not simple to aggregate as the sensor readings are taken with respect to time rather than flight cycles or schedule records. Furthermore, maintenance records are often provided as unstructured records; in many cases, the records are paper- based records. With the data residing in multiple locations, even if one entity maintains organised records, the data are potentially not simply joined or merged with records from another entity. Existing approaches or recommendations for maintenance of aircraft engines are mainly based on a selection of related data rather than a complete set of related information due to a difficulty in merging multiple sources of data together in a reliable manner.

Moreover, when a process of predicting a maintenance schedule is executed manually, the process is dependent not only on an ability to process the multiple sources of data, but also on an availability of experts to analyse the data and generate accurate predictions. On account of an availability of experts being limited, often by temporal constraints, a time taken for an expert to devote sufficient effort to analyse the data and generate a meaningful prediction results in the prediction being available potentially too late to avert a failure or accident from occurring.

Therefore, on account of problems encountered with known approaches as aforementioned, there arises a need to semi-automate or automate recommendations of maintenance operations or scheduling operations of aircraft engines and generating a "state-of- health", namely an operating status, of aircraft for maintaining their one or more aircraft engines.

In a published US patent application US2018/0012423A1 (" Maintenance Systems and Methods for ECS Elements", Honeywell), there is described a maintenance system for an environment conditioning element of an environmental control system (ECS) of a vehicle. The system includes a data collection module configured to receive geographical areas of travel for the vehicle over respective periods of time. The data collection module is configured to determine a pollution value and a time value for each of the geographic areas of travel. The system further includes a pollution count module coupled to the data collection module and receiving the pollution values and the time values. The pollution count module is configured to determine a pollution count for the environment conditioning element based on the pollution values and the time values. The system further includes a reporting module coupled to the pollution count module and receiving the pollution count. The reporting module is configured to generate a report for a user that includes the pollution count.

Earlier published patent applications EP3200039A1, US20173232741 and US2018012423A1 are concerned with aircraft engines and issues of contamination experienced by such engines.

SUMMARY The present disclosure seeks to provide an aircraft engine maintenance system that generates a data output representative of an operating status of a given aircraft engine by ingesting data related to the given aircraft engine and also supplementary data and proactively recommends a maintenance or a scheduling operation of the given aircraft engine in an automated way. Moreover, the present disclosure seeks to provide an improved monitoring system that processes sensor data acquired from sensors coupled to a given aircraft engine, and also supplementary data, to generate an output indicative of a "state-of-health" of the given aircraft engine.

According to a first aspect, there is provided an aircraft engine maintenance system comprising at least one input interface that receives, when in operation, data relating to a given aircraft engine from a sensor arrangement that is connected to the given aircraft engine and supplementary data from a database arrangement, wherein the supplementary data describes operational and maintenance information that characterizes the given aircraft engine; characterized in that the aircraft engine maintenance system further comprises a data processing arrangement that receives, when in operation, the data related to the given aircraft engine as an input data and the supplementary data and processes the input data and the supplementary data to generate normalized data, wherein the normalized data is generated by ingesting the input data and the supplementary data, wherein ingesting the input data and the supplementary data includes normalizing relative weightings of the input data and the supplementary data to generate the normalized data, wherein the data processing arrangement:

(i) processes the normalized data to determine an "state-of-health" of the given aircraft engine by applying the normalized data to a numerical model of the given aircraft engine; and

(ii) derives from the numerical model at least one predicted aircraft health parameter for the given aircraft engine and uses or provides the derived at least one predicted aircraft health parameter as an input for scheduling a maintenance intervention for the given aircraft.

The invention is of advantage in that a synergistic combination of ingesting a wider spectrum of input data, normalizing the wide spectrum of input data, and using more advanced mathematic models enables the at least one predicted aircraft health parameter to be more accurately and reliably determined, and therefore the corresponding scheduled maintenance intervention to be more appropriately determined to reduce a risk of engine failure on a first hand, and not incurring unnecessary cost and down-time by excessive maintenance work on a second hand.

Optionally, in the aircraft engine maintenance system, machine learning is used to determine at least one of the following : a selection of a numerical model to be used in the aircraft engine maintenance system, coefficients of a numerical model to be used in the aircraft engine maintenance system, or a set of features used in a numerical model to be used in the aircraft engine maintenance system. More optionally, in the aircraft engine maintenance system, at least one measure of spatial or temporal uncertainty is determined using a statistical or machine learning model arranged to receive values of a contaminant determined from at least one of: a database, a web API, a NWP model, a measurement. Optionally, in the aircraft engine maintenance system, at least one environmental contaminant exposure is determined using a statistical or machine learning model arranged to compensate for bias or spatial or temporal uncertainty in values of a contaminant determined from at least one of: a database, a web API, a NWP model, a measurement.

Optionally, in the aircraft engine maintenance system, machine learning is used to determine at least one of the following : the selection of a numerical model, the coefficients of a numerical model, or a set of features used in a numerical model. The machine learning beneficially employs one or more algorithms that are adaptive in response to data being processed therethrough; in such a manner, the machine learning improves an accuracy of normalization of data when the system makes recommendations, and also an accuracy of various models employed in the system. Training data is provided to the machine learning to configure the machine learning, for example by using historical data obtained from previous aircraft flights and previous weather conditions.

Optionally, in the aircraft engine maintenance system, the data processing arrangement employs machine learning to optimize adaptively ingestion and normalization to provide the normalized data, and to optimize an accuracy of the numerical model. More optionally, the machine learning is arranged, for example pursuant to training data provided, to monitor temporal rates of change or temporal patterns of change of one or more parameters indicative of the "state-of-health" of the aircraft engine, to determine a rate of degradation of the aircraft engine; for example, the machine learning monitors temporal rates of change of a plurality of the parameters; optionally, the parameters are derived from processing the sensor data.

Optionally, the aircraft engine maintenance system comprises at least one storage medium that is coupled to the data processing arrangement, wherein the at least one storage medium stores the input data and the supplementary data.

Optionally, the aircraft engine maintenance system comprises at least one output interface that provides to a user the data output representative of the at least one predicted aircraft health parameter for the given aircraft engine.

Optionally, in the aircraft engine maintenance system, the data processing arrangement, when in operation, processes normalized data comprising a reference to a serial number of the given aircraft engine that is used to select at least a part of the supplementary data for use in the aircraft engine maintenance system.

Optionally, in the aircraft engine maintenance system, the data processing arrangement, when in operation, processes the input data comprising at least one of a carrier data, a contaminant exposure data, a temperature record, humidity records, and Automatic dependent surveillance - broadcast (ADS-B) positional data, wherein the carrier data comprises at least one of schedules, maintenance or cycles of operation of the given aircraft engine.

Optionally, in the aircraft engine maintenance system, the data processing arrangement, when in operation, processes the supplementary data comprising raw data, wherein the raw data comprises a logging filename, a data type and an associated metadata of the input data. Optionally, in the aircraft engine maintenance system, the data processing arrangement, when in operation, processes the supplementary data comprising records of an amount of maintenance intervention historically associated with an aircraft or engine.

Optionally, the aircraft engine maintenance system cleans the supplementary data to obtain cleaned supplementary data by performing at least one of:

(i) correcting spelling errors of the supplementary data;

(ii) identifying erroneous flight details, correcting or filtering identified erroneous flight details, timestamps or cycle counts;

(iii) reformatting the supplementary data; and

(iv) filtering null or duplicate values of the supplementary data.

Optionally, the aircraft engine maintenance system extrapolates the supplementary data to obtain extended supplementary data by performing at least one of:

(i) identifying changepoints in the data;

(ii) encoding unstructured, textual or classification data in a numerical representation;

(iii) grouping similar events;

(iv) computing a count or accumulated value since an event; and

(v) computing a count of the number of occasions a given value has crossed a threshold since an event.

More optionally, the aircraft engine maintenance system merges at least one of:

(i) the cleaned supplementary data into the normalized data; and

(ii) extended supplementary data into the normalized data.

Optionally, in the aircraft engine maintenance system, the numerical model is pre-trained for at least one of:

(i) classifying a performance of the given aircraft engine; and

(ii) scoring a health condition of the given aircraft engine. Optionally, in the aircraft engine maintenance system, the output data includes a reference to changes in running costs, recommended maintenance actions, or operational actions to be undertaken by a user.

According to a second aspect, there is provided a method for (of) operating an aircraft engine maintenance system, characterized in that the method includes: using at least one input interface of the aircraft engine maintenance system to receive data relating to a given aircraft engine from a sensor arrangement that is connected to the given aircraft engine and supplementary data from a database arrangement, wherein the supplementary data describes operational and maintenance information that characterizes the given aircraft engine; using a data processing arrangement of the aircraft engine maintenance system to receive the data related to the given aircraft engine as an input data and the supplementary data and using the input data and the supplementary data to generate normalized data, wherein the normalized data is generated by ingesting the input data and the supplementary data, wherein ingesting the input data and the supplementary data includes normalizing relative weightings of the input data and the supplementary data to generate the normalized data, using the data processing arrangement:

(i) to process the normalized data to determine an operating status of the given aircraft engine by applying the normalized data to a numerical model of the given aircraft engine; and

(ii) to derive from the numerical model at least one predicted aircraft health parameter for the given aircraft engine and to use or provide the derived at least one predicted aircraft health parameter as an input for scheduling a maintenance intervention for the given aircraft. Optionally, in the method, machine learning is used to determine at least one of the following : a selection of a numerical model to be used in the aircraft engine maintenance system, coefficients of a numerical model to be used in the aircraft engine maintenance system, or a set of features used in a numerical model to be used in the aircraft engine maintenance system. More optionally, in the method, at least one measure of spatial or temporal uncertainty is determined using a statistical or machine learning model arranged to receive values of a contaminant determined from at least one of: a database, a web API, a NWP model, a measurement. More optionally, in the method, at least one environmental contaminant exposure is determined using a statistical or machine learning model arranged to compensate for bias or spatial or temporal uncertainty in values of a contaminant determined from at least one of: a database, a web API, a NWP model, a measurement.

Optionally, in the method, machine learning is used to determine at least one of the following : the selection of a numerical model, the coefficients of a numerical model, or a set of features used in a numerical model. The machine learning beneficially employs one or more algorithms that are adaptive in response to data being processed therethrough; in such a manner, the machine learning improves an accuracy of normalization of data when the system makes recommendations, and also an accuracy of various models employed in the system. Training data is provided to the machine learning to configure the machine learning, for example by using historical data obtained from previous aircraft flights and previous weather conditions.

Optionally, the method includes arranging for the data processing arrangement to employ machine learning to optimize adaptively data ingestion to provide the normalized data, and an accuracy of the numerical model. More optionally, the machine learning is arranged, for example pursuant to training data provided, to monitor temporal rates of change or temporal patterns of change of one or more parameters indicative of the "state-of- health" of the aircraft engine, to determine a rate of degradation of the aircraft engine; for example, the machine learning monitors temporal rates of change of a plurality of the parameters; optionally, the parameters are derived from processing the sensor data.

Optionally, in the method, the normalized data comprises a reference to a serial number of the given aircraft engine, for use when selecting the supplementary data.

Optionally, in the method, the input data comprises at least one of a carrier data, a contaminant exposure data, a temperature record, humidity records, and Automatic dependent surveillance - broadcast (ADS-B) positional data, wherein the carrier data comprises at least one of schedules, maintenance or cycles of the given aircraft engine.

Optionally, in the method, the supplementary data comprises raw data, wherein the raw data comprises a logging filename, a data type and an associated metadata of the input data.

Optionally, in the method, the supplementary data comprises records of an amount of maintenance intervention associated with an aircraft or engine.

Optionally, the method comprises cleaning the supplementary data to obtain cleaned supplementary data by performing at least one of:

(i) correcting spelling errors of the supplementary data,

(ii) identifying and correcting or filtering erroneous flight details, timestamps or cycle counts;

(iii) reformatting the supplementary data; and

(iv) filtering null or duplicate values of the supplementary data.

Optionally, the method comprises extending the supplementary data to obtain extended supplementary data by performing at least one of: (i) identifying changepoints in the data;

(iii) grouping similar events;

(iv) computing a count or accumulated value since an event; and

More optionally, the method comprises merging at least one of:

(i) the cleaned supplementary data into the normalized data; and

(ii) extended supplementary data into the normalized data.

Optionally, the method comprises pre-training the numerical model for at least one of:

(i) classifying a performance of the given aircraft engine; and

(ii) scoring a health condition of the given aircraft engine.

Optionally, in the method, the output data includes a reference to changes in running costs, recommended maintenance actions, or operational actions to be undertaken by a user.

According to a third aspect, there is provided a computer program product comprising instructions to cause the system of the first aspect to carry out the method of the second aspect.

Embodiments of the present disclosure substantially eliminate or at least partially address the aforementioned drawbacks in existing approaches for generating a data output representative of an operating status of a given aircraft engine by ingesting an input data and a supplementary data.

Additional aspects, advantages, features and objects of the present disclosure are made apparent from the drawings and the detailed description of the illustrative embodiments construed in conjunction with the appended claims that follow.

It will be appreciated that features of the present disclosure are susceptible to being combined in various combinations without departing from the scope of the present disclosure as defined by the appended claims.

BRIEF DESCRIPTION OF THE DRAWINGS

The summary above, as well as the following detailed description of illustrative embodiments, is better understood when read in conjunction with the appended drawings. For the purpose of illustrating the present disclosure, exemplary constructions of the disclosure are shown in the drawings. However, the present disclosure is not limited to specific methods and instrumentalities disclosed herein. Moreover, those in the art will understand that the drawings are not to scale. Wherever possible, like elements have been indicated by identical numbers.

Embodiments of the present disclosure will now be described, by way of example only, with reference to the following diagrams wherein :

FIG. 1 is a schematic illustration of a system according to an embodiment of the present disclosure;

FIG. 2 is an illustration of a flow diagram depicting an operation of an aircraft engine maintenance system according to an embodiment of the present disclosure;

FIG. 3 is an illustration of a flow diagram of a data ingestion process according to an embodiment of the present disclosure;

FIG. 4 is an illustration of a flow diagram of steps of a method for (of) operating an aircraft engine maintenance system according to an embodiment of the present disclosure; FIG. 5 is an illustration of an exploded view of a distributed computing system or cloud computing implementation according to an embodiment of the present disclosure;

FIG. 6 is a schematic illustration of a system in accordance with an embodiment of the present disclosure;

FIG. 7 is a flowchart illustrating steps of a method for (of) training the system of FIG. 6, namely a method for (of) training an engine health model for sequent use thereafter to predict an engine health parameter of an aircraft engine in accordance with an embodiment of the present disclosure; FIG. 8 is a flowchart illustrating steps of a method for (of) predicting an engine health parameter of the engine or the aircraft using a predictive engine health model in accordance with an embodiment of the present disclosure;

FIG. 9 is a graphical illustration of a predicted engine health in an ordinate Y- axis plotted against a number of cycles since last shop visit in an abscissa X-axis to predict an expected remaining useful life in accordance with an embodiment of the present disclosure;

FIG. 10 is a graphical illustration of a predicted engine health in the ordinate Y- axis plotted against a number of cycles since last shop visit in the abscissa X- axis to predict an expected lifespan of the engine for different levels of exposure to contaminants in accordance with an embodiment of the present disclosure;

FIG. 11 illustrates a graphical illustration of a flight trajectory model of flight phases and contaminant density at the engine at the flight phases in accordance with an embodiment of the present disclosure;

FIGS. 12A to 12B are graphical illustrations of an integrated method for (of) using the system of FIG. 6 to determine a cost awareness (CA) and a risk mitigation (RM) for predicting a risk of high altitude ice (HAI) being encountered on an aircraft route at different periods from a departure time, in accordance with an embodiment of the present disclosure;

FIGS. 13A to 13B are flow diagrams illustrating steps of a method for (of) training the system of FIG. 6 to provide a predictive engine health model using machine learning and using the predictive engine health model to predict an engine health in accordance with an embodiment of the present disclosure;

FIG. 14 is an illustration of steps of a computing arrangement for determining contaminant data using one or more servers, a supercomputing and/or a distributed computing platform according to an embodiment of the present disclosure; and

FIG. 15 is an illustration of an exploded view of a distributed or cloud computing implementation in accordance with an embodiment of the present disclosure.

In the accompanying drawings, an underlined number is employed to represent an item over which the underlined number is positioned or an item to which the underlined number is adjacent. A non-underlined number relates to an item identified by a line linking the non-underlined number to the item. When a number is non-underlined and accompanied by an associated arrow, the non- underlined number is used to identify a general item at which the arrow is pointing.

DETAILED DESCRIPTION OF EMBODIMENTS

The following detailed description illustrates embodiments of the present disclosure and ways in which they can be implemented. Although some modes of carrying out the present disclosure have been disclosed, those skilled in the art would recognize that other embodiments for carrying out or practicing the present disclosure are also possible.

In overview, embodiments of the present disclosure are concerned with an aircraft engine maintenance system comprising at least one input interface that receives, when in operation, data relating to a given aircraft engine from a sensor arrangement that is connected to the given aircraft engine and supplementary data from a database arrangement, wherein the supplementary data describes operational and maintenance information that characterizes the given aircraft engine; characterized in that the aircraft engine maintenance system further comprises a data processing arrangement that receives, when in operation, the data related to the given aircraft engine as an input data and the supplementary data and uses the input data and the supplementary data to generate normalized data, wherein the normalized data is generated by ingesting the input data and the supplementary data, wherein ingesting the input data and the supplementary data includes normalizing relative weightings of the input data and the supplementary data to generate the normalized data, wherein the data processing arrangement:

(i) processes the normalized data to determine an operating status of the given aircraft engine by applying the normalized data to a numerical model of the given aircraft engine; and

Pursuant to embodiments of the present disclosure, the aircraft engine maintenance system, when in operation, generates a data output representative of "state of health” of the given aircraft engine in an automated way, namely to a higher degree of reliability and accuracy than has hitherto been feasible. The aircraft engine maintenance system thus improves an operating safety of the given aircraft engine by appropriately scheduling maintenance of the given aircraft engine. Optionally, the aircraft engine maintenance system reduces a risk of damage to the given aircraft engine or an occurrence of in-flight failures of the given aircraft engine, and thereby increases the given aircraft engine's lifetime. Optionally, the aircraft engine maintenance system schedules maintenance interventions for the given aircraft engine in response to the exposure of the given aircraft engine to atmospheric contaminants. Moreover, optionally, the system schedules maintenance interventions based on an exposure rate including both a short-term exposure, and a long-term exposure of the aircraft engine to the contaminants. Such "short-term exposure" relates to a time period in a range of circa 1 to 200 flight cycles, or 1 day to 1 month, whereas such "long-term exposure" relates to a time period in a range of circa 100 flight cycles and above. Beneficially, long-term exposure may be considered to be the exposure of an engine since its last engine wash, shop visit or major overhaul, while short-term exposure assesses a period shorter than this.

Optionally, the aircraft engine maintenance system prioritises maintenance for the given aircraft engine according to the exposure rate of the aircraft engine to the contaminants. A prioritised maintenance intervention is, for example, an inspection. A prioritised maintenance intervention is, for example, a shop visit or major overhaul, where the engine is stripped down, worn or damaged parts are repaired or replaced, and the engine is rebuilt. A prioritised maintenance intervention is optionally an engine wash, where an engine or part of an engine is cleaned with a liquid, solid or thermal cleaning process intended to remove deposits of corrosion, dust and other contaminants that potentially have accumulated. Both shop visits and engine washes can increase an engine's efficiency and reduce its exhaust gas temperature. A particular, a benefit of the present invention is that it facilitates scheduling such as maintenance interventions for greatest effectiveness and efficient use of resources, balancing a cost of making an intervention and an operational value gained. The aircraft engine maintenance system is capable of providing a benefit reduces emission of carbon dioxide (C0₂) and other pollutants, for example micro- particulates, generated by the aircraft engine when in operation.

Optionally, the at least one input interface obtains the input data including at least one of a carrier data, a contaminant exposure data, a temperature record, humidity measurement data, and Automatic dependent surveillance - broadcast (ADS-B) positional data from a server. The aircraft engine maintenance system is beneficially communicatively connected to a user device for receiving the input data from the user device; the user device optionally comprises the at least one input interface. The at least one input interface optionally obtains the input data from an Automatic Dependent Surveillance-Broadcast (ADS-B system), for example as aforementioned. The input data is optionally communicated to the user device/the automatic system through a communication network. The user device optionally comprises the at least one output interface.

Hardware components employed to implement the aircraft engine maintenance system of the present disclosure optionally include a data processing arrangement and the aforesaid at least one input interface. In an embodiment, the aforesaid hardware components include a server, and the at least one input interface is disposed with a client side of the server or the user device. The server is optionally a tablet computer, a desktop computer, a personal computer or an electronic notebook. In an example embodiment, the server is optionally a cloud service. The user device optionally comprises a personal computer, a smartphone, a tablet computer, a laptop computer or an electronic notebook computer. The components of the aircraft engine maintenance system are coupled in communication via the communication network that is optionally a wired network or a wireless network, or a combination of both. In an embodiment, the server optionally performs the one or more steps of the data processing arrangement. The server optionally comprises a database that includes the input data and the supplementary data. The input data includes at least one of a carrier data, a contaminant exposure data, a temperature record, humidity data, and Automatic dependent surveillance - broadcast ADSB positional data. The carrier data comprises at least one of schedules, maintenance, and cycles of the given aircraft engine on a periodic basis (for example, a weekly basis). The supplementary data includes a raw data, wherein the raw data comprises a logging filename, a data type and an associated metadata of the input data.

The aircraft engine maintenance system optionally determines historical contaminant exposure by analyzing a plurality of historical trajectories of a plurality of aircraft engines as a function of spatial location and time and estimating contaminant concentration of the plurality of aircraft engines for each trajectory using an atmospheric model. In an embodiment, the contaminant exposure is optionally estimated using a trajectory or trajectory model. When preparing data for training a machine learning model, the plurality of aircraft engines are beneficially associated with sets of training, test or validation engines.

The aircraft engine maintenance system optionally determines at least one of: an average contaminant exposure, a total and cumulative estimate of exposure to each contaminant type, per cycle and/or per time interval, and incorporates one or more of these estimates in the engine health model. The aircraft engine maintenance system beneficially determines estimates of contaminant exposure as a function of at least one flight phase and incorporates the estimate for at least one flight phase in the engine health model. The system optionally determines values for aggregates of types of contaminant, for example totalling dust of different particle sizes, with an optional weighting of different parameters. The aircraft engine maintenance system optionally estimates a cost associated with actual engine usage to determine a cost of contaminant exposure of the aircraft engine. In an embodiment, calculating an exposure to a contaminant for a flight follows methods set out in a published patent application US2018/0012423A1 (" Maintenance Systems and Methods for ECS Elements", Honeywell), mutatis mutandis considering that the asset of interest is an aircraft engine. Cumulative exposure is estimated by summing exposure to a contaminant over two or more flight cycles between a maintenance event and a present time. Average contaminant exposure is susceptible to being estimated by methods including dividing a cumulative exposure by the number of cycles, and calculating a moving average of contaminant exposure (namely, a short-term exposure as defined above).

In an embodiment, contaminant exposure is calculated by integrating contaminant exposure values determined from a database or an output of a numerical weather prediction (NWP) model according to a given trajectory. The computing arrangement is beneficially provided with a flight plan that includes takeoff and landing airports and a trajectory defined by list of one or more waypoints defined by expected time parameters, latitude parameters, longitude parameters, altitude parameters. Waypoints are optionally provided from Automatic dependent surveillance - broadcast (ADS-B) positional data. Reference is made a published patent application GB1909354.1 (Applicant: Satavia) that elucidates a numerical weather prediction (NWP) model in greater detail.

In another embodiment, where only the takeoff and landing airports are provided, a trajectory is susceptible to being estimated using methods known in the art, for example by preparing a typical trajectory between the two airports representative of one or more previous paths taken by an aircraft. A trajectory optionally also comprises only one point or a sequence of points representing a part of a flight path of interest, for example to traverse a region known to be subject to adverse weather. In such an example, the system processes to determine a risk only for the provided subset trajectory. In an embodiment, where time parameters are not otherwise provided, time is determinable from the scheduled departure or arrival time of a flight.

In another embodiment, where only one time parameter is provided, times of each waypoints is determinable iteratively from another known waypoint in turn by dividing the distance between the two waypoints by the aircraft's typical speed during that part of the flight.

It will be appreciated that waypoints are known on a similar grid to a numerical weather prediction (NWP) model, where waypoints are widely separated and they are susceptible to being interpolated or extrapolated, for an example on a straight-line basis, so that spatial locations are sampled with a resolution approximating to that of the NWP model.

The aircraft engine maintenance system optionally estimates an expected contamination exposure of the aircraft engine over a given period of time by estimating a probability, a probability density function, or multiple parameters including expected mean and standard deviations of estimated contaminant exposure. The system optionally determines either of, or both of, the mean and standard deviation of the contaminant exposure or density for at least one of a time within a year and a time-of-day. For example, the historical data sets are grouped by calendar month and hour of the day, thus providing an estimate of the mean and standard deviation of the engine health parameter that varies from month to month, and over the course of a day.

The aircraft engine maintenance system optionally determines a mean or standard deviation, or both a mean and a standard deviation, of contaminant exposure or density by using the machine learning or regression methods, such as at least one of:

(i) linear regression;

(ii) neural network comparisons or correlations between data sets; (iii) decision trees, decision forest or gradient boosted decision tree, where time within a year and/or a time of day are included as regression parameters or features.

Machine learning may beneficially be arranged to estimate current or future evolution of an output or target value that is at least one of: an engine health measure, engine sensor value or contaminant exposure or density as follows. For at least a training set of aircraft or aircraft engines, a set of target values and features are determined. Output (target) values of contaminant exposure or contaminant density may be obtained from a measurement, a NWP model or analysis of a trajectory in respect of an actual or simulated aircraft flight as described above. Features may additionally include class or numerical representations of the respective takeoff and arrival airports, regions, terrain, proximity to the ocean, position (including altitude, latitude and longitude), temperature, precipitation and wind speed. Features may be normalised. The at least one machine learning or regression method is then trained to use the features to predict the output (target) using methods known in the art of machine learning.

Machining learning is described in greater detail below. Beneficially, in embodiments of the present disclosure, at least one environmental contaminant exposure is estimated using a statistical or machine learning model arranged to correct a NWP model for bias or uncertainty.

Once trained, a machine learning or regression method is beneficially provided with coefficients determined from a previous training, and features calculated similarly from a new case such as a new aircraft flight, to make an estimate of the (unknown) output or target value.

Beneficially, a model is retrained or adapted with new data, to update coefficients used by the model and to make more accurate predictions in future. A model is optionally arranged to estimate separately contaminant exposure due to takeoff, and due to arrival, by considering separately the relevant phases of flight in proximity to takeoff (namely, taxi from gate, takeoff, climb and optionally part of cruise) and landing (optionally, a part of cruise, descent, landing and taxi to gate).

The arrangement of a machine learning method to estimate an engine health measure or engine sensor value is further described below.

In an example embodiment, the aircraft engine maintenance system determines a probability that a given contaminant is present in a given flight trajectory, rather than or in addition to computing an average concentration for certain types of contaminant, such as ice crystals. The presence of a contaminant is beneficially determined by comparing to a threshold, for example, in a case of ice crystals at a spatial concentration of 1 gram per cubic metre. Where a plurality of conditions for which the presence or absence of contaminant is known or is derivable, a probability or a risk of the contaminant being present is determinable from the respective probability distribution of such observations over a balanced set of conditions. Embodiments of the present disclosure beneficially employ a machine learning method that is able to output a probability estimate, including without limitation logistic regression, neural network with logistic output, random forest, cross-validation ensemble model, wherein the machine learning method is beneficially trained to predict such a probability estimate according to the methods set out in the present disclosure.

In an example embodiment, an aircraft engine that has a value of an engine health parameter that is below-expected value of such a parameter, given a same cycle count, would be prioritised for maintenance ahead of an engine with an above-expected engine health parameter. The system optionally generates a notification or provide alerts to the user device based on the generated data output that is representative of the operating status of the given aircraft engine.

The aircraft engine maintenance system optionally schedules maintenance interventions for the given aircraft engine in response to a "state of health” of the aircraft's engine as computed by the maintenance system. Moreover, the aircraft engine maintenance system optionally schedules maintenance interventions based on a containment exposure rate including both a short-term exposure, and a long-term exposure of the given aircraft engine. "Short term" and "long term" have been elucidated in the foregoing.

The aircraft engine maintenance system of the present disclosure is capable of being applied in a manually-supervised manner. However, optionally, the aircraft engine maintenance system is capable of functioning in an automated manner, namely the aircraft engine maintenance system automates a process of receiving, processing and managing data related to operational aircraft engines. Additionally, the system provides a platform on which decision support tools can be built. The decision support tools optionally utilize analyzed data and predictions from a machine learning (ML) model to provide alerts and recommendations in addition to scheduled reports providing an overview of the aircraft engine.

In the present disclosure, "machine learning” (ML) relates to adaptive algorithms that are capable of reconfiguring themselves depending on training data provided to configure the algorithms. Such adaptive algorithms are beneficially implemented as one or more software products that are executable on computing hardware. Optionally, the adaptive algorithms include explicit functions such as correlation functions, data comparison functions based on adaptive neural networks, as well as multi-dimensional solution spaces whose states are capable of being dynamically adjusted in response to receiving training data. In an example embodiment of the aircraft engine maintenance system, the machine learning comprises explicit mathematical functions including any one of: linear regressions, neural networks, decision trees, decision forests or gradient boosted decision trees, radial basis functions, support vector machines, Gaussian processes, or principal component analysis. The machine learning optionally includes at least one of time within a year and/or time of day as regression parameters. However, for determining or for estimating an exposure associated with a trajectory of a given aircraft, it is beneficial to employ machine learning (ML) in combination with explicit mathematical functions as image correlation, resampling of data, and extrapolation of data.

The aircraft engine maintenance system of the present disclosure optionally collects the input data from mutually different sources, wherein the mutually different sources optionally include a Mandatory Occurrence Report (MOR) of the aircraft engine and airlines, but are not limited thereto. The input data from the mutually different data is normalized and then subject to data merging and data comparisons or correlations. For example, in such a manner, the input data is combined with additional supplementary data such as an environmental contaminant exposure to determine a "state of health”, namely an operating status of the aircraft engine. The aircraft engine maintenance system optionally employs a machine learning model to process the input data and the supplementary data, for example as aforementioned, to produce alerts and reports with respect to operation cost and maintenance needs and recommendations for the given aircraft engine.

In an embodiment, the data output optionally includes data relating to schedule records, engine sensor data, flight cycles, maintenance action records, and inspection records; such records and data is often of diverse formats and structures and has to be normalized to corresponding normalized data that is subject to various mathematical functions such as correlation, comparison analysis, linear transformation, regression and such like. Without normalization of data being performed, certain input data would have an unrepresentative influence on determining the "health of state" of the given aircraft engine, with a risk that the maintenance schedule that is recommended is not optimal. Such data is optionally beneficially be used as features for a machine learning model, for example as aforementioned.

In an embodiment, the data processing arrangement is capable of producing several outputs, for example indicative of aircraft engine "state of health", and a recommended maintenance schedule for the aircraft engine taking into account its "state of health". Outputs optionally include outputs from a previous stage and target predictions output by a machine learning model. The normalized data describing the engine history is optionally available through a web Application Programming Interface (API) or a web user-interface. The normalized data is optionally a combined dataset which combines data from all data sources into a normalized table sorted by an engine serial number (ESN) and a time or cycles as a new time since new (TSN) / cycles since new (CSN) with a row given for each unique combination of the engine serial number and CSN / TSN. In an embodiment, the output data comprises an engine report containing data visualizations and key statistics related to the "state of health", for example operating status, of the given aircraft engine that are produced on demand or via a schedule. The statistics optionally include average aircraft engine running costs or at-risk aircraft engine categories (for example, engine model or version types). A term "risk" is defined as relating to engines which are expected to require maintenance action within a given amount of cycles. The risk is optionally determined using statistical models within the system, as aforementioned.

In an embodiment, the carrier data is optionally processed using one or more methods such as via a shared cloud drive or by access to a carrier's API. In an example embodiment, the aircraft engine maintenance system is designed to process batch data related to the aircraft engine and runs end-to-end at a frequency matching the delivery of this data. The frequency of delivering this data is optionally a daily or a weekly basis. The aircraft engine maintenance system optionally controls the maintenance of the aircraft engine using existing scheduling tools such as cron scripts or through a scheduling framework such as Apache Airflow®. While the system is designed to run in batches, the architecture of the aircraft engine maintenance system is optionally used for streaming the input data.

The aircraft engine maintenance system optionally functions as follows:

(i) Initially, a sensor arrangement checks whether or not new data has been exposed or becomes available. In an embodiment, the sensor arrangement optionally functions by checking the data created / date modified tag on the data transferred to a shared cloud drive on an incremental basis within a given delivery window or a time period.

(ii) When the sensor arrangement has detected that new data is available, the sensor arrangement optionally triggers a data ingestion process. At this stage, the raw data is archived and the archived data is optionally stored on a local server or a cloud storage drive. The raw data is optionally archived in a SQL database, logging filename, data type and associated metadata.

(iii) During the data ingestion stage, all carrier data are merged together into a normalized data/dataset, and sorted with a reference to an engine serial number and cycles since new or time since new. In an embodiment, following the data ingestion stage, the supplementary data is retrieved. The retrieved supplementary data optionally includes contaminant exposure records. Furthermore, information from the carrier provided schedules (e.g. specifically landing and departure timestamps and airports) optionally allows for the contaminant data to be queried through a service provider API. The supplementary data is optionally merged together with the normalized data. Such a merge is optionally as simple as the normalized data being sorted and the supplementary data being found in reference to the ESN and CSN/TSN. The merged data is optionally processed to provide features to a machine learning (ML) model which optionally includes cumulative sums, moving averages and other operations undertaken on time series data, for example as aforementioned. Pre-trained machine learning models are optionally used to classify a performance of the aircraft engine and a score relative to a "state of health " of the aircraft engine. In an embodiment, the classifications optionally predict whether or not a given aircraft engine is serviceable or unserviceable relative, wherein the classification are generated from the machine learning model.

In an embodiment, the scores optionally include an overall metric which describes the "state of health" of the given aircraft engine with respect to when a maintenance is required. The aircraft engine maintenance system optionally employs other predictive models that produce remaining useful cycles that are feasible for operation of the given aircraft engine. The aircraft engine maintenance system optionally forecasts predictions by extrapolating modelled scores. The aircraft engine maintenance system optionally employs a method of (for) producing forecasted predictions which use additional data such as upcoming schedule data for forecasting contaminant reports. The output from all of the data processing stages is optionally stored within the SQL database, for example as aforementioned. In an embodiment, the aircraft engine maintenance system optionally comprises a web user-interface (UI) or API that are used to provide access to this dataset. The system optionally provides key statistics and visualizations to a customer through the web UI or through scheduled engine reports.

In an embodiment, the aircraft engine maintenance system optionally provides alerts with reference to changes in running costs, recommended maintenance actions or operational actions to be undertaken by a pilot for the given aircraft engine. In an embodiment, the aircraft engine maintenance system optionally triggers the maintenance actions of the aircraft engine using a classification or machine learning model which has detected an unserviceable component in the aircraft engine. In an embodiment, the system optionally triggers maintenance alerts using an extrapolated forecast which predicts that the aircraft engine potentially needs to be taken "off wing" within a certain amount of cycles. The alerts are optionally provided to a user (e.g. a pilot) who is operating the aircraft engines. The alert is optionally triggered towards an end of lifetime of the aircraft engine with a suggestion to reduce strain on the aircraft engine during operation.

In an embodiment, when the data arrangement has triggered the ingestion process, the input data and the supplementary data is optionally collected from a delivery source and downloaded locally to a device. In an embodiment, a batch of the input data includes multiple data types, with a data type defined by the information contained. For example, if a carrier/aircraft engine supplies four datasets covering schedules, sensors, cycles and maintenance that would be considered four data types. Maintenance data optionally includes many data types rather than a single record with each action (e.g. wash, engine overhauls, and installations) supplied separately in addition to inspection reporting. The input data is optionally logged in a database with respect to metadata such as filename, date and data type.

In an embodiment, the local supplementary data is loaded in a memory and is cleaned (namely cleansed) to generate fixed organized data. Cleaning of the supplementary data optionally includes correcting spelling errors, reformatting data and filtering null or duplicated values. Cleaning is performed on each data type of the supplementary data in isolation. At a merge stage, the cleaned supplementary data is brought together into a normalized table. The supplementary data is typically provided with reference to ESN and either absolute time or CSN and absolute time. The supplementary data containing only timestamps is optionally compared to the supplementary data containing both CSN and time to provide a nearest cycle count for each data type which is optionally used to merge data types together. The merged data/dataset is then recorded in a SQL database. In an embodiment, the carrier optionally provides access to data relating to sensors (namely measurement data obtained from engine hardware), schedules, maintenance and cycles on a periodic basis. The aircraft engine maintenance system optionally detects when the input data has been delivered for processing it in a timely manner. In an embodiment, the contaminant exposure data is requested from a service provider to compliment the original input data. The input data is processed for modelling with one or more pre-trained models. One pre-trained model is for classifying whether or not an at-risk component of the aircraft engine is found serviceable during an inspection and another pre-trained model is for providing a score of the overall "state of health” of the aircraft engine with respect to a time until it should be taken off wing. In an embodiment, alerts are optionally triggered if the component of the aircraft engine is classified unserviceable or the extrapolated end of life prediction of the aircraft engine changes beyond an expected rate. In another embodiment, the alerts are optionally delivered through an email and a text to a customer/user device. Additionally, the customer/user optionally accesses a full dataset through a web API and views summary reports through scheduled reports delivered via an email.

According to an embodiment, the aircraft engine maintenance system comprises at least one storage medium that is coupled to the data processing arrangement and stores the input data and the supplementary data.

According to another embodiment, the aircraft engine maintenance system comprises at least one output interface that provides the data output representative of the "state of health", namely operating status, to a user.

According to yet another embodiment, the normalized data comprises a reference to a serial number of the given aircraft engine.

According to yet another embodiment, the input data comprises at least one of a carrier data, a contaminant exposure data, a temperature record, humidity records, and Automatic dependent surveillance - broadcast (ADS-B) positional data, wherein the carrier data comprises at least one of schedules, maintenance or cycles of the given aircraft engine.

According to yet another embodiment, the supplementary data comprises raw data, wherein the raw data comprises a logging filename, a data type and an associated metadata of the input data.

According to yet another embodiment, supplementary data comprises records of an amount of maintenance intervention associated with an aircraft or engine. By "amount of maintenance intervention associated with an aircraft or engine", there is meant thereby the total cost incurred in the maintenance of the aircraft so as to provide a fully serviceable aircraft flight when it is required. Such a cost is inclusive of the personnel cost and overhead cost required to maintain the aircraft or its engine or engines. Moreover, such a cost also includes logistics cost related to material management, as well as any cost incurred due to depreciation of spare parts.

According to yet another embodiment, the aircraft engine maintenance system cleans the supplementary data, for example as aforementioned, to obtain cleaned supplementary data by performing at least one of: correcting spelling errors of the supplementary data, identifying and correcting or filtering erroneous flight details, timestamps or cycle counts, reformatting the supplementary data and/or filtering null or duplicated values of the supplementary data. Such cleansing is very important to perform prior to performing data normalization, as aforementioned, (or as a part of data normalization) so that the "state of health" parameter is accurate representative of a condition of the aircraft engine.

According to yet another embodiment, the aircraft engine maintenance system extends, namely extrapolates, the supplementary data to obtain extended supplementary data by performing at least one of: (i) identifying changepoints in the data;

(ii) encoding unstructured, textual or classification data in a numerical representation ;

(iii) grouping similar events;

(iv) computing a count or accumulated value since an event;

(v) computing a count of the number of occasions a value has crossed a threshold since an event.

An event optionally includes at least one of:

(i) an engine new event;

(ii) an engine installed event;

(iii) an engine removed event;

(iv) an engine washed event;

(v) an engine serviced event;

(vi) a maintenance performed event;

(vii) an inspection performed event;

(viii) a changepoint detected event;

(ix) a fault identified event; and

(x) a bird strike event has occurred.

An accumulated value optionally includes the number of cycles since an event. The process of extending supplementary data optionally involves multiple stages of transformations, for example multiple iterative extrapolations of database data.

According to yet another embodiment, the aircraft engine maintenance system merges at least one of the cleaned supplementary data and extended supplementary data into the normalized data.

According to yet another embodiment, the numerical model is pre-trained for at least one of:

(i) classifying a performance of the given aircraft engine; and

(ii) scoring a health condition of the given aircraft engine. According to yet another embodiment, the output data includes a reference to changes in running costs, recommended maintenance actions, or operational actions to be undertaken by a user.

In overview, the present disclosure also provides a method for (of) operating an aircraft engine maintenance system, characterized in that the method includes: using at least one input interface of the aircraft engine maintenance system to receive data relating to a given aircraft engine from a sensor arrangement that is connected to the given aircraft engine and supplementary data from a database arrangement, wherein the supplementary data describes operational and maintenance information that characterizes the given aircraft engine; using a data processing arrangement of the aircraft engine maintenance system to receive the data related to the given aircraft engine as an input data and the supplementary data and using the input data and the supplementary data to generate normalized data, wherein the normalized data is generated by ingesting the input data and the supplementary data, wherein ingesting the input data and the supplementary data includes normalizing relative weightings of the input data and the supplementary data to generate the normalized data, using the data processing arrangement:

(ii) to derive from the numerical model at least one predicted aircraft health parameter for the given aircraft engine and to use or provide the derived at least one predicted aircraft health parameter as an input for scheduling a maintenance intervention for the given aircraft. According to an embodiment, in the method, the normalized data comprises a reference to a serial number of the given aircraft engine.

According to another embodiment, in the method, the input data comprises at least one of a carrier data, a contaminant exposure data, a temperature record, humidity data, and Automatic dependent surveillance - broadcast (ADS-B) positional data, wherein the carrier data comprises at least one of schedules, maintenance or cycles of the given aircraft engine on a periodic basis.

According to yet another embodiment, in the method, the supplementary data comprises raw data, wherein the raw data comprises a logging filename, a data type and an associated metadata of the input data.

According to yet another embodiment, in the method, the supplementary data comprises records of an amount of maintenance intervention associated with an aircraft or engine. According to yet another embodiment, the method comprises cleaning the supplementary data to obtain cleaned supplementary data by performing at least one of:

(i) correcting spelling errors of the supplementary data; and

(ii) identifying and correcting or filtering erroneous flight details, timestamps or cycle counts, reformatting the supplementary data or filtering null or duplicated values of the supplementary data.

According to yet another embodiment, the method comprises extending, for example using extrapolation, the supplementary data to obtain extended supplementary data by performing at least one of:

(i) identifying changepoints in the data;

(iii) grouping similar events; (iv) computing a count or accumulated value since an event; and

(v) computing a count of the number of occasions a value has crossed a threshold since an event. According to yet another embodiment, the method comprises merging into the normalized data at least one of: the cleaned supplementary data, the extended supplementary data.

According to yet another embodiment, the method comprises pre-training the statistical model for classifying a performance of the given aircraft engine and scoring a "state of health" condition of the given aircraft engine.

According to yet another embodiment, in the method, the output data includes a reference to changes in running costs, recommended maintenance actions, or operational actions to be undertaken by a user.

The advantages of the present method are thus identical to those disclosed above in connection with the aircraft engine maintenance system and the embodiments listed above in connection with the aircraft engine maintenance system apply mutatis mutandis to the method. The present disclosure provides a computer program product comprising instructions to cause the above system to carry out the above method.

The advantages of the present computer program product are thus identical to those disclosed above in connection with the aircraft engine maintenance system and the embodiments listed above in connection with the aircraft engine maintenance system apply mutatis mutandis to the computer program product.

Embodiments of the present disclosure beneficially proactively recommend a maintenance or a scheduling operation of the given aircraft engine in an automated way. Embodiments of the present disclosure beneficially generate a data output representative of a "state of health" (namely an operating status of) the given aircraft engine. Embodiments of the present disclosure are susceptible to improving the safety of the aircraft engine by scheduling the maintenance of the aircraft engine. Embodiments of the present disclosure potentially prevent damage to the aircraft engine or in-flight failures, and thereby increases the aircraft engine's lifetime. Embodiments of the present disclosure beneficially schedule maintenance interventions for the aircraft engine in response to the exposure of the aircraft engine with contaminants present in the atmosphere. Embodiments of the present disclosure beneficially schedule maintenance interventions based on an exposure rate including both a short-exposure, and a long-exposure of the aircraft engine to the contaminants. Embodiments of the present disclosure are capable of causing a reduction in emission of the carbon dioxide and other pollutants generated by the aircraft engine when in operation.

It will be appreciated that embodiments describe in the foregoing beneficially include an aircraft atmospheric contamination determination system. The embodiments beneficially employ an aircraft atmospheric contamination determination system, comprising:

(a) a computing arrangement including at least one input interface and at least one output interface, wherein :

(i) the computing arrangement executes when in operation a predictive aircraft health model, wherein the computing arrangement accesses when in operation a database of atmospheric contamination data defined as a function of location and time;

(ii) the at least one input interface receives flight data relating to a target aircraft to be evaluated and provides the flight data to the computing arrangement; and

(iii) the at least one output interface provides at least one predicted aircraft health parameter to a user; and (b) wherein, the computing arrangement uses the predictive aircraft health model and the inputs to:

(i) determine a contaminant exposure measure for a target aircraft by retrieving atmospheric contaminant data at a location and a time in proximity to at least one flight of the target aircraft; and

(ii) derive at least one predicted aircraft health parameter for the target aircraft based upon at least the target aircraft flight history and contaminant data as a function of location and time.

Optionally, in an aircraft atmospheric contamination determination system, the computing arrangement is trained by interrogating the database of contamination data to:

(i) obtain a set of training data by retrieving the atmospheric contaminant data at a location and a time in proximity to at least one flight in the aircraft flight data; and

(ii) provide the historical contaminant exposure and the aircraft service data as inputs for use in training the predictive aircraft health model.

The present system, when in operation, thus determines a contaminant exposure of the target aircraft, for example target engine, and predicts an aircraft health parameter, for example engine health parameter, of the target aircraft, for example target aircraft engine, for scheduling maintenance interventions. The present system thus predicts, when in operation, engine health or remaining useful life of the target aircraft, for example target aircraft engine. The present system thus schedules maintenance interventions for the target aircraft, for example target aircraft engine, based on the predicted aircraft health parameter, for example aircraft engine parameter. The present system potentially assists to prioritise the maintenance interventions based on the predicted engine health parameter.

In embodiments of the present disclosure, the flight data, engine service data, atmospheric contaminant data at a location and a given time, an aircraft flight history, an assignment of engines to aircraft and engine service data are obtained from a server. A user device is communicatively connected to the aircraft engine contamination determination system for receiving flight data relating to the aircraft engine to be evaluated, as required by a user. The system optionally generates, for example, for ease of processing, a table of the data for the training set defining at least one feature that is susceptible to being used to predict a target output of interest. For example, for sensor readings, a column of the table defines readings relating to a contaminant and rows of the table provide values for a plurality of flights. In an example embodiment, the system beneficially uses linear regression to calculate weights such that a weighted sum of the columns plus an optional offset weight provides a predictive engine health model. In another embodiment, the system beneficially uses a variant of regression such as "Least Squares Regression" or any suitable regression models to calculate weightings to provide a predictive engine health model. In yet another example embodiment, the generation of the predictive engine health model is beneficially generated using at least one of artificial intelligence (AI), Machine Learning or a neural network algorithm; such artificial intelligence (AI), Machine Learning and the neural network algorithm will be understood by a person skilled in the art of computer system design, and concerns computer software products whose one or more algorithms are able to adapt themselves in response to data being processed using the one or more algorithms; such adaptation is, for example, assisting by using training data. In an alternative embodiment, the weights or parameters of a predictive engine health model are optionally determined or adjusted manually by a person skilled in the art to provide a prediction of an engine health parameter. It will be appreciated in embodiments of the present disclosure that a predictive model of aircraft health is trained on measurement data, to define various parameters of the model, and then, thereafter, the predictive model is applied to data pertaining to the target aircraft to predict its health, for example its aircraft engine health; in other words, there are two distinct phases in operation of embodiments of the present disclosure. In an example embodiment, the computing arrangement includes a control module that is communicatively connected to a server via a communication network. The control module is connected to at least one input interface that receives, when in operation, flight data relating to a target aircraft engine to be evaluated and the at least one input interface provides the flight data to the control module. In one embodiment, the at least one input interface receives the flight data relating to the aircraft engine to be evaluated from a user device of a user. The control module is connected to at least one output interface that provides, when in operation, at least one predicted engine health parameter to the user through the user device. The user device optionally comprises a personal computer, a smartphone, a tablet computer, a laptop computer or an electronic notebook computer. The communication network is beneficially a wired network or a wireless network, or a combination of both a wired and wireless network. The server is optionally a tablet computer, a desktop computer, a personal computer or an electronic notebook computer. In an embodiment, the server is optionally a cloud service.

The server optionally comprises a contaminant database of contamination data as a function of spatial location and time, and an historical database of historical data relating to a plurality of aircraft engines. In one example embodiment, the server comprises a database of assignment of engines to aircraft and engine service data.

The system optionally determines historic contaminant exposure by analyzing a plurality of historical trajectories of a plurality of aircraft engines associated to the set of training data as a function of spatial location and time, and estimates contaminant concentration of the plurality of aircraft engines for each trajectory using an atmospheric model. In an embodiment, the contaminant exposure is optionally estimated using a trajectory or trajectory model.

The system optionally determines at least one of: an average estimate of exposure, a total estimate of exposure and a cumulative estimate of exposure to each contaminant type, per cycle and/or per time interval, and incorporates one or more of these estimates in the engine health model. The system beneficially determines estimates of contaminant exposure by flight phase and incorporates the estimate for at least one flight phase in the engine health model. The system optionally determines values for aggregates of types of contaminant, for example totaling dust of different particle sizes, with an optional weighting of different parameters. The system optionally estimates a cost associated with actual engine usage to determine a cost of contaminant exposure of the aircraft engine.

The system optionally estimates an expected contamination exposure of the aircraft engine over a period of time by estimating a probability, a probability density function, or multiple parameters including expected mean and standard deviation of estimated contaminant exposure. The system optionally determines either of, or both of, the mean and standard deviation of the contaminant exposure or density for at least one of a time within year and time of day. For example, the historical data sets are grouped by calendar month and hour of the day, thus providing an estimate of the mean and standard deviation of the engine health parameter that varies from month to month and over the course of a day. The system optionally determines a mean or standard deviation, or both a mean and a standard deviation, of future contaminant exposure or density by using the machine learning or regression methods, such as linear regression, neural network, decision tree, decision forest or gradient boosted decision tree, where time within a year and/or a time of day are included as regression parameters. In an example embodiment, the system determines a probability that the contaminant is present, rather than the average concentration for certain types of contaminant, such as ice crystals.

The system optionally determines at least one contaminant parameter for the set of training data and at least one engine health parameter for each aircraft engine associated with the set of training data. The system optionally determines the at least one target health parameter based on the at least one contaminant parameter, using techniques known in art of statistics, regression and machine learning.

The input interface further receives one or more input variables that are utilised by a processor of the computing arrangement in the engine health model. The one or more input variables include one or more of: engine or aircraft control, loading or operational parameters, engine or aircraft model and type information, sensor data, weather parameters, temperatures, speeds, altitudes, mass flow rates, fuel flow rates, vibration measures, inspection results or wear estimates. In one embodiment, the_target input variable of the aircraft engine includes one or more of accumulated cycles or hours, count of excursions outside an operating threshold, for example "engine service data". The one or more input variables may be determined per flight, flight phase, or at a finer resolution such as every minute. In an example embodiment, transformation and feature engineering methods known in the art of machine learning may be employed to determine at least one of the input variables or output variables. The feature engineering methods known in the art of machine learning may include calculating a moving average, moving standard deviation, L-norm, cumulative value, non-linear transform of a variable, product or weighted combination of variables.

The predictions are optionally made using any known machine learning or regression method, such as linear regression, neural network, decision tree, decision forest or gradient boosted decision tree, radial basis function, support vector machine, Gaussian process, principal component analysis. In an embodiment, an expected lifetime of the aircraft engine beneficially determined from an output of the predictive engine health model . The expected lifetime of the aircraft engine is optionally used to plan when the engine requires replacement or removal for a shop visit. The predictive engine health model optionally additionally outputs a confidence interval on its remaining useful life estimate, so that an aircraft engine is beneficially prepared for when the end of life is likely to occur, for example, to ensure that a spare aircraft is available. The system beneficially calculates the remaining useful life (RUL) of the aircraft engine or the aircraft using a table, or example to assist processing or understand of data. The table includes recorded data of the aircraft engine and the last row indicates a target life when the aircraft engine is ending. For example, the row corresponding to when an aircraft engine is, or is due to be, removed for service shows with a zero (0) and the second last row shows with a one (1), and hence the third last row shows a two (2). The system beneficially uses the RUL to provide servicing and schedule maintenance accordingly when an aircraft has an aircraft engine at 50% health and optionally swaps the aircraft engine if needed or for those events to coincide with the end of life (0) of the aircraft engine. For example, an aircraft engine El is presently in poor condition and an aircraft engine E2 is in good condition, at the present time engine El has a RUL value that is lower than the RUL value of engine E2, and at the end of life both the aircraft engines El, E2 would have approximately the same RUL measure such as zero.

In one embodiment, the system determines the engine health index using a transform. For example, when all aircraft engines are new it is considered as hundred (100). The end of life of an aircraft engine is assigned an index of zero (0). For training of the model, a target engine health parameter is calculated for each training engine by interpolating between the new value (such as 100) and the end of life value (such as 0) according to the cycles since new of the engine. Where CSN(flight, engine) is the cycles since new of the engine in each flight in the training set, and CEOL(engine) is cycles at end of life, that is the value of CSN(engine) where the engine is at, or is estimated to be at, an end of remaining useful life or requires a major service, the target engine health index EHI is beneficially determined using the following or an equivalent formula :

EHI (flight, engine) = 100 * [1 - ^CSN(flight _' engine) j c OL(engine)]

In another example embodiment, such calculation is beneficially also, or alternatively, based on the hours since new of each engine. In an embodiment, the determination of health Index requires an extrapolation step to be performed to determine a remaining useful life of the aircraft engine.

A particular benefit of embodiments of the present disclosure is that it can be used to identify temporal rates of change, patterns or deviations from normal performance that are potentially indicative of failures, including at least one of: engine sensor values unexpected given the short-term and long-term history of the engine, and temporal rates of change of a sensor value or engine health parameter that potentially merit inspection or intervention to avoid adverse impact to the fuel usage or life of the engine. For example, a machine learning model is trained to predict a subsequent value of an engine health measure or sensor value, given features representative of other parameters of the subsequent or a previous flight and the short-term and long-term history of the engine, its sensor values, carrier data or contaminant exposure. A subsequent value is, for example, an engine's exhaust gas temperature margin (EGTM) in a subsequent flight. A prediction error is then calculated as the difference between the subsequent value estimated using the model, and the actual value. A typical distribution, or a parameter such as a standard deviation, of prediction errors is determinable by reference to a training or test set of engines or flights. At least one threshold or likelihood determined with reference to a probability distribution function, is then beneficially applied to determine whether the subsequent value is likely to be within the normal range.

The system is optionally used to schedule maintenance interventions for the aircraft engine in response to the predicted engine health parameter. In an embodiment, an engine is considered to be due for maintenance when an engine health parameter computed for the engine falls below a threshold. The predicted remaining useful life or an extrapolation of engine health index, or both, determines the number of cycles before such maintenance is required. The system optionally schedules maintenance interventions including both short- term interventions, such as engine washes to clean out dust, and long-term interventions such as major maintenances or shop visits. Engine washing is described above.

The system beneficially adjusts one or more quantities indicative of engine health to consider maintenance events. In an example embodiment, when an engine has a major service, its engine health parameter is reset to a value representative of an engine that has undergone such a major service, and for subsequent modelling counts of cycles since such major service, and accumulated input variables, are reset to zero or some other representative value. On account of engine performance being improved after such a major, corresponding savings in carbon dioxide emissions are beneficially attributed to "carbon offsets" that can assist towards paying for costs of the major service.

The system further prioritises maintenance for the aircraft engine according to the determined engine health and/or lifetime. In an example embodiment, an engine that has a value of an engine health parameter that is below the expected value of such a parameter, given the same cycle count, would be prioritised for maintenance ahead of an engine with above expected engine health parameter.

According to an embodiment, the at least one input interface receives historical data relating to the target aircraft engine to be analysed. According to another embodiment, the system further comprises at least one storage medium coupled directly or indirectly to the control module and containing a database of historical data relating to a plurality of aircraft engines that the system may be used to evaluate.

According to yet another embodiment, the predictive engine health model is trained by machine learning using the set of training data comprising historical data sets. In an embodiment, the historical data sets comprise aircraft flight data, atmospheric contaminant data over location and time and engine service data. According to yet another embodiment, the predicted engine health parameter is selected from a group comprising a cycle count, hour count, remaining cycles, remaining hours or health index.

According to yet another embodiment, the processor further calculates and outputs at least one of a standard deviation, variance or confidence interval for the predicted health parameter.

According to yet another embodiment, the predictive engine health model is trained to determine:

(i) at least one contaminant parameter for the set of training data; and (ii) at least one engine health parameter for each engine or aircraft associated with the set of training data.

Once trained, the predictive engine health model is beneficially used to calculate the at least one aircraft health parameters, for example engine health parameter, corresponding to the at least one contaminant parameter for an engine, aircraft or flight not present in the training set.

According to yet another embodiment, the flight data includes flight trajectory information that is selected from at least one of flight phase, estimation of airspeed or engine mass flow. In an embodiment, the airspeed or engine mass flow is estimated to weigh the at least one contaminant and to provide estimates of rate, density or mass of contaminants passing through the aircraft engine.

According to yet another embodiment, the input interface further receives one or more input variables that are utilized by the processor in the engine health model, wherein the one or more input variables include one or more of: engine or aircraft control, loading or operational parameters, engine or aircraft model and type information, sensor data, weather parameters, temperatures, speeds, altitudes, mass flow rates, fuel flow rates, vibration measures, inspection results or wear estimates. According to yet another embodiment, the system comprises a distributed computing system, wherein the control module and at least one data storage medium are centralised, and at least one input interface and at least one output interface are provided at local computing nodes. According to yet another embodiment, the historic contaminant exposure is determined by:

(i) analyzing a plurality of historical trajectories of a plurality of aircraft engines associated the set of training data over location and time; and

(ii) estimating contaminant concentration of the plurality of aircraft engines for each trajectory using an atmospheric model.

In the forgoing embodiments, there are beneficially included an aircraft atmospheric contamination determination system, comprising :

(iii) the at least one output interface provides at least one predicted aircraft health parameter to a user; and (b) wherein, the computing arrangement uses the predictive aircraft health model and the inputs to: (i) determine a contaminant exposure measure for a target aircraft by retrieving atmospheric contaminant data at a location and a time in proximity to at least one flight of the target aircraft; and

(ii) derive at least one predicted aircraft health parameter for the target aircraft based upon at least the target aircraft flight history and contaminant data as a function of location and time; and

(iii) provide an alert or schedule a maintenance intervention for the

target aircraft based on the predicted aircraft health parameter. Optionally, in the aircraft atmospheric contamination determination system, the computing arrangement is trained by interrogating the database of contamination data to

Optionally, in the aircraft atmospheric contamination determination system, the system further schedules:

(i) a maintenance intervention when an engine contamination measure passes or forecasts to pass a threshold; and

(ii) a maintenance intervention when an engine health measure passes or forecasts to pass a threshold,

wherein the maintenance intervention comprises at least one of engine wash, engine inspection, engine removal or engine shop visit.

The system optionally generates a notification or provide alerts to the user device through the at least one output interface based on the predicted engine health parameter. According to an embodiment, a processor of the computing arrangement uses the engine health model to provide estimates of future contamination exposure of the aircraft engine over time by estimating a probability, a probability density function, or multiple parameters including the expected mean and standard deviation of estimated contaminant exposure and the contamination determination system utilises the estimates to schedule interventions.

Embodiments of the disclosure described in the foregoing optionally employ an asset maintenance system, comprising a computing arrangement including at least one input interface and at least one output interface, wherein the computing arrangement, when in operation :

(i) accesses a database of atmospheric contamination data defined as a function of location and time;

(ii) receives flight data relating to an asset to be evaluated;

(iii) determines a flight trajectory specifying at least one location of the asset over time for each of a plurality of flights in the asset flight data ;

(iv) retrieves the atmospheric contaminant data at a location and a time in proximity to each of a plurality of flight trajectories of the at least one flight;

(v) determines a contaminant exposure measure for the asset; and

(vi) outputs a recommended maintenance intervention to the asset according to the value of at the contaminant exposure measure of the asset.

Optionally, in the asset maintenance system, the system schedules a maintenance intervention when a contaminant exposure measure passes or forecasts to pass a threshold .

Optionally, in the asset maintenance system, wherein the maintenance intervention comprises at least one of engine wash, engine inspection, engine removal or engine shop visit. Optionally, the asset comprises any of an aircraft body, a flight surface, an engine, an instrument, s sensor, a control equipment or a landing gear, or a component of the foregoing.

Embodiments of the disclosure as described in forgoing optionally employs a method for (of) determining aircraft health using a computing arrangement coupled to a sensor arrangement, wherein the method comprises:

(a) providing a computing arrangement including at least one input interface and at least one output interface, wherein the method includes:

(i) arranging for the computing arrangement to execute when in operation a predictive aircraft health model, wherein the computing arrangement accesses when in operation a database of atmospheric contamination data defined as a function of location and time;

(ii) receiving at the at least one input interface flight data relating to a target aircraft to be evaluated and providing the flight data to the computing arrangement; and

(iii) providing to the at least one output interface at least one predicted aircraft health parameter to a user; and

(b) arranging for the computing arrangement to use the predictive engine health model and the inputs to:

(ii) derive at least one predicted aircraft health parameter for the target aircraft based upon at least the target aircraft flight history and contaminant data as a function of location and time. Optionally, training the computing arrangement is achieved by interrogating the database of contamination data to: (i) obtain a set of training data by retrieving the atmospheric contaminant data at a location and a time in proximity to at least one flight in the aircraft flight data; and

According to an embodiment, the method further comprises training the engine health model using machine learning to provide a predictive engine health model, wherein the predictive engine health model receives at least one contaminant parameter and to output at least one engine health parameter.

According to another embodiment, the step of training the engine health model utilises the aircraft flight data and the atmospheric contaminant data as inputs and engine health data derived from service data to provide at least one health parameter. Such training is undertaken before the engine health model is subsequently used to predict aircraft health of a given target aircraft.

According to yet another embodiment, the engine health parameter comprises at least one of an engine health index or a useful remaining life prediction.

According to yet another embodiment, the machine learning comprises any one of: linear regression, neural network, decision tree, decision forest or gradient boosted decision tree, radial basis function, support vector machine, Gaussian process, or principal component analysis.

According to yet another embodiment, the machine learning includes at least one of time within a year or time of day as regression parameters. According to yet another embodiment, the predictive model provides a standard deviation, variance or confidence interval for the predicted engine health parameter. According to yet another embodiment, the atmospheric contaminant data over location and time comprises an atmospheric model of the concentration or mixing ratio of contaminants at a plurality of locations and times.

According to yet another embodiment, the contaminants include a plurality of contaminants selected from: dust, organic particles, volcanic ash, salt, sulphur dioxide and sulphate ions, or any combination thereof.

According to yet another embodiment, the atmospheric model derives estimates of at least one of the mixing ratio or the concentration of contaminants at the plurality of locations and times and provides an expected mass of the respective contaminant per unit mass of air at a particular location, height measure and time.

According to yet another embodiment, the atmospheric model provides at least one of estimated average or standard deviation of contaminants for each location by time-of-day, by time-of-year, or by both time-of-day and time-of-year. According to yet another embodiment, the method further comprises scheduling maintenance interventions for the aircraft engine in response to the predicted engine health parameter.

According to yet another embodiment, the aircraft flight data comprises flight phase and estimation of at least one of airspeed or engine mass flow and associated location, time and date information.

According to yet another embodiment, using the predictive model to predict an engine health parameter comprises a Monte Carlo model. Monte Carlo model is especially useful for determining boundary limits for a given mathematical model by employing a plurality of computations with stochastic variation to map a solution space pertaining the Monte Carlo model.

According to yet another embodiment, the predictive engine health model includes one or more further input variables selected from : engine or aircraft control, loading or operational parameters, engine or aircraft model and type information, sensor data, weather parameters, temperatures, speeds, altitudes, mass flow rates, fuel flow rates, vibration measures, inspection results or wear estimates. The advantages of the present method are thus identical to those disclosed above in connection with the present system and the embodiments listed above in connection with the system apply mutatis mutandis to the method.

It will be appreciated that the aforesaid present method is not merely a "method of doing a mental act", but has a technical effect in that the method functions as a form of technical control using machine learning of a technical artificially intelligent system. The method involves building an artificially intelligent machine learning model and using the machine learning model to solve a technical problem of predicting an aircraft health parameter of a target aircraft, for example for predicting an engine health parameter of an engine of a target aircraft, and scheduling maintenance interventions for the target aircraft, for example for the target aircraft engine.

The present disclosure also provides a computer program product that executable on the aforementioned computing arrangement, wherein the computer program product comprises instructions to cause the above system to carry out the aforesaid method. The advantages of the present computer program product are thus identical to those disclosed above in connection with the present system and the embodiments listed above in connection with the present system apply mutatis mutandis to the computer program product.

In one example embodiment, the system for scheduling aircraft operations is provided, wherein the system comprises: a control module, comprising a processor; and a memory, containing executable instructions for generating a predictive engine contamination model for use by the processor; and at least one storage medium coupled directly or indirectly to the control module and containing a database of contamination data over location and time; a route input interface configured to receive schedule information related to a plurality of aircraft flights; a fleet input interface configured to receive aircraft flight data and aircraft engine data relating to a target aircraft fleet being scheduled; and wherein the control module configured to receive the schedule information from the route input interface and the aircraft flight data and the aircraft engine data from the fleet input interface and the processor interrogates the database of contamination data to provide inputs for use in the predictive engine contamination model, the processor using the predictive engine contamination model and inputs to derive: an estimated historical contamination for an aircraft or each aircraft engine to at least one contaminant; an estimated expected contaminant exposure for the aircraft or each aircraft engine; and the control module is further configured to identify at least one aircraft engine with higher estimated historical contamination or with lower estimated historical contamination; identify at least one route with lower expected exposure or highest expected exposure to the at least one contaminant; and perform recommendations on route allocation by assigning an aircraft with the higher estimated historical contamination being scheduled to at least one route with lower expected exposure to ensure that total exposure to contamination, based upon the estimated historical contamination and expected contaminant exposure, is reduced or maintained within a required range; the system further comprising an output interface configured to receive recommended route allocation information from the control module.

The system thus helps to schedule aircraft operations by generating a model for estimating aircraft engine contamination based upon known historical data sets including aircraft flight data, atmospheric contaminant data over location and time, and engine service data. The system helps to schedule aircraft operations based on the estimation of contamination exposure, thereby managing the cost of operating aircraft or engines. The system optionally associates a fee, a charge or a contractual payment to estimated contaminants exposure or to provide recommendations derived from contaminants exposure, or both; for example, carbon dioxide emission savings achieved by performing engine maintenance is beneficially recorded as a "carbon offset" and subject to financial reward payment that assists to improve aircraft maintenance and operating safety. The system helps to adjust the service of aircraft engines to increase overall engine cycles between shop visits. This practice beneficially reduces an engine vendor's or owner's maintenance costs (as well as material resources that are required to be employed), and part of this saving is beneficially shared with the airline or maintenance organization to compensate for the cost of making the adjustment and to motivate them to adopt the practice.

In one embodiment, the system optionally beneficially assigns aircraft with lower estimated historical contamination to routes with higher expected exposure to the at least one contaminant during route allocation.

In another embodiment, the expected contaminant exposure of the aircraft or the aircraft engine is estimated by determining an exposure index which is a measure of a contaminant exposure or an amount of contaminant material that passes through the aircraft or the aircraft engine in the aircraft fleet based on its trajectory. The control module allocates the exposure index to each aircraft engine in the fleet based upon the estimated historic contamination.

In yet another embodiment, the system further comprises a maintenance interface that is communicated with the control module to input data for adjusting the estimated historical contamination for the aircraft or each aircraft engine as a result of maintenance interventions and to output alerts for required interventions.

In yet another embodiment, the historic contamination of the aircraft is estimated by determining at least one of an average contamination exposure of the aircraft or each aircraft engine per cycle, an average contamination exposure of the aircraft or each aircraft engine per time interval or a cumulative contamination exposure of the aircraft or each aircraft engine and determining historic contamination of at least one flight phase of the aircraft or each aircraft engine from the database of contamination data.

Embodiments of the present disclosure concern aircraft flight planning as a function of measured atmospheric contaminants that are potentially encountered that potentially adversely affect aircraft operating performance and reliability, for example aircraft engine operating performance and reliability. In one embodiment, the system for aircraft flight planning is provided, wherein the system comprises: a first storage medium including a measure of at least one atmospheric contaminant with respect to a location, an altitude or a pressure, and a time; a control module in communication with the first storage medium, the control module having a processor and a memory containing executable instructions to provide the indications related to an estimated contamination risk to at least one aircraft at selected locations and altitudes or pressures; an input interface in communication with the control module for receiving at least one aircraft flight plan data including at least one of a time, a pressure or an altitude, a trajectory and a location representing at least one aircraft flight; and an output interface in communication with the control module for providing indications related to the estimated atmospheric contamination risk to the at least one aircraft, wherein the control module receives at least one aircraft flight plan data from the input interface; determines an estimated contamination risk using the measure of the at least one atmospheric contaminant for the at least one aircraft flight based upon a location, an altitude or pressure, a trajectory and a time information extracted from the at least one aircraft flight plan data; and provides a resultant indication related to the estimated contamination risk of the at least one aircraft to the output interface. In one embodiment, the system thus reduces or mitigates a risk of contaminant exposure to the aircraft engine. The system thus allows operators to determine, before an aircraft takes off, whether or not it is appropriate to alter the service (for example, to cancel, delay or re-route a flight) to avoid a contamination hazard. The system beneficially enables operators or automated systems to determine possible changes to the aircraft flight path, height or speed when an aircraft is in flight, to avoid or reduce the risk of encountering a hazard. The system beneficially runs Weather Prediction Model (WRF) using boundary conditions (e.g. dust boundary data) for predicting estimated contamination, and the output is an estimate of weather, water, and contaminants by locations, altitudes and time. The system optionally produces a three-dimensional (3D) matrix for each contaminant separately and store or combine them to obtain four dimensional (4D) weather model using time. In one embodiment, the system calculates at least one modified aircraft flight plan for the at least one aircraft flight, an estimated contamination risk for the modified aircraft flight plan being generated for comparison with the estimated contamination for at least one aircraft flight plan.

Embodiments of the present disclosure beneficially incorporate an effect of a wide variety of contaminants and beneficially avoids a need for aircraft sensors to measure exposure of contaminants and avoids the weight, drag, and cost of such sensors. Embodiments of the present disclosure beneficially monitor engine condition with greater accuracy. Embodiments of the present disclosure beneficially monitor the substantial variations in environmental contaminant exposure over time and as aircraft operate different routes, to improve accuracy of engine condition modelling. Embodiments of the present disclosure beneficially enable the aircraft engine condition to be modeled with improved accuracy and only limited access, or no access, to measurement data from an engine. Embodiments of the present disclosure beneficially recommend or schedule maintenance events that optimise aircraft engine life, including both short-term interventions such as engine washes to clean out dust, and long-term interventions such as major maintenances or shop visits, for example as aforementioned with reference to FIGs. 1 to 5. Embodiments of the present disclosure beneficially recommend interventions such as engine washes when they are most needed (e.g. immediately after contaminant exposure) rather than when the results of damage of the aircraft engine are observed by sensors. Embodiments of the present disclosure beneficially improve the accuracy of predictions of when maintenance is due by modeling the effect of corrosive contaminants such as salt and sulphate ion that are not typically sensed. Embodiments of the present disclosure beneficially analyze variations by time of day and year, and long-term trends, and enables airlines to make changes such as scheduling take-offs at times of day when contaminants are lower to reduce the impact or costs of contaminant exposure. Embodiments of the present disclosure beneficially enable a user to resell the contaminant data to engine manufacturers and eventually selling into the Airlines in a data as a service model (DaaS).

DETAILED DESCRIPTION OF THE DRAWINGS

FIG. 1 is a schematic illustration of an aircraft engine maintenance system 100 according to an embodiment of the present disclosure. The system 100 comprises component parts including a data processing arrangement 102, an input interface 104 and a sensor arrangement 106 and a database arrangement 108; these component parts mutually interact together in a manner as described in the foregoing; optionally, data encryption is employed to protect an integrity of data employed in the aircraft engine maintenance system, when in operation, to improve a reliability of the aircraft engine maintenance system 100. The data processing arrangement 102 optionally comprises a processor and a database.

FIG. 2 is an illustration of a flow diagram including steps of an operation of an aircraft engine maintenance system in accordance with an embodiment of the present disclosure. At a step 202, the carrier data is optionally exposed through one or more methods such as a shared cloud drive or by access to a carrier's API. At a step 204, a sensor arrangement checks whether or not new data has been exposed. At a step 206, when the sensor arrangement has detected that new data is available, the sensor arrangement optionally triggers a data ingestion process. At a step 208, the aircraft engine maintenance system processes batch data relating to the aircraft engine and runs an end-to-end check at a frequency matching the delivery of this data. At a step 210, raw data generated from the ingestion process is archived and the archived data is optionally stored on a local server or a cloud storage drive. At a step 212, following the data ingestion process, supplementary data is retrieved. The retrieved supplementary data optionally includes contaminant exposure records. At a step 214, the input and the supplementary data is modelled. The supplementary data is optionally merged together with the normalized data. The merged data is optionally processed to provide features to a machine learning model which optionally include cumulative sums, moving averages and other operations undertaken on time series data, for example as described in the foregoing. Pre-trained machine learning models are optionally used to classify a performance of the aircraft engine and a score relative to a "state of health” of the aircraft engine. In an embodiment, the classifications optionally predict whether or not an aircraft engine is serviceable, as determined using the machine learning model. At a step 216, the output from all of the data processing stages is optionally stored within a SQL database. At a step 218, the aircraft engine maintenance system optionally comprises a web user-interface (UI) or API that is used to provide access to this dataset. At a step 220, the aircraft engine maintenance system optionally provides alerts in reference to changes in running costs, recommended maintenance actions or operational actions to be undertaken by a pilot of the aircraft engine.

FIG. 3 is an illustration of a flow diagram of a data ingestion process in accordance with an embodiment of the present disclosure. At a step 302, the carrier data is optionally exposed through one or more methods such as a shared cloud drive or by access to a carrier's API. At a step 304, a sensor arrangement checks whether or not new data has been exposed (namely has become available). At a step 306, when the sensor arrangement has detected that new data is available, the sensor arrangement optionally triggers a data ingestion process and input data and supplementary data is optionally collected from a delivery source and downloaded locally to a device. At a step 308, the aircraft engine maintenance system processes batch data relating to the aircraft engine and runs an end-to-end at a frequency matching the delivery of this data. At a step 310, the raw data is archived and the archived data is optionally stored on a local server or a cloud storage drive. At a step 312, the supplementary data is cleaned (namely cleansed) to obtain cleaned supplementary data by performing at least one of: correcting spelling errors of the supplementary data, identifying and correcting or filtering erroneous flight details, timestamps or cycle counts, reformatting the supplementary data or filtering null and/or duplicated values of the supplementary data. At a step 314, the cleaned supplementary data is merged into the normalized data. At a step 316, the merged data/dataset is then recorded in an engine database.

FIG. 4 is an illustration of a flow diagram pertaining to a method for (of) operating the aircraft engine maintenance system 100 in accordance with an embodiment of the present disclosure. At a step 402, data relating to a given aircraft engine as an input data and supplementary data are received. At a step 404, the input data and the supplementary data are used to generate normalized data by ingesting the input data and the supplementary data. At a step 406, the normalized data is used to determine an operating status of the given aircraft engine by applying the normalized data to a statistical model of the given aircraft engine, for example as described in the foregoing. At a step 408, a data output representative of the "state of health" (namely operating status) of the given aircraft engine is generated.

FIG. 5 is an illustration of an exploded view of a distributed computing system or cloud computing implementation in accordance with an embodiment of the present disclosure. The exploded view comprises an input interface 502, a control module that comprises a processor 504, a memory 506 and a non- volatile storage 508, processing instructions 510, a shared/ distributed storage 512, a server that comprises a server processor 514, a server memory 516 and a server non-volatile storage 518 and an output interface 520. The function of the server processor 714, the server memory 516 and the server non-volatile storage 518 are thus identical to the processor 504, the memory 506 and the non-volatile storage 508 respectively. The functions of these parts are as has been described above.

FIG. 6 is a schematic illustration of a system according to an embodiment of the present disclosure. The system comprises a computing arrangement, wherein the computing arrangement includes elements including a control module 1102 that includes a processor 1104 and a memory 1106, an input interface 1108, an output interface 1110, a communication network 1112, a server 1114, a contamination database 1116 of contamination data over location and time and a historical database 1118 of historical data. The functions of these parts are as has been described above.

FIG. 7 is a flowchart illustrating steps of a method for (of) training an engine health model to predict an engine health parameter of an aircraft engine according to an embodiment of the present disclosure. Such training is performed and concluded before the model is subsequently applied to data, for example measured sensor data, to provide predictions. At a step 1202 of the method of training, one or more input variables that include engine or aircraft control, loading or operational parameters, engine or aircraft model and type information, sensor data, weather parameters, temperatures, speeds, altitudes, mass flow rates, fuel flow rates, vibration measures, inspection results or wear estimates, are obtained. At a step 1204 of the method of training, atmospheric contaminant data over location and time are obtained from a contamination database. At a step 1206 of the method of training, aircraft flight history is obtained from a historical database. At a step 1208 of the method of training, assignment data of aircraft engines and engine service data are obtained from the historical database. At a step 1210 of the method of training, historical aircraft trajectories are determined from the historical database. At a step 1212 of the method of training, contaminant exposure measure for a target aircraft engine is determined by retrieving atmospheric contaminant data at a location and a time in proximity to at least one target aircraft engine; such real-time data acquisition requires use of one or more sensors for sensing physical engine characteristics and engine environs. At a step 1214 of the method of training, engine run time, cycles and hours are determined from the assignment data of engines to the aircraft, the engine service data, and the aircraft flight history. At a step 1216 of the method of training, inputs such as the historical contaminant exposure, the engine service data and the one or more input variables are provided to an engine health model. At a step 1218 of the method of training, engine health and engine life data are derived, namely is computed, from maintenance records. At a step 1220 of the method of training, output variables are selected for the engine. At a step 1222 of the method of training, the engine health model is trained to provide a predictive model using machine learning. At a step 1224 of the method of training, an engine health parameter of the target aircraft engine is predicted, namely is computed, and the effect of contaminants is incorporated using the predictive model.

FIG. 8 is a flowchart illustrating steps of a method for (of) predicting an engine health parameter of the engine or the aircraft using a predictive engine health model in accordance with an embodiment of the present disclosure; in other words, FIG. 7 concerns training the model, whereas FIG. 8 concerns subsequently applying the model. At a step 1302 of the method of predicting, one or more input variables are obtained. At a step 1304 of the method of predicting, atmospheric contaminant data over location and time are obtained from a contamination database of contamination data over location and time. At a step 1306 of the method of predicting, aircraft flight history is obtained from a historical database of historical data. At a step 1308 of the method of predicting, assignment data of engines to aircrafts and engine service data are obtained. At a step 1310 of the method of predicting, historic aircraft trajectories are determined from the historical database. At a step 1312 of the method of predicting, contaminant exposure measure for a target aircraft engine is determined by retrieving atmospheric contaminant data at a location and a time in proximity to at least one flight of the target aircraft engine; contaminant exposure measure is determined, for example, by employing a sensor arrangement including one or more sensors. At a step 1314 of the method of predicting, engine run time, cycles and hours are determined, namely are computed, from the assignment of engines to the aircraft and the engine service data and the aircraft flight history. At a step 1316 of the method of predicting, inputs such as the historical contaminant exposure and the engine service data and the one or more input variables are provided to an engine health model. At a step 1318 of the method of predicting, an engine health parameter of the target aircraft engine is predicted, namely is computed, and an effect of contaminants are incorporated using the predictive model. At a step 1320 of the method of predicting, aircraft health is predicted using the predictive model, for example engine health and/or remaining useful life are predicted using the predictive model.

FIG. 9 is a graphical illustration of a predicted engine health in an ordinate Y- axis plotted against a number of cycles since last shop visit in an abscissa X-axis to predict an expected remaining useful life in accordance with an embodiment of the present disclosure. An expected end of life of an aircraft engine are beneficially determined from an output of the predictive engine health model. The expected end of life of an aircraft engine is beneficially used to plan when the aircraft engine requires replacement or removal or a shop visit. In an example embodiment, a life of the aircraft, for example aircraft engine, is reduced from a last service / shop visit if the environmental contaminant exposure affects negatively the aircraft, for example aircraft engine. The life of the aircraft, for example aircraft engine, is extended as expected to be at the end of life of the aircraft, for example aircraft engine, by predicting, namely computing, an engine health parameter using the predictive engine health model. For example, a number of cycles since last shop visit is increased from 5000 to 6000 when the engine health parameter of at least one aircraft engine is predicted. In an embodiment, an expected remaining useful life in cycles for an aircraft engine is automatically determined, namely is computed, from the last shop visit or service. The engine health index is calculated on a scale of 0 to 100. In an embodiment, a value of 100 in the scale indicates a new engine with an ideal performance and a value of 0 indicates that the aircraft engine requires a shop visit.

FIG. 10 is a graphical illustration of a predicted engine health in an ordinate Y- axis plotted against a number of cycles since last shop visit in an abscissa X-axis to predict an expected lifespan of the engine for different levels of exposure to contaminants in accordance with an embodiment of the present disclosure. The graphical illustration demonstrates an effect of an intervention on an aircraft engine that had previously been subjected to high contaminant exposure is indicated. An expected lifespan of the aircraft engine is reduced if a current exposure continues (for example, a lifespan of the of the aircraft engine may be 4000 cycles when the engine is subjected to high contaminant exposure). The expected lifespan of the aircraft engine increases if a service is modified or a maintenance practice is improved to reduce future contaminant exposure and optimise maintenance to reduce effects of contaminants (for example, the lifespan of the engine or the aircraft is increased from 4000 cycles to 5000 cycles). The graphical illustration further elucidates that the expected lifespan of an aircraft engine is increased when the aircraft engine is subject to a low contaminant exposure.

FIG. 11 is a graphical illustration of a flight trajectory model of flight phases and contaminant density at the engine at the flight phases in accordance with an embodiment of the present disclosure. The graphical illustration elucidates that the distance is plotted in an abscissa X-axis and the airspeed is plotted in an ordinate Y-axis. The graphical illustration further elucidates that the distance is plotted in the X-axis and the height is plotted in the Y-axis. The graphical illustration elucidates that the airspeed at different flight phase includes take- off, climb, cruise, descent and landing of a flight and the height of the flight which is varied at departure, cruise and arrival of the flight. A control module beneficially determines, namely computes, an exposure index which is a measure of a contaminant exposure or an amount of contaminant material that passes through the aircraft engine in the aircraft fleet based on its flight trajectory. The total contaminant exposure during flight is obtained using integration of contaminant density or mass flow at the aircraft engine.

FIGs. 12A to 12B are graphical illustrations of an integrated method pertaining to computing a cost awareness (CA) and a risk mitigation (RM) for predicting a risk of high-altitude ice (HAI) being encountered on an aircraft route at different periods from a departure time, in accordance with an embodiment of the present disclosure. The graphical illustrations elucidate forecasting, nowcasting and post-validation of an airline or aircraft engine. A system beneficially determines a possibility of a specific route at risk by assembling a numerical weather prediction (NWP) model and a trajectory models for forecasting a future of the aircraft engine. In an embodiment, the system provides a method for (of) forecasting in future using a specially-configured atmospheric model. The system may determine nowcasting of the aircraft engine by integration of various nowcasting products (implemented at ~10-minute updates). The system may determine post-validation to estimate a risk of a hazard being present at a location and time in future, by applying a spatial and/or temporal uncertainty calculation to an indication related to the estimated atmospheric contamination risk. For example, an uncertainty in future movement of air due to wind is beneficially represented with a standard deviation in units of distance, and a kernel function such as a radial Gaussian with a radius proportional to such standard deviation are beneficially applied to a matrix of hazard probability estimates to derive a spatially smoothed probability. In an example embodiment, the post-validation of the aircraft engine is determined using machine learning, namely algorithms that adaptively modify their operating parameters in response to training data being processed by the algorithms. In an embodiment, the uncertainty estimation may use an approximation to Bayes' Theorem. In an embodiment, the graphical illustration shows that a prediction uncertainty decreases closer to a departure of an aircraft. The graphical illustrations further elucidate an event indication that has occurred before the departure of the aircraft. The event indication comprises a high risk of HAI being encounter on route, a medium risk of HAI being encounter on route and a low risk of HAI being encounter on route ( en route). The indications for the estimated risk on a specific route comprise cost awareness (CA) and risk mitigation (RM). The graphical illustration further elucidates actual costs for the aircraft with forecasted costs of the aircraft. In the medium risk of a HAI encounter on the route, the cost of the aircraft increases by 20% due to extra fuel cost, if the flight is cancelled. The system may prefer a risk mitigation as alternative flight routes for passengers to reduce the cost for aircraft. In the high risk of a HAI encounter on route, the cost of the airline increases by 50% due to extra fuel, if the flight is cancelled. The system may provide alerts including a risk mitigation to the flight passengers to rebook or cancel the flight. In the medium risk of a HAI encounter on the route, the cost for the aircraft increases by 20% due to extra fuel cost, if the flight is cancelled and the system may delay the flight by two hours. The cost awareness and the risk mitigation option for the low risk of HAI encounter on route comprise no extra fuel and no mitigation needed respectively. The cost awareness and the risk mitigation may validate flight data that comprises forecasted cost, actual cost, one or more turbulence encounters and pilot reports. Such fuel savings also represent an energy efficiency of the aircraft when in operation, and are optionally recorded as "carbon offsets" subject to financial repayment to assist to address costs of, for example, executing engine maintenance.

FIGs. 13A to 13B are flow diagrams illustrating steps of a method for (of) training a predictive engine health model using machine learning and using the predictive engine health model to predict an engine health in accordance with an embodiment of the present disclosure. At a step 1802 of the method of training, a set of training data comprising historical data sets for a training set of at least one aircraft engine is provided. At a step 1804 of the method of training, at least one measure of historic contaminant exposure of the set of training data is determined by retrieving the atmospheric contaminant data at a location and a time in proximity to at least one flight in the aircraft flight data. At a step 1806 of the method of training, the historical contaminant exposure and the engine service data are provided to an engine health model. At a step 1808 for prediction purposes, a contaminant exposure measure for a target aircraft engine is determined by retrieving atmospheric contaminant data at a location and a time in proximity to at least one flight of the target aircraft engine. At a step 1810 for prediction purposes, a predictive engine health model is used to predict an engine health parameter of the target aircraft engine based on at least one of target aircraft engine flight history and contaminant data over location and time.

FIG. 14 is an illustration of steps of a method for (of) determining contaminant data using one or more servers, a supercomputing or a distributed computing platform according to an embodiment of the present disclosure. At a step 1902, satellite data is obtained. At a step 1904, flight monitoring data is obtained. At a step 1906, surface measures are obtained. At a step 1908, radiosondes data is obtained. At a step 1910, the satellite data is processed. At a step 1912, the flight monitoring data is processed. At a step 1914, the surface measures are processed. At a step 1916, the radiosondes data is processed; it will be appreciated that radiosondes data is equivalent to sensor data. At a step 1918, import data is calibrated. At a step 1920, boundary conditions for the import data are determined and configured. At a step 1922, the determined boundary conditions are provided to a supercomputing or a distributed computing platform. At a step 1924, the determined boundary conditions are validated. At a step 1926, weather models are determined. At a step 1928, the boundary conditions, the weather models and the validated boundary conditions are obtained and processed by the supercomputing or distributed computing platform. At a step 1930, the processed data is stored in a storage or distributed storage. At a step 1932, the processed data are stored in a working memory. At a step 1934, contaminant distribution of at least one of latitude and longitude, altitude and/or pressure level or time are forecasted using the supercomputing or the distributed computing platform. In an embodiment, the forecast of contaminant data is generated for various latitude, longitude, altitude, pressure level or time.

FIG. 15 is an illustration of an exploded view of a distributed computing system or cloud computing implementation in accordance with an embodiment of the present disclosure; such computing elements are convenient regarding as being a "computing arrangement'. The exploded view comprises an input interface 2002, a control module that comprises a processor 2004, a memory 2006 and a non-volatile storage 2008, processing instructions 2010, a shared/ distributed storage 2012, a server that comprises a server processor 2014, a server memory 2016 and a server non-volatile storage 2018 and an output interface 2020. The function of the server processor 2014, the server memory 2016 and the server non-volatile storage 2018 are thus identical to the processor 2004, the memory 2006 and the non-volatile storage 2008 respectively. The functions of these parts are as has been described above.

Additionally, it should be noted that the present system and method helps to reduce or eliminate certain aircraft emissions through improved maintenance and therefore reduces carbon dioxide (C0₂) emissions.

In embodiments of the disclosure as described in the foregoing, for example where machine learning is employed, it will be appreciated that an estimate of atmospheric contaminant risk is beneficially computed by employing at least one of: a statistical model, a regression model, a machine learning model; at least one historical or present {"live") data set of atmospheric contaminant data is beneficially employed. Moreover, the statistical model or machine learning model is calibrated, trained or optimized by using the at least one historical or present {"live") data set of atmospheric contaminant data or simulations of atmospheric contaminant risk. For example, a statistic method is used to estimate a risk that atmospheric contamination may pose to a given aircraft or aircraft engine. For example, the risk can be calculated using Bayes' theorem as a statistical method :

A represents a contaminant risk, B represents contaminant estimates from an NWP model, p(A) represents prior knowledge of contaminant risk, and p(B) represents statistical uncertainty relating the location, height or time position of a weather event, wherein the risk in the presence of contaminants is susceptible to being estimated using Bayes' theorem by using : p(A I B) = p(B I A) * p(A) / p(B) . p(A) by marginalisation of p(A | B), while p(B | A) and determined by comparison of estimated contaminants from similar NWP models that are evaluated using historical data representative of the respective contaminants over regions where one or more aircraft were known to be affected by adverse weather events. Historical data potentially comprises contaminant mixing ratios, concentrations or integrated amounts estimated using NWP models, or measurements from, radiosondes, ground-based sensors, sensors mounted on aircraft, for events or time points in the past. Moreover, present ("//Ve") atmospheric contaminant data beneficially comprises contaminant mixing ratios, concentrations or integrated amounts estimated using NWP models, or measurements from, radiosondes, ground-based sensors, sensors mounted on aircraft, for times within 24 hours before or after the flight, and beneficially use most recent data that is available. Moreover, a statistical method optionally also comprises a method that estimates at least one probability density, histogram or statistical distribution parameter related to a mean, standard deviation, variance or kurtosis of a distribution of a contaminant, and then uses such at least one parameter to determine a measure indicative of the risk, probability or likelihood that an atmospheric contaminant is potentially present at levels (i.e. volumetric concentrations) above at least one threshold, by comparison with the cumulative distribution function of the contaminant's probability density function associated with the at least one parameter. It will be appreciated that embodiments of the present disclosure benefit from consideration of spatial or temporal uncertainty. Such consideration is achievable by considering estimated values of a contaminant at multiple spatial points in proximity in location, height or time in the vicinity of a waypoint. Optionally, such values have associated therewith a probability density function. Alternatively, a probability density function, or a parameter such as a standard deviation of a probability density function, is optionally estimated from the multiple points. A risk is calculated for each point in proximity. The overall risk is determinable from the maximum risk, wherein the maximum risk is optionally weighted by the probability density function.

Beneficially, embodiments of the present disclosure employ a numerical model, including a regression model or a machine learning model including without limitation a linear regression, logistic regression, decision tree, random forest, support vector machine, Gaussian process or neural network; the numerical model is beneficially used to estimate atmospheric contaminant risk. The numerical model's coefficients are optionally hand-tuned or trained using a machine learning process.

Beneficially, embodiments of the present disclosure employ a numerical model to estimate a target value, a target category, a probability associated with at least one target value or category, or an uncertainty parameter including a standard deviation, variance or covariance matrix.

Models employed in embodiments of the present disclosure are beneficially trained using historical data of measurements of a contaminant at a plurality of spatial locations and at times in the past, for example as aforementioned. Spatial locations and times are determinable from waypoints or from the use of global satellite navigation systems. Measurements are obtainable using various sensing apparatus including ground-based sensors, aircraft-mounted sensors, satellite sensing, radiosondes. Loss of power or loss of flight control of aircraft (adverse events) and normal power or control (benign conditions and events) are beneficially also considered as measurements of high-risk conditions and normal conditions.

When using models of the present disclosure, measurements are used to derive at least one target value for the model to predict. Values obtained from a NWP model of the contaminant and of other relevant parameters (which optionally include temperature, pressure, altitude, wind speed, flight speed) in the vicinity (in time, altitude or location) of the measurement are beneficially used as features for the models. Derived features are calculable using methods of machine learning, including smoothing, averaging, operations including mean, maximum and standard deviation. Optionally, features are normalised, using normalization weighting as described in the foregoing. Features are also determinable from other sources of atmospheric data, including without limitation, ground-based sensors, aircraft-mounted sensors, satellite sensing, radiosondes. Beneficially, features are indicative of the target atmospheric contaminant and of the physical processes that transport it or influence its risk to an aircraft.

Aforementioned models are beneficially trained using corresponding sets of features and targets using methods employed for machine learning. The estimated target value, and any associated probability or uncertainty parameter, is beneficially used to determine an atmospheric contaminant risk as described in the foregoing.

After training, the model is then applied to features similarly determined from historical data or present ("//Ve") data, to provide a measure of atmospheric contaminant risk at corresponding spatial locations and times. Live data may include recent measurements, recent forecasts, nowcasts and other ground- based sensors, aircraft-mounted sensors, satellite sensing, radiosondes, beneficially within 24 hours and beneficially as close in location and time to a waypoint of interest. A probability or uncertainty parameter determined by a model beneficially provides a measure of spatial or temporal uncertainty. An overall risk for a flight or when traversing a given region is optionally determinable by combining an estimated atmospheric contaminant risk from two or more waypoints using at least one of the following operations: mean, maximum, Lp-norm. Beneficially, an index of risk is determinable by comparing an atmospheric contaminant risk (including a probability distribution or value) with a pre- determined scale. Beneficially, values of an index of risk are determinable by comparing at atmospheric pressure with at least one threshold. For example, an index of 1 corresponds to very low risk and to ice water content values below 0.01g/m³, and index of 2, with a higher risk, to values in a range of 0.01 to 0.1 g/m³, an index of 3 to values in a range of 0.1 to 1 g/m³, and so on. An index of risk is beneficially used to represent an atmospheric contaminant risk.

When presenting atmospheric contaminant risk at the output interface, the system beneficially identifies at least one section of the flight plan with a highest risk or where a risk exceeds a threshold.

Next, embodiments of the present disclosure, as described above, methods of recommending flight plan variants will be considered, for example where aircraft trajectory re-routing is recommended. When making recommendations for flight plan variants, a computing arrangement of a system further comprises a database of at least one of aircraft flight plan variables or air traffic constraints, wherein in operation the computing arrangement calculates at least one modified aircraft flight plan for the at least one aircraft flight based on the at least one of aircraft flight plan variables or air traffic constraints, and generates an estimated atmospheric contamination risk for the modified aircraft flight plan for comparison with the estimated atmospheric contamination risk for the at least one aircraft flight.

The resultant indication related to the estimated atmospheric contamination risk provided via the output interface comprises at least one of: an index of the risk of atmospheric contamination associated with an aircraft flight; a recommendation on modification of at least one aircraft flight plan to mitigate or reduce atmospheric contamination exposure; an indication of portions of the aircraft flight where risk of exposure exceeds a threshold.

Beneficially, the method further comprises using the measure of at least one atmospheric contaminant to calculate at least one estimate of the atmospheric contaminant risk in the vicinity of a modified flight plan and adjusting the flight plan by performing a comparison of a plurality of estimated atmospheric contamination risk associated with each of the modified flight plan and the flight plan.

Beneficially, the method further comprises applying an optimisation to reduce the atmospheric contamination risk versus at least one cost metric associated with alternate flight plan.

Beneficially, when the process determines that an unacceptable risk potentially pertains for a given flight plan, the method presents an alternative flight plan having a lower risk. By reference to air traffic control constraints such as approved corridors, historical flights of other aircraft in a region, at least one candidate alternative flight plan is determinable using flight planning methods, for example known contemporary flight planning methods.

Thus, the methods described in the foregoing are susceptible to being used to determine an atmospheric contaminant risk associated with the at least one alternative flight plan. When an alternative flight plan is considered to offer lower atmospheric contaminant risk, it is beneficially recommended as a possible modification.

Models of the present disclosure, as described in the foregoing, beneficially employ a cost function that optionally includes or combines one or more of estimated cost associated with a flight plan, a differential cost from another flight plan, an estimated duration, a differential duration, an atmospheric contaminant risk. Such a cost function optionally weight each cost input with a mutually different weight representative of their relative magnitude. Moreover, such a cost function beneficially allows the system or flight planners to compare alternative strategies, taking into account that increases in flight duration or cost are potentially acceptable if they provide a lower atmospheric contaminant risk.

Embodiments of the disclosure described in the foregoing are implemented using a computing arrangement, wherein the computing arrangement beneficially receives aircraft flight plan data from the input interface; wherein the at least one aircraft flight plan data includes at least one of a time, a pressure or an altitude, a trajectory and a location representing at least one aircraft flight. Optionally, the flight plan includes take-off and landing airports and a trajectory defined by a list of waypoints defined by expected time, latitude, longitude, altitude. When only the takeoff and landing airports are provided, a trajectory is estimable using numerical methods, for example by preparing a typical trajectory between the two airports representative of one or more previous paths taken by a given aircraft.

Optionally, a trajectory includes only of one point or a sequence of points representing a part of a flight path of interest for a given aircraft, for example to traverse a region known to be subject to adverse weather conditions. In such an example, the system optionally processes risk only for the part of the flight path of interest for a given aircraft.

Where time data is not otherwise provided, synthesized time data determinable from the scheduled departure or arrival time of a flight. Where only one time is provided, times of each waypoints are determinable iteratively from another known waypoint in turn by dividing a distance between the two waypoints by an aircraft's typical speed during that part of the flight. In embodiments of the disclosure described in the foregoing, it is convenient that waypoints are known on a similar grid to a numerical weather prediction (NWP) model. Where waypoints are widely separated, they are susceptible to being interpolated or extrapolated, for example on a straight line basis, so that spatial locations are sampled with resolution approximating that of the NWP model.

From the foregoing, it will be appreciated that the estimated atmospheric contamination risk is determined using a measure of at least one atmospheric contaminant for the at least one aircraft flight based upon a location, an altitude or pressure, a trajectory and a time information extracted from the at least one aircraft flight plan data. Beneficially, an NWP model is arranged to compute the concentration or mixing ratio of a contaminant at a plurality of locations, heights and times in proximity to the flight, using methods known in the art such as the WRF and ECMWF models. At one or more waypoints, the computing arrangement beneficially determines an atmospheric contaminant risk by querying the results of an NWP model for that location, time and height to determine a concentration of a contaminant. For example, the NWP model defines the concentration or mixing ratio of ice water, dust, sulphate, sulphur dioxide or sodium chloride. Where the NWP model defines a mixing ratio (for example, in kg of contaminant per kg of air), the concentration is derivable by multiplying the mixing ratio by the density of air at the temperature and pressure estimated at the respective location, time and height.

Where locations, heights or times in the NWP model do not correspond exactly to a waypoint, the closest location, height or time is optionally determined and used to query the NWP model results. Alternatively, interpolation such as linear or bicubic interpolation is used to interpolate the NWP model mixing ratio or concentration at the waypoint.

Next, "risk" as a parameter in embodiments of the disclosure described in the foregoing will be elucidated in greater detail. The expected concentration of a contaminant is optionally used directly when computing a risk. For example, it is known that ice water content poses a particular hazard to aircraft and its concentration is susceptible to being presented to user in embodiments of the present disclosure. Optionally, a threshold is susceptible to being applied to determine a risk. For example, if ice water content exceeds 1 gram per cubic metre, it is potentially considered to pose an unacceptable hazard and therefore will be highlighted as a risk. A pre-determined relationship is optionally used to relate a concentration of a contaminant to a risk. For example, the risk is 10% at a contaminant concentration of 0.1 g/m³, the risk is 30% at a contaminant concentration of 1.0 g/m³, and the risk is 100% at a contaminant concentration of 3 g/m³. Intermediate risk values may be determined by interpolation.

Next, it will be appreciated that embodiments of the disclosure described in the foregoing include a computing arrangement that computes at least one of spatial or temporal uncertainty to derive the indication related to the estimated atmospheric contamination risk. Optionally, a measure of uncertainty is beneficially calculated or derived for the waypoint or closest NWP result location. Moreover, a measure of uncertainty may be estimated by the NWP model by Monte Carlo modelling as known in the art. Moreover, a measure of uncertainty is optionally a value for a particular waypoint or location, or is optionally beneficially averaged or smoothed in respect of regions, heights or times, for example using a Gaussian convolution, to reflect that an estimation of the measure of uncertainty is potentially itself stochastically noisy.

Beneficially, in embodiments of the present disclosure, a probability density function is determined representative of the contaminant density or mixing ratio. By comparing with the probability density function, an estimated concentration and uncertainty is used to determine at least one risk. For example, given an estimated ice water content, a standard deviation of the estimate determined from the measure of spatial or temporal uncertainty, and that the estimation error is considered to follow a Normal distribution, the probability that atmospheric contaminant density at a location potentially exceeds a threshold that is determinable and usable as a measure of the risk. For example, the risk that they contaminant exceeds 1 g/m³ is beneficially presented at the output interface together with its respective location or waypoint. It is known in the art of NWP that models potentially make errors in locating an event such as a storm or dust storm in space or time. Known methods of NWP estimate a magnitude of such errors, for example by estimating cloud visibility as seen from space and by comparing with satellite images of corresponding events. An uncertainty, namely potential error, is susceptible to vary as a function of time or spatial location. Temporal and spatial uncertainty are related, as is known in the art of Kalman filtering, and thus are potentially interchangeable, for example by considering the speed of movement of a weather event or the underlying wind speed. In the present disclosure, spatial uncertainty is optionally defined by at least one standard deviation. By multiplying by a Gaussian function, this spatial uncertainty is converted to a probability density function sampled over space that is representative of a spatial normal distribution.

When a spatial or a temporal uncertainty is defined over a region (optionally, by a probability density function), a contaminant's uncertainty or distribution is susceptible to being estimated at a location using statistical methods. For example, the NWP results pertaining to the region are susceptible to being considered as an ensemble of samples (optionally, with likelihood determined from the probability density function). The proportion of samples that exceed at least one threshold is susceptible to being considered to approximate the risk (optionally, weighted using the probability density function), providing a beneficial method to estimate the risk, as a probability, that the contaminant exceeds each threshold in the approximate location.

Modifications to embodiments of the present disclosure described in the foregoing are possible without departing from the scope of the present disclosure as defined by the accompanying claims. Expressions such as "including", "comprising", "incorporating", "have", "is" used to describe and claim the present disclosure are intended to be construed in a non-exclusive manner, namely allowing for items, components or elements not explicitly described also to be present. Reference to the singular is also to be construed to relate to the plural.

Claims

1. An aircraft engine maintenance system comprising at least one input interface that receives, when in operation, data relating to a given aircraft engine from a sensor arrangement that is connected to the given aircraft engine and supplementary data from a database arrangement, wherein the supplementary data describes operational and maintenance information that characterizes the given aircraft engine; characterized in that the aircraft engine maintenance system further comprises a data processing arrangement that receives, when in operation, the data related to the given aircraft engine as an input data and the supplementary data and uses the input data and the supplementary data to generate normalized data, wherein the normalized data is generated by ingesting the input data and the supplementary data, wherein ingesting the input data and the supplementary data includes normalizing relative weightings of the input data and the supplementary data to generate the normalized data, wherein the data processing arrangement:

2. An aircraft engine maintenance system of claim 1, characterized in that machine learning is used to determine at least one of the following : a selection of a numerical model to be used in the aircraft engine maintenance system, coefficients of a numerical model to be used in the aircraft engine maintenance system, or a set of features used in a numerical model to be used in the aircraft engine maintenance system.

3. An aircraft engine maintenance system of claim 2, characterized in that at least one measure of spatial or temporal uncertainty is determined using a statistical or machine learning model arranged to receive values of a contaminant determined from at least one of: a database, a web API, a NWP model, a measurement.

4. An aircraft engine maintenance system of claim 2 or 3, characterized in that characterized in that at least one environmental contaminant exposure is determined using a statistical or machine learning model arranged to compensate for bias or spatial or temporal uncertainty in values of a contaminant determined from at least one of: a database, a web API, a NWP model, a measurement.

5. An aircraft engine maintenance system of claim 1, 2, 3 or 4, characterized in that the machine learning is arranged, pursuant to training data provided, to monitor temporal rates of change or temporal patterns of change of one or more parameters indicative of the "state-of-health" of the given aircraft engine, to determine a rate of degradation of the given aircraft engine.

6. An aircraft engine maintenance system of claim 5, characterized in that the machine learning monitors temporal rates of change of a plurality of the parameters

7. An aircraft engine maintenance system of any one of claims 1 to 6, characterized in that the aircraft engine maintenance system comprises at least one storage medium that is coupled to the data processing arrangement, wherein the at least one storage medium stores the input data and the supplementary data.

8. An aircraft engine maintenance system of any one of claims 1 to 7, characterized in that the aircraft engine maintenance system comprises at least one output interface that provides to a user the data output representative of the at least one predicted aircraft health parameter for the given aircraft engine.

9. An aircraft engine maintenance system of any one of claims 1 to 8, characterized in that the data processing arrangement, when in operation, processes normalized data comprising a reference to a serial number of the given aircraft engine that is used to select at least a part of the supplementary data for use in the aircraft engine maintenance system.

10. An aircraft engine maintenance system of any one of the preceding claims, characterized in that the data processing arrangement, when in operation, processes the input data comprising at least one of a carrier data, a contaminant exposure data, a temperature record, humidity, and Automatic dependent surveillance - broadcast (ADS-B) positional data, wherein the carrier data comprises at least one of schedules, maintenance or cycles of the given aircraft engine.

11. An aircraft engine maintenance system of any one of the preceding claims, characterized in that the data processing arrangement, when in operation, processes the supplementary data comprising raw data, wherein the raw data comprises a logging filename, a data type and an associated metadata of the input data.

12. An aircraft engine maintenance system of any one of the preceding claims, characterized in that the data processing arrangement, when in operation, processes the supplementary data comprising records of an amount of maintenance intervention associated with an aircraft or engine.

13. An aircraft engine maintenance system of any one of the preceding claims, characterized in that the aircraft engine maintenance system cleans the supplementary data to obtain cleaned supplementary data by performing at least one of:

(i) correcting spelling errors of the supplementary data;

(iii) reformatting the supplementary data; and

(iv) filtering null or duplicate values of the supplementary data.

14. An aircraft engine maintenance system of any one of the preceding claims, characterized in that the aircraft engine maintenance system extends the supplementary data to obtain extended supplementary data by performing at least one of:

(i) identifying changepoints in the data;

(iii) grouping similar events;

(iv) computing a count or accumulated value since an event; and

15. An aircraft engine maintenance system of any one of the preceding claims, characterized in that the aircraft engine maintenance system merges at least one of:

(i) the cleaned supplementary data into the normalized data; and (ii) extended supplementary data into the normalized data.

16. An aircraft engine maintenance system of any one of the preceding claims, characterized in that the numerical model is pre-trained for at least one of: (i) classifying a performance of the given aircraft engine; and

(ii) scoring a health condition of the given aircraft engine.

17. An aircraft engine maintenance system of any one of the preceding claims, characterized in that the output data includes a reference to changes in running costs, recommended maintenance actions, or operational actions to be undertaken by a user.

18. A method for (of) operating an aircraft engine maintenance system, characterized in that the method includes: using at least one input interface of the aircraft engine maintenance system to receive data relating to a given aircraft engine from a sensor arrangement that is connected to the given aircraft engine and supplementary data from a database arrangement, wherein the supplementary data describes operational and maintenance information that characterizes the given aircraft engine; using a data processing arrangement of the aircraft engine maintenance system to receive the data related to the given aircraft engine as an input data and the supplementary data and using the input data and the supplementary data to generate normalized data, wherein the normalized data is generated by ingesting the input data and the supplementary data, wherein ingesting the input data and the supplementary data includes normalizing relative weightings of the input data and the supplementary data to generate the normalized data, using the data processing arrangement: (i) to process the normalized data to determine an operating status of the given aircraft engine by applying the normalized data to a numerical model of the given aircraft engine; and

(ii) to derive from the numerical model at least one predicted aircraft health parameter for the given aircraft engine and to use or provide the derived at least one predicted aircraft health parameter as an input for scheduling a maintenance intervention for the given aircraft.

19. A method of claim 18, characterized in that machine learning is used to determine at least one of the following : a selection of a numerical model to be used in the aircraft engine maintenance system, coefficients of a numerical model to be used in the aircraft engine maintenance system, or a set of features used in a numerical model to be used in the aircraft engine maintenance system.

20. A method of claim 19, characterized in that at least one measure of spatial or temporal uncertainty is determined using a statistical or machine learning model arranged to receive values of a contaminant determined from at least one of: a database, a web API, a NWP model, a measurement.

21. An aircraft engine maintenance system of claim 19 or 20, characterized in that characterized in that at least one environmental contaminant exposure is determined using a statistical or machine learning model arranged to compensate for bias or spatial or temporal uncertainty in values of a contaminant determined from at least one of: a database, a web API, a NWP model, a measurement.

22. A method of claim 18, characterized in that the method includes arranging for the data processing arrangement to employ machine learning to optimize adaptively ingestion to provide the normalized data, and an accuracy of the numerical model.

23. A method of claim 22, characterized in that the machine learning is arranged, for example pursuant to training data provided, to monitor temporal rates of change or temporal patterns of change of one or more parameters indicative of the "state-of-health" of the given aircraft engine, to determine a rate of degradation of the aircraft engine.

24. A method of claim 18, characterized in that the machine learning monitors temporal rate of change of a plurality of the parameters

25. A method of any one of claims 18 to 24, characterized in that the normalized data comprises a reference to a serial number of the given aircraft engine, for use in selecting the supplementary data.

26. A method of any one of claims 18 to 25, characterized in that the input data comprises at least one of a carrier data, a contaminant exposure data, a temperature record, humidity records, and Automatic dependent surveillance - broadcast (ADS-B) positional data, wherein the carrier data comprises at least one of schedules, maintenance or cycles of the given aircraft engine.

27. A method of any one of claims 18 to 26, characterized in that the supplementary data comprises raw data, wherein the raw data comprises a logging filename, a data type and an associated metadata of the input data.

28. A method of any one of claims 18 to 27, characterized in that the supplementary data comprises records of an amount of maintenance intervention associated with an aircraft or engine.

29. A method of any one of claims 18 to 28, characterized in that the method comprises cleaning the supplementary data to obtain cleaned supplementary data by performing at least one of:

(i) correcting spelling errors of the supplementary data, (ii) identifying and correcting or filtering erroneous flight details, timestamps or cycle counts;

(iii) reformatting the supplementary data; and

(iv) filtering null or duplicate values of the supplementary data.

30. A method of any one of claims 18 to 29, characterized in that the method comprises extending the supplementary data to obtain extended supplementary data by performing at least one of:

(i) identifying changepoints in the data;

(iii) grouping similar events;

(iv) computing a count or accumulated value since an event; and

31. A method of claim 30, characterized in that the method comprises merging at least one of:

(i) the cleaned supplementary data into the normalized data; and

(ii) extended supplementary data into the normalized data.

32. A method of any one of claims 18 to 31, characterized in that the method comprises pre-training the numerical model for at least one of:

(i) classifying a performance of the given aircraft engine; and

(ii) scoring a health condition of the given aircraft engine.

33. A method of any one of claims 18 to 32, characterized in that the output data includes a reference to changes in running costs, recommended maintenance actions, or operational actions to be undertaken by a user.

33. A computer program product comprising instructions that are executable on computing hardware to cause the system of any one of claims 1 to 16 to carry out the method of any one of claims 18 to 33.