CN111680875A - Unmanned aerial vehicle state risk fuzzy comprehensive evaluation method based on probability baseline model - Google Patents
Unmanned aerial vehicle state risk fuzzy comprehensive evaluation method based on probability baseline model Download PDFInfo
- Publication number
- CN111680875A CN111680875A CN202010372798.8A CN202010372798A CN111680875A CN 111680875 A CN111680875 A CN 111680875A CN 202010372798 A CN202010372798 A CN 202010372798A CN 111680875 A CN111680875 A CN 111680875A
- Authority
- CN
- China
- Prior art keywords
- unmanned aerial
- aerial vehicle
- index
- risk
- comprehensive evaluation
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0635—Risk analysis of enterprise or organisation activities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N7/00—Computing arrangements based on specific mathematical models
- G06N7/01—Probabilistic graphical models, e.g. probabilistic networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
Abstract
The invention discloses an unmanned aerial vehicle state risk fuzzy comprehensive evaluation method based on a probability baseline model, which comprises the following steps: training and constructing a probability baseline model of multiple indexes of an unmanned aerial vehicle comprehensive evaluation index system by using CM data in a historical normal state of the unmanned aerial vehicle; the method comprises the steps of obtaining real-time CM data of multiple indexes of an unmanned aerial vehicle comprehensive evaluation index system when the unmanned aerial vehicle flies, and respectively calculating quantitative risk indications of the real-time CM data of the multiple indexes by using a probability baseline model of the multiple indexes of the unmanned aerial vehicle comprehensive evaluation index system; and performing dynamic fuzzy comprehensive evaluation on the state risk of the unmanned aerial vehicle by using the calculated quantitative risk indication.
Description
Technical Field
The invention relates to an unmanned aerial vehicle state risk evaluation method, in particular to an unmanned aerial vehicle state risk fuzzy comprehensive evaluation method based on a probability baseline model.
Background
Unmanned aerial vehicles are one of the rapidly developing components in the aviation industry, which have been applied in various civil and commercial fields, such as crop and infrastructure management, emergency management, search and rescue, and the like. Therefore, ensuring the flight safety of the unmanned aerial vehicle is crucial to guarantee the mission capability and safety thereof, and there is an increasing urgent need in the industry to research and improve the risk assessment technology of the unmanned aerial vehicle.
In order to accurately evaluate the flight risk of the unmanned aerial vehicle and thus effectively avoid the occurrence of safety problems, in recent years, some researches on unmanned aerial vehicle risk evaluation methods have been carried out in the industry [3 ]. In consideration of population density, shelter, ground obstacle and other factors, document [4] proposes a risk evaluation method for unmanned aerial vehicles based on a risk map and uses the risk evaluation method for unmanned aerial vehicles to plan paths. Similarly, in consideration of the flight environment factors, document [2] proposes an unmanned aerial vehicle flight risk evaluation method based on a gaussian process model, and finds a minimum risk path through a path integral formula. In addition to the unmanned aerial vehicle risk evaluation method combined with path planning, unmanned aerial vehicle risk evaluation combined with collision theory also receives wide attention. For example, document [6] proposes a simplified model to measure the collision probability of a drone, using the azimuth and collision core angular time rate. In addition, with the flight speed, flight direction and relevant environmental factors as inputs, document [7] constructs a more complex deterministic model to calculate the safe flight limits of the drone, thus avoiding potential conflicts.
The method provides multiple ideas for unmanned aerial vehicle risk evaluation from different angles. However, none of these unmanned aerial vehicle flight risk evaluation methods considers the state information of the unmanned aerial vehicle itself: they often reduce the complex drone system to one mass point in the flight environment and assume that it is always in a normal state or simply represent the drone state with some predefined failure rate, which results in less accurate drone risk assessment results. Due to the fact that modeling is too ideal, corresponding risk assessment results obtained by ignoring unmanned aerial vehicle state information may be inaccurate. The main reasons are that: 1) the risk evaluation model is too ideal, and the state information of the unmanned aerial vehicle which has important influence on flight safety is ignored; 2) the unmanned aerial vehicle comprises various key subsystems, a single machine and massive state monitoring data, and for the complex actual conditions, a quantitative and effective data analysis method, a system and a comprehensive risk evaluation theory are lacked.
Disclosure of Invention
The invention aims to provide a fuzzy comprehensive evaluation method for unmanned aerial vehicle state risk based on a probability baseline model, which is oriented to massive unmanned aerial vehicle state monitoring (CM) data.
The unmanned aerial vehicle state risk fuzzy comprehensive evaluation method based on the probability baseline model for achieving the purpose comprises the following steps:
training and constructing a probability baseline model of multiple indexes of an unmanned aerial vehicle comprehensive evaluation index system by using CM data in a historical normal state of the unmanned aerial vehicle;
the method comprises the steps of obtaining real-time CM data of multiple indexes of an unmanned aerial vehicle comprehensive evaluation index system when the unmanned aerial vehicle flies, and respectively calculating quantitative risk indications of the real-time CM data of the multiple indexes by using a probability baseline model of the multiple indexes of the unmanned aerial vehicle comprehensive evaluation index system;
performing dynamic fuzzy comprehensive evaluation on the state risk of the unmanned aerial vehicle by using the calculated quantitative risk indication;
wherein the CM is referred to as "condition monitoring".
Preferably, the training and establishing a probability baseline model of multiple indexes of the unmanned aerial vehicle comprehensive evaluation index system by using the state monitoring data of the historical normal state of the unmanned aerial vehicle comprises;
extracting the multidimensional risk sensitive characteristics of each index by using a time domain statistical function;
performing dimensionality reduction processing on the extracted multidimensional risk sensitive characteristics of each index by a principal component analysis method, and removing redundant components to obtain a dimensionality reduction characteristic sequence of each index;
removing outliers in the dimensionality reduction feature sequence of each index by using a quartile locus threshold value method to obtain each training feature sequence of the probability base line model for training each index;
and respectively training corresponding models by utilizing each training characteristic sequence to obtain a probability baseline model of a plurality of indexes of the unmanned aerial vehicle comprehensive evaluation index system.
Preferably, the multiple indexes of the unmanned aerial vehicle comprehensive evaluation index system include: pitch control and roll control of the flight control subsystem; a flap, a rudder and a lift of the steering engine subsystem; gas circuits and lubricating oil of the engine subsystem; a main generator and 28V battery for the electrical subsystem; and landing gear of the landing subsystem.
Preferably, the extracting the multidimensional risk sensitive feature of each index by using the time domain statistical function comprises:
for each index in the multiple indexes, selecting CM data capable of effectively reflecting the risk condition of the unmanned aerial vehicle from the CM data in the historical normal state as an index sensitive parameter, and combining the index sensitive parameters into each index sensitive parameter set;
and extracting time domain statistical characteristics of each index sensitive parameter set to obtain the multidimensional risk sensitive characteristics of each index.
Preferably, the extracting the time-domain statistical features of each index-sensitive parameter set includes: and respectively calculating a mean value, a standard deviation value and a root mean square value of each index sensitive parameter in each index sensitive parameter set.
Preferably, the obtaining of the multidimensional risk sensitivity characteristic of each index comprises: and respectively calculating the mean value, the standard deviation value and the root mean square value of each index sensitive parameter, and combining the mean value, the standard deviation value and the root mean square value to form a feature vector to serve as the multi-dimensional risk sensitive feature.
Preferably, the training the corresponding model by using each training feature sequence respectively comprises: training a corresponding Gaussian mixture model by using each training characteristic sequence, and quantitatively characterizing the Gaussian mixture model; and taking the trained Gaussian mixture model as a probability baseline model of the corresponding index.
Preferably, the acquiring real-time CM data corresponding to a plurality of indexes of the comprehensive unmanned aerial vehicle evaluation index system during flight of the unmanned aerial vehicle includes:
collecting real-time CM data of the unmanned aerial vehicle during flying;
extracting the multidimensional risk sensitive characteristics of each index from the acquired real-time CM data of the unmanned aerial vehicle during flight by using a time domain statistical function;
performing dimensionality reduction processing on the extracted multidimensional risk sensitive characteristics of each index by a principal component analysis method, and removing redundant components to obtain a dimensionality reduction characteristic sequence of each index;
and eliminating wild values in the dimensionality reduction feature sequence of each index by using a quartile site threshold value method to obtain real-time feature data of a plurality of indexes corresponding to the unmanned aerial vehicle comprehensive evaluation index system.
Preferably, the calculating the quantitative risk indicator of the real-time CM data of the multiple indexes respectively by using the constructed probability baseline model of the multiple indexes of the unmanned aerial vehicle comprehensive evaluation index system comprises:
and performing characteristic transformation on the obtained real-time CM data corresponding to each index of the unmanned aerial vehicle comprehensive evaluation index system to obtain real-time characteristic data, inputting the real-time characteristic data into the probability baseline model of the corresponding index, and outputting the real-time quantitative risk indication of the corresponding index.
Preferably, the dynamic fuzzy comprehensive evaluation of the risk of the unmanned aerial vehicle state by using the calculated quantitative risk indication comprises:
dynamically and adaptively adjusting the pre-calculated initial weight according to the quantitative risk indication of the real-time CM data of each index, wherein indexes with larger quantitative risk indications obtain higher dynamic weights;
converting quantitative risk indications of the real-time CM data of the multiple indexes into probability form membership degree vectors by using membership degree functions;
and carrying out fuzzy comprehensive operation on indexes of each level by using the dynamic weight and the local membership vector, and finally obtaining membership representation of the overall state risk of the unmanned aerial vehicle.
The invention has the beneficial technical effects that:
(1) the probability baseline model can be used for carrying out self-adaptive modeling representation on the historical normal distribution of the CM data of the unmanned aerial vehicle, and further converting the real-time CM data into quantitative risk indication with probability significance, so that the effective utilization of the state information of the unmanned aerial vehicle and the accurate measurement of the state risk are realized in a data-driven mode.
(2) The adopted dynamic fuzzy comprehensive evaluation model can be combined with quantitative risk indication calculated in real time, and index weight is dynamically adjusted in a self-adaptive manner through variable weight factors, so that early slight state risk of the unmanned aerial vehicle is discovered in time; in addition, comprehensive evaluation of the state risk of the unmanned aerial vehicle can be completed in a probability form through fuzzy operation based on the membership function, so that more reasonable information support is provided for relevant decision making under uncertain conditions.
(3) The unmanned aerial vehicle state fuzzy comprehensive evaluation method based on the probability baseline model provides a set of method flows for effectively integrating the logic composition knowledge of the qualitative unmanned aerial vehicle and the state information of the quantitative unmanned aerial vehicle, and can be effectively popularized and used in various types of unmanned aerial vehicles and even similar complex systems.
Drawings
FIG. 1 is a schematic diagram of a fuzzy comprehensive evaluation method for unmanned aerial vehicle state risk based on a probability baseline model;
FIG. 2 is a schematic of the anomaly injection data of the present invention;
FIG. 3 is a diagram illustrating feature extraction, dimension reduction, and outlier rejection according to the present invention;
FIG. 4 is a probabilistic baseline model and multiple anomaly state feature space schematic of the present invention;
FIG. 5 is a schematic diagram of the quantitative risk indication and dynamic weight adjustment of the present invention;
FIG. 6 is a representation of the risk quantification of the overall state of an UAV using static weights and membership vectors of the present invention;
fig. 7 is a representation of the risk quantification of the overall state of the drone using dynamic weights and membership vectors of the present invention.
Detailed Description
In the flight phase of the unmanned aerial vehicle, the method can provide real-time quantitative indexes for the state risk of the unmanned aerial vehicle, so that guarantee is provided for the safe flight of the unmanned aerial vehicle. Because the unmanned aerial vehicle is complex in composition and difficult to directly measure the state risk by using an accurate mathematical model, the invention adopts a strategy of graded evaluation: the method comprises the steps of firstly decomposing a complex unmanned aerial vehicle system into a multi-level evaluation index system, analyzing CM data of each local index by adopting a data driving method, quantifying risks of the CM data, and finally obtaining the whole state risk of the unmanned aerial vehicle in a probability membership form through fuzzy comprehensive operation. The method can effectively mine the state information in the mass CM data of the complex unmanned aerial vehicle, provide real-time and quantitative state risk indication for the complex unmanned aerial vehicle, and provide powerful information support for flight safety and relevant decision making of the unmanned aerial vehicle.
The invention provides an unmanned aerial vehicle state risk fuzzy comprehensive evaluation method based on a probability baseline model, which is realized by the following steps:
step one, establishing an unmanned aerial vehicle comprehensive evaluation index system. And establishing a multi-level index system for risk evaluation of the unmanned aerial vehicle by combining the key subsystem of the unmanned aerial vehicle and the logic composition and function positioning of the single machine. Meanwhile, an initial weight among all levels of indexes is defined by using an order relation method, and the risk levels of the indexes are divided, so that the subsequent membership function design and the generation of risk membership vectors are supported.
And step two, in an off-line stage, training and constructing a probability baseline model of each local index by using state monitoring data of the historical normal state of the unmanned aerial vehicle. And selecting the CM data in the historical normal state for each local index of the index system in the step one. Firstly, extracting multidimensional risk sensitive characteristics of the multi-dimensional risk sensitive characteristics by using a time domain statistical function; secondly, performing dimensionality reduction treatment on the redundant components through a principal component analysis method so as to remove the redundant components; and finally, performing unsupervised fitting on the characteristic space distribution by using a Gaussian mixture model, so that the data distribution characteristics of each local index in a normal state are embedded and represented in a model parameter space, wherein the parameter space is a probability baseline.
And step three, in an online stage, calculating quantitative risk indication of the real-time CM data by using the trained probability baseline model. The state risk of each local index is embodied as the deviation of the real-time CM data distribution from the historical normal CM data distribution, and for the purpose, the deviation likelihood of the probability baseline model is utilized to quantify the data distribution difference, and the data distribution difference is used as the quantitative risk indication of the real-time CM data.
And step four, carrying out dynamic fuzzy comprehensive evaluation on the state risk of the unmanned aerial vehicle. Firstly, according to the quantitative risk indication calculated by each local index data in real time, the initial weight is dynamically and adaptively adjusted by using a variable weight factor, and the index with larger quantitative risk indication obtains higher dynamic weight; secondly, converting QRI of the local indexes into probability form membership degree vectors by using membership degree functions; and finally, carrying out fuzzy comprehensive operation on indexes of each level according to the dynamic weight and the local membership vector, and finally obtaining membership representation of the overall state risk of the unmanned aerial vehicle.
The present invention will be described in further detail with reference to the accompanying drawings and examples.
FIG. 1 is a flow chart of a method of the present invention, as shown in FIG. 1, comprising the steps of:
step one, establishing an evaluation index system
In order to more effectively evaluate the state risk of the complex unmanned aerial vehicle with a plurality of key subsystems and single machines, an evaluation index system is firstly constructed, and specifically the evaluation index system comprises the steps of construction of the evaluation index system, initial weight definition and risk grade division.
Step 101, building an evaluation index system
A drone is a complex system with multiple subsystems and a single machine, whose overall state during flight is synthetically affected by the real-time state of its components. Therefore, firstly, the flight state of the unmanned aerial vehicle is set as a primary index of an evaluation index system; then, according to the logical composition structure and the component functional positioning of unmanned aerial vehicle during flight, set five key unmanned aerial vehicle subsystems as the secondary index: it is respectively a flight control subsystem, a redirector subsystem, an engine subsystem, an electrical subsystem and a landing subsystem; on the basis, ten important components of the subsystems are set as three-level indexes: it is pitch control, roll control, flaps, rudders, elevators, gas circuits, oil, main generator, 28V battery and landing gear respectively; finally, for each three-level index uijThe selection of the invention can effectively reflectCM parameters of risk conditions thereof are taken as index sensitive parameters and combined into an index sensitive parameter setWhere E represents the number of sensitive CM parameters,indicating a certain sensitive CM parameter carrying risk related information. Therefore, a three-level index system for unmanned aerial vehicle state risk evaluation is constructed, and detailed information of the three-level index system is shown in table 1.
TABLE 1 unmanned aerial vehicle risk evaluation index system
Step 102, initial weight definition
Defining the initial weight of the index is another key step in the state risk assessment of the unmanned aerial vehicle, which determines the degree to which each local index affects the overall state risk of the unmanned aerial vehicle. The method adopts the order relation method to define the initial weight of the index system, does not need consistency check in the analytic hierarchy process, and is more suitable for weight calculation of a complex index system. The process of defining the initial weight set by using the order relation method mainly comprises 3 steps: index importance ranking, index relative importance distribution and weight calculation.
Step 102-1, index importance ranking
For index set u of the same level in the index system1,u2,...,ui,...,un-1,unThe index sets are sorted in descending order according to the relative importance degree among the indexes and are recorded asWhereinIndicating that the ranked index i is relatively more important or equally important than the ranked index j.
102-2, index relative importance distribution
ri=Wi-1/Wi
wherein, Wi-1And WjRespectively representing the weights of the i-1 th index and the i-th index after sorting, riIs the relative importance between the two indexes. The assignment of relative importance and evaluation rules are shown in table 2.
TABLE 2 rule for relative importance of indexes
Step 102-3, index weight calculation
The relative importance between the indexes is subjected to successive multiplication operation to obtain:
setting k to 2, and summing the continuous multiplication results k to n to obtain:
since the sum of all index weights is specified to be 1, the following derivation can be made:
as can be seen from the above derivation formula, the relative importance r of the index based on the assigned valueiThe final index weight can be derived:
similarly, based on the last index weight WnThe other index weights can be derived one by one:
Wn-1=rnWn
by using the order relation method, the initial weight of any layer of index set in the index system can be obtained.
Step 103, risk level definition
In addition to defining an initial set of weights, a risk level is a hierarchical representation of the risk assessment results by fuzzy language, which may be expressed as V ═ V (V ═ V)1,v2,...,vk,...,vK) Where K is the number of risk classes. In the present invention, K ═ 4 is defined herein, and specifically expressed as: health, sub-health, danger, and failure. The risk level setting will serve as a reference basis for subsequent membership function design and will determine the physical meaning and expression form of the final fuzzy risk vector.
Step two, constructing an offline probability baseline model
The difference between the real-time CM data distribution and the historical normal CM data distribution reflects the state risk of indexes corresponding to the unmanned aerial vehicle, and quantitative representation needs to be carried out on the historical normal distribution of each index CM data. In the invention, the probability baseline model is constructed by using the historical normal CM data of the unmanned aerial vehicle offline so as to realize the quantitative characterization of the CM baseline state. Specifically, the method comprises the steps of feature extraction, dimension reduction compression, outlier rejection and probability baseline model construction.
Step 201, time domain statistical feature extraction
Feature extraction is a necessary operation in data mining, and can extract useful information from original data, so as to achieve the purposes of highlighting the intrinsic characteristics of the data and improving the subsequent analysis performance. In consideration of the data characteristics of the unmanned aerial vehicle, the invention extracts 3 time domain statistical characteristics for each parameter of the unmanned aerial vehicle: mean (M), standard deviation (Std) and Root Mean Square (RMS) values to construct a feature space for each sensitive CM parameter of the drone.
Given a CM parameter time series containing L sampling points, the three time domain statistical characteristics are respectively expressed as follows:
by time domain statistical feature extraction, to a set containing E-sensitive parametersThe invention will obtain a feature vector F of 3E × 1ijIt can be expressed as:
step 202, dimension reduction compression processing
On the basis of feature extraction, the invention performs dimension reduction compression processing on the obtained feature vector, thereby eliminating redundant features and improving algorithm performance.
In the invention, the principal component analysis method is used for carrying out dimension reduction compression on the feature vector of each index, and the principal component analysis method is most widely applied due to good performance and low calculation complexity. By a principal component analysis method, the feature vector after dimensionality reduction is obtainedO is less than 3E, wherein,also called principal component, which contains most of the useful information in the original feature vector.
Step 203, outlier rejection
Abnormal values existing in the characteristic sequence can damage the effect of unmanned aerial vehicle state risk evaluation. In this regard, the present invention utilizes outlier culling based on a quartile threshold to eliminate outliers. For index uijGiven a feature matrix whose dimensionality reduced is N × OWhereinIs a feature sequence of the o-th dimension with N points. For each feature sequenceWherein a threshold interval is exceededIs determined as a outlier, wherein in each case a lower threshold valueAnd upper thresholdThe calculation is defined as follows:
wherein the content of the first and second substances,andare respectively a characteristic sequenceC is a threshold coefficient. After outlier rejection, a new feature matrixWill be used to train the probabilistic baseline model.
Step 204, probability baseline model construction
By an index uijFor example, given a feature matrix extracted from the CM data in its historical normalcyThe method is characterized in that a Gaussian mixture model is trained to quantitatively represent the indexes, and the trained Gaussian mixture model is used as a probability baseline model of the indexes.
The distribution of the CM data under the historical normal state can be quantitatively characterized in the form of weighted summation of M Gaussian components by a Gaussian mixture model, and a limited number of mixture models can be expressed as follows:
wherein the content of the first and second substances,is the proportional coefficient of the mixed model,i.e. the sum of their coefficients is 1. Probability density p (x | theta) of each Gaussian mixture componentm) Are a gaussian probability density function and each component is represented by a set of parameters thetam=(μm,Sm) Is represented bymIs the mean vector, σmIs a covariance matrix. The parameters finally form a parameter vector set phi ═ pi1,...,πm;θ1,...,θm)。
In the training stage, the invention utilizes the expectation maximization algorithm to maximize the subordinate likelihood functions to fit the optimal parameter vector set of the model
When likelihood functionConverging to a local optimum, the invention obtains the optimum parameter vector setThe Gaussian mixture model at this moment is taken as a probability baseline model which realizes the index uijIs determined by the baseline state feature spatial distribution.
Step three, online quantitative risk indication generation
The probability baseline model trained in the off-line stage quantifies and represents the baseline state distribution of parameter data of each index CM. In the online stage, the real-time CM parameter data of each index is input into the probability baseline model, and the model can output 0-1 deviation likelihood as the real-time quantitative risk indication of the index. Specifically, the method comprises feature extraction, dimensionality reduction compression, outlier rejection and model training.
Step 301, time domain statistical feature extraction
Similar to step 201, the description is omitted here.
Step 302, dimension reduction compression processing
Similar to step 202, further description is omitted here.
Step 303, outlier rejection
Similar to step 203, it is not described herein.
Step 304, quantitative risk indication generation
On-line phase, givenFeature vector converted from real-time CM data of each indexThe invention calculates the probability density of the model through the corresponding probability base line modelThis probability density quantification characterizes the difference between the real-time state data distribution of each index and its historical state data distribution. The closer the real-time CM data of the index is to the CM data sample in the baseline state, the normalized probability density thereofApproaching 1; normalized probability density of the CM data of an index the more it deviates from the CM data sample at baselineThe closer to 0 will be. To conform to the risk expression convention, the present invention uses operationsWill normalize the probability densityConversion to deviation likelihood QijAnd the real-time risk is taken as a quantitative risk indication reflecting the real-time risk of each index.
For index set ui=(ui1,ui2,...,uij,...,uin) The operation is repeated for each index in the set, and the corresponding quantitative risk indication set can be obtained finally by the method
Step four, dynamic fuzzy comprehensive evaluation
In order to timely capture the slight risk of the index in the early abnormal state and realize the comprehensive evaluation of the overall state risk of the unmanned aerial vehicle. The invention uses a dynamic fuzzy comprehensive evaluation method and utilizes the obtained three-level indexes to quantify the risk indication in real time to carry out comprehensive evaluation step by step. Specifically, the method comprises dynamic weight adjustment based on variable weight factors, membership matrix operation and fuzzy comprehensive evaluation of the overall state risk of the unmanned aerial vehicle.
Step 401, dynamic weight adjustment based on variable weight factor
To capture the slight risk in early abnormal states, first a quantitative risk indication measured in real timeAccording to the weight factor, corresponding to the initial weight set Wi=(wi1,wi2,...,wij,...,win) And carrying out dynamic adjustment. The process of dynamic weight adjustment can be expressed as:
wherein, wijIs an index uijInitial weight of, QijIs an index uijIn real-time to quantify the risk indication,is a new weight dynamically adjusted by a variable weight factorα are variable weight sensitivity coefficients (i.e., variable weight factors) ranging from 0-1 for adjusting the sensitivity of the index weights to real-time quantified risk indicators the closer the variable weight sensitivity coefficient α is to 0, the more new weights are assigned to those indicators where the quantified risk indicators change significantly, which makes the slight local risk more likely to be reflected in the overall risk of the unmanned aerial vehicle, conversely, the closer the variable weight sensitivity coefficient α is to 1, the less sensitive the weight adjustment is to the change in the quantified risk indicators in real time for each indicator, and when the variable weight sensitivity coefficient α is equal to 1, the dynamic weights based on the variable weights will degrade to static weights, which means that any change in the real-time quantified risk indicators of the indicators will not cause any change in the quantified risk indicatorsNew weight adjustments.
Step 402, membership matrix operation
To generate a probabilistic risk assessment result under uncertain conditions, a real-time quantitative risk indicator Q for each indicator is requiredijPerforming membership degree vectorAnd calculating, wherein the membership degree vector represents the probability that the real-time risk of the index is subordinate to each risk level.
The invention uses a rigid-type membership function with a smoother probability curve form to calculate the membership vector, and the membership function is set as follows with reference to the risk level setting in step 103:
wherein Q isijFor a given index uijA real-time quantitative risk indication of k1-k6Is the membership boundary for each risk level,is an index uijWherein the number of elements in the membership vector is the same as the number of risk classes, and the quantification thereof characterizes the current index uijThe probability belonging to each risk level repeats the above operation for the rest indexes in the index set to obtain the index set ui=(ui1,ui2,...,uij,...,uin) Is given by the membership matrix Ri:
Ri=[ri1,ri2,...,rij,...,rin]T
Step 403, fuzzy comprehensive risk evaluation of overall state of unmanned aerial vehicle
By using the weight sets and the corresponding risk membership degree matrixes to carry out step-by-step weighting operation, the fuzzy comprehensive evaluation of the overall state risk of the unmanned aerial vehicle can be realized. The above process can be expressed as:
wherein, Bi=(bi1,bi2,..,bik,...,biK) Is a fuzzy risk vector of the first-level comprehensive evaluation, and is a new membership vector supporting the next-level comprehensive evaluation. By repeating the operations step by step, the invention obtains the fuzzy risk vector of the overall state of the unmanned aerial vehicle, and the vector reflects the probability that the whole unmanned aerial vehicle is at each risk level, thereby completing the fuzzy evaluation of the state risk of the unmanned aerial vehicle in a quantitative mode. .
Description of the examples
The method is tested by adopting real state monitoring data of a certain type of unmanned aerial vehicle. For each sensitive CM parameter, the invention uses 25 time series samples collected under normal state of the droneAnd constructing a probability baseline model, wherein the number of sampling points of each time series sample is set to be L-5000.
Meanwhile, the effectiveness of the unmanned aerial vehicle risk state evaluation method is tested. The invention aims at 2 three-level indexes (gas path u)3128V storage battery u41) Subordinate 4 sensitive CM parameters (Cylinder head temperature)Cylinder head temperature28V battery voltage28V battery voltage) And 5 time sequence samples are selected respectively, and 4 kinds of abnormity (decline states I-IV) with gradually increased degrees are injected respectively, so that the risk evaluation effect of the method is more intuitively displayed.
Specifically, anomalies are injected into the normal-sensitive CM parameter time series in a manner that superimposes gaussian random sequences. By sensing CM parametersA time series of samples ofFor example, the fault injection process is as follows:
wherein the content of the first and second substances,are time series samples after the fault injection,is formed by L Gaussian distributionsAn abnormal sequence of random numbers is sampled. WhereinRespectively being time series samplesMean and variance ofc1、c2Mean and variance abnormal coefficients are respectively. In the invention, 4 groups of abnormal coefficients (0.1 ), (0.2,0.2), (0.3 ) and (0.4 ) are respectively set so as to simulate 4 kinds of abnormal state data with gradually increased degrees. FIG. 2 shows a three-level index gas path u31Two subordinate CM parameters (28V Battery Voltage)28V battery voltage) A certain normal time series and an abnormal time series after fault injection.
In step 101, the present invention uses a sequence relation method to perform initial weight definition on each level of index. According to the logic composition and function setting of the unmanned aerial vehicle, the relative importance of 5 secondary indexes is firstly sequenced: flight control subsystem>Steering engine subsystem>Engine subsystem>Electrical subsystem>And the landing subsystem, wherein the relative importance between the two indexes is set to be (1.2,1.1,1.0 and 1.3). Thus, the present invention obtains the initial weight W of the secondary index set (0.254,0.212,0.193,0.193,0.148) as the calculation process described in step 101-1 to step 101-3. Meanwhile, the three-level indexes subordinate to each second-level index are set to be equally important for each second-level index, namely the relative importance degrees of the three-level indexes are all 1. Similarly, the calculation process described in step 101-1 to step 101-3 yields initial weights for 5 sets of three-level metrics: w1=(0.500,0.500),W2=(0.333,0.333,0.333),W3=(0.500,0.500),W4=(0.500,0.500)and W5=(1.000)。
After the initial weight of an index system is defined, the method uses the CM data of each index in a historical normal state to construct a probability base line model, and outputs the quantitative risk indication of the real-time CM data of each index by using the probability base line model. In contrast, as described in step 201-203, time domain feature extraction, dimension reduction compression and outlier elimination operations need to be performed first to obtain the feature space of the CM data. Specifically, the dimensionality reduction in the principal component analysis method is set to 2, and the wild is setThe threshold coefficient in value elimination is set to 1.41, and the hyper-parameters (the number of gaussian models and the covariance types) in each gaussian mixture model are automatically obtained by cross-validated grid search. FIG. 3 shows a three-level index gas path u31Two subordinate CM parameters (28V Battery Voltage)28V battery voltage) The effects of feature extraction, dimension reduction compression and wild value elimination are achieved.
As shown in FIG. 3, the invention firstly uses a three-level index gas path u316-dimensional feature vectors are extracted from subordinate 2 sensitive CM parameter data, and the features highlight the intrinsic state characteristics of the CM data from different aspects; on the basis, through dimension reduction compression and outlier rejection, outliers in a feature space are well rejected, 2-dimensional feature vectors are obtained, and the feature vectors are input into a Gaussian mixture model to support the construction of a probability baseline model. In this regard, a probabilistic baseline model constructed using historical normative feature data and feature data distributions under different abnormal conditions are shown in fig. 4.
As shown in fig. 4, the constructed probability baseline model is shown in the form of normalized probability density contour, and it can be seen that the probability baseline model well represents the feature space in the historical normality. In addition, as shown in the distribution of the triangular identification features in fig. 4, the distances between the feature distributions at different abnormal degrees and the normal feature distributions are increased, and the distribution deviation can be quantitatively characterized by the quantitative risk indication of the probability baseline model.
Repeating the above operations on other three-level indexes to obtain corresponding QRI set with all three-level quantitative risk indications1=(Q11,Q12)、QRI2=(Q21,Q22,Q23)、QRI3=(Q31,Q32)、QRI4=(Q41,Q42) And QRI5=(Q51). According toThese quantized risk indication sets, corresponding three-level initial weight sets, are dynamically adjusted by variable weight factors and converted into new weight setsBased on the three-level new weight sets and the corresponding three-level quantitative risk indication sets, the method can perform weighting operationObtaining a second-level quantization risk set Qi. From these secondary quantitative risk indications, the secondary weight set W ═ can be defined in a similar manner (W)1,w2,w3,w4,w5) Dynamically adjusting to obtain a new weight setThe dynamic adjustment process of the secondary weight set is shown in fig. 5.
As shown in fig. 5, the weights of the 2 secondary indicators injected with the anomaly also show continuous increase as the corresponding quantitative risk indicator increases. It follows that dynamic weight adjustment based on variable weight factors shows good sensitivity to abnormal changes in the index, which is beneficial for finding slight risks of local indexes in early abnormal states.
By using the dynamic weights of all levels and the corresponding quantitative risk indications, the method can perform fuzzy comprehensive evaluation step by step and finally obtain the risk membership vector of the overall state risk of the unmanned aerial vehicle. In the invention, a membership boundary hyperparameter of a membership function is set to k1=0.1、k2=0.3、k3=0.4、k4=0.6、k50.7 and k60.9. In contrast, fig. 6 and 7 show the overall risk fuzzy comprehensive evaluation results of the overall state of the unmanned aerial vehicle obtained by using the static weight and the dynamic weight, respectively.
As shown in the figure, by combining with real-time quantitative risk indication, along with the increase of the injection abnormal degree, the probability that the overall state risk of the unmanned aerial vehicle obtained by evaluation of the method disclosed by the invention belongs to the fourth-level risk level (failure) is continuously increased, and the method is more reasonable compared with a fuzzy comprehensive evaluation result based on static weight, so that the method disclosed by the invention has higher sensitivity to local index abnormality and can be used for capturing and judging the early abnormal state of the unmanned aerial vehicle in time.
Although the present invention has been described in detail hereinabove, the present invention is not limited thereto, and various modifications can be made by those skilled in the art in light of the principle of the present invention. Thus, modifications made in accordance with the principles of the present invention should be understood to fall within the scope of the present invention.
Claims (10)
1. An unmanned aerial vehicle state risk fuzzy comprehensive evaluation method based on a probability baseline model comprises the following steps:
training and constructing a probability baseline model of multiple indexes of an unmanned aerial vehicle comprehensive evaluation index system by using CM data in a historical normal state of the unmanned aerial vehicle;
the method comprises the steps of obtaining real-time CM data of multiple indexes of an unmanned aerial vehicle comprehensive evaluation index system when the unmanned aerial vehicle flies, and respectively calculating quantitative risk indications of the real-time CM data of the multiple indexes by using a probability baseline model of the multiple indexes of the unmanned aerial vehicle comprehensive evaluation index system;
performing dynamic fuzzy comprehensive evaluation on the state risk of the unmanned aerial vehicle by using the calculated quantitative risk indication;
wherein, the CM refers to condition monitoring.
2. The unmanned aerial vehicle state risk fuzzy comprehensive evaluation method based on the probability baseline model according to claim 1, wherein the training and constructing the probability baseline model of the multiple indexes of the unmanned aerial vehicle comprehensive evaluation index system by using the state monitoring data of the historical normal state of the unmanned aerial vehicle comprises;
extracting the multidimensional risk sensitive characteristics of each index by using a time domain statistical function;
performing dimensionality reduction processing on the extracted multidimensional risk sensitive characteristics of each index by a principal component analysis method, and removing redundant components to obtain a dimensionality reduction characteristic sequence of each index;
removing outliers in the dimensionality reduction feature sequence of each index by using a quartile locus threshold value method to obtain each training feature sequence of the probability base line model for training each index;
and respectively training corresponding models by utilizing each training characteristic sequence to obtain a probability baseline model of a plurality of indexes of the unmanned aerial vehicle comprehensive evaluation index system.
3. The unmanned aerial vehicle state risk fuzzy comprehensive evaluation method based on the probability baseline model according to claim 1 or 2, wherein the plurality of indexes of the unmanned aerial vehicle comprehensive evaluation index system comprise: pitch control and roll control of the flight control subsystem; a flap, a rudder and a lift of the steering engine subsystem; gas circuits and lubricating oil of the engine subsystem; a main generator and 28V battery for the electrical subsystem; and landing gear of the landing subsystem.
4. The unmanned aerial vehicle state risk fuzzy comprehensive evaluation method based on the probability baseline model according to claim 1, 2 or 3, wherein the extracting the multidimensional risk sensitive feature of each index by using the time domain statistical function comprises:
for each index in the multiple indexes, selecting CM data capable of effectively reflecting the risk condition of the unmanned aerial vehicle from the CM data in the historical normal state as an index sensitive parameter, and combining the index sensitive parameters into each index sensitive parameter set;
and extracting time domain statistical characteristics of each index sensitive parameter set to obtain the multidimensional risk sensitive characteristics of each index.
5. The unmanned aerial vehicle state risk fuzzy comprehensive evaluation method based on the probability baseline model according to claim 1, 2 or 3, wherein the time domain statistical feature extraction for each index sensitive parameter set comprises:
and respectively calculating a mean value, a standard deviation value and a root mean square value of each index sensitive parameter in each index sensitive parameter set.
6. The unmanned aerial vehicle state risk fuzzy comprehensive evaluation method based on the probability baseline model according to claim 5, wherein the obtaining of the multidimensional risk sensitivity characteristic of each index comprises:
and respectively calculating the mean value, the standard deviation value and the root mean square value of each index sensitive parameter, and combining the mean value, the standard deviation value and the root mean square value to form a feature vector to serve as the multi-dimensional risk sensitive feature.
7. The unmanned aerial vehicle state risk fuzzy comprehensive evaluation method based on the probability baseline model according to claim 2, wherein the training of the corresponding model with each training feature sequence respectively comprises:
training a corresponding Gaussian mixture model by using each training characteristic sequence, and quantitatively characterizing the Gaussian mixture model;
and taking the trained Gaussian mixture model as a probability baseline model of the corresponding index.
8. The fuzzy comprehensive evaluation method for unmanned aerial vehicle state risk based on probability baseline model according to claim 2, wherein the obtaining of real-time CM data corresponding to multiple indexes of the comprehensive evaluation index system of the unmanned aerial vehicle during flight comprises:
collecting real-time CM data of the unmanned aerial vehicle during flying;
extracting the multidimensional risk sensitive characteristics of each index from the acquired real-time CM data of the unmanned aerial vehicle during flight by using a time domain statistical function;
performing dimensionality reduction processing on the extracted multidimensional risk sensitive characteristics of each index by a principal component analysis method, and removing redundant components to obtain a dimensionality reduction characteristic sequence of each index;
and eliminating wild values in the dimensionality reduction feature sequence of each index by using a quartile site threshold value method to obtain real-time feature data of a plurality of indexes corresponding to the unmanned aerial vehicle comprehensive evaluation index system.
9. The unmanned aerial vehicle state risk fuzzy comprehensive evaluation method based on the probability baseline model according to claim 2, wherein the step of respectively calculating quantitative risk indicators of the real-time CM data of the plurality of indicators by using the probability baseline model of the plurality of indicators of the constructed unmanned aerial vehicle comprehensive evaluation indicator system comprises:
and performing characteristic transformation on the obtained real-time CM data corresponding to each index of the unmanned aerial vehicle comprehensive evaluation index system to obtain real-time characteristic data, inputting the real-time characteristic data into the probability baseline model of the corresponding index, and outputting the real-time quantitative risk indication of the corresponding index.
10. The fuzzy comprehensive evaluation method for unmanned aerial vehicle state risk based on probability baseline model according to claim 2, wherein the dynamic fuzzy comprehensive evaluation for unmanned aerial vehicle state risk using the calculated quantitative risk indication comprises:
dynamically and adaptively adjusting the pre-calculated initial weight according to the quantitative risk indication of the real-time CM data of each index, wherein indexes with larger quantitative risk indications obtain higher dynamic weights;
converting quantitative risk indications of the real-time CM data of the multiple indexes into probability form membership degree vectors by using membership degree functions;
and carrying out fuzzy comprehensive operation on indexes of each level by using the dynamic weight and the local membership vector, and finally obtaining membership representation of the overall state risk of the unmanned aerial vehicle.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010372798.8A CN111680875B (en) | 2020-05-06 | 2020-05-06 | Unmanned aerial vehicle state risk fuzzy comprehensive evaluation method based on probability baseline model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010372798.8A CN111680875B (en) | 2020-05-06 | 2020-05-06 | Unmanned aerial vehicle state risk fuzzy comprehensive evaluation method based on probability baseline model |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111680875A true CN111680875A (en) | 2020-09-18 |
CN111680875B CN111680875B (en) | 2022-11-01 |
Family
ID=72433916
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010372798.8A Active CN111680875B (en) | 2020-05-06 | 2020-05-06 | Unmanned aerial vehicle state risk fuzzy comprehensive evaluation method based on probability baseline model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111680875B (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112465145A (en) * | 2020-12-02 | 2021-03-09 | 西北工业大学 | Unmanned cluster intelligent qualitative evaluation method based on logical reasoning and fuzzy synthesis |
CN112463853A (en) * | 2020-11-27 | 2021-03-09 | 重庆医药高等专科学校 | Financial data behavior screening working method through cloud platform |
CN113031647A (en) * | 2021-02-25 | 2021-06-25 | 浙江工业大学 | Power supply type unmanned aerial vehicle optimal path planning method based on fuzzy comprehensive evaluation |
CN114398769A (en) * | 2021-12-29 | 2022-04-26 | 中国人民解放军92728部队 | Automatic score obtaining method for unmanned helicopter flight control system |
CN114459644A (en) * | 2021-12-30 | 2022-05-10 | 南京航空航天大学 | Undercarriage drop load identification method based on optical fiber strain response and Gaussian process |
CN115630531A (en) * | 2022-12-19 | 2023-01-20 | 中南大学 | Automated safety assessment method for unmanned aerial vehicle control system |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20170124615A1 (en) * | 2015-11-04 | 2017-05-04 | Wal-Mart Stores, Inc. | Systems, method, and non-transitory computer-readable storage media for evaluating, storing, and managing labels for classification model evaluation and training |
CN108073088A (en) * | 2017-12-21 | 2018-05-25 | 天津职业技术师范大学 | Unmanned aerial vehicle control system assessment and Simulation Methods based on interior dynamic mapping |
CN109711715A (en) * | 2018-12-25 | 2019-05-03 | 国家电网有限公司 | Electric power telecommunication network risk assessment method |
CN109885907A (en) * | 2019-01-29 | 2019-06-14 | 南京航空航天大学 | A kind of Satellite Attitude Control System health state evaluation and prediction technique based on cloud model |
-
2020
- 2020-05-06 CN CN202010372798.8A patent/CN111680875B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20170124615A1 (en) * | 2015-11-04 | 2017-05-04 | Wal-Mart Stores, Inc. | Systems, method, and non-transitory computer-readable storage media for evaluating, storing, and managing labels for classification model evaluation and training |
CN108073088A (en) * | 2017-12-21 | 2018-05-25 | 天津职业技术师范大学 | Unmanned aerial vehicle control system assessment and Simulation Methods based on interior dynamic mapping |
CN109711715A (en) * | 2018-12-25 | 2019-05-03 | 国家电网有限公司 | Electric power telecommunication network risk assessment method |
CN109885907A (en) * | 2019-01-29 | 2019-06-14 | 南京航空航天大学 | A kind of Satellite Attitude Control System health state evaluation and prediction technique based on cloud model |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112463853A (en) * | 2020-11-27 | 2021-03-09 | 重庆医药高等专科学校 | Financial data behavior screening working method through cloud platform |
CN112463853B (en) * | 2020-11-27 | 2022-06-17 | 重庆医药高等专科学校 | Financial data behavior screening working method through cloud platform |
CN112465145A (en) * | 2020-12-02 | 2021-03-09 | 西北工业大学 | Unmanned cluster intelligent qualitative evaluation method based on logical reasoning and fuzzy synthesis |
CN112465145B (en) * | 2020-12-02 | 2024-04-05 | 西北工业大学 | Unmanned cluster intelligent qualitative evaluation method based on logic reasoning and fuzzy synthesis |
CN113031647A (en) * | 2021-02-25 | 2021-06-25 | 浙江工业大学 | Power supply type unmanned aerial vehicle optimal path planning method based on fuzzy comprehensive evaluation |
CN114398769A (en) * | 2021-12-29 | 2022-04-26 | 中国人民解放军92728部队 | Automatic score obtaining method for unmanned helicopter flight control system |
CN114459644A (en) * | 2021-12-30 | 2022-05-10 | 南京航空航天大学 | Undercarriage drop load identification method based on optical fiber strain response and Gaussian process |
CN115630531A (en) * | 2022-12-19 | 2023-01-20 | 中南大学 | Automated safety assessment method for unmanned aerial vehicle control system |
Also Published As
Publication number | Publication date |
---|---|
CN111680875B (en) | 2022-11-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111680875B (en) | Unmanned aerial vehicle state risk fuzzy comprehensive evaluation method based on probability baseline model | |
CN109766583B (en) | Aircraft engine life prediction method based on unlabeled, unbalanced and initial value uncertain data | |
CN107941537B (en) | A kind of mechanical equipment health state evaluation method | |
CN113837477B (en) | Method, device and equipment for predicting power grid faults under typhoon disasters driven by data | |
CN105973594B (en) | A kind of rolling bearing fault Forecasting Methodology based on continuous depth confidence network | |
CN111368885B (en) | Gas circuit fault diagnosis method for aircraft engine | |
CN109829236A (en) | A kind of Compressor Fault Diagnosis method based on XGBoost feature extraction | |
CN112508442B (en) | Transient stability assessment method and system based on automatic and interpretable machine learning | |
CN109187025A (en) | A kind of rolling bearing remaining life prediction technique of integrated KELM | |
An et al. | A data-driven method for transient stability margin prediction based on security region | |
CN112069727B (en) | Intelligent transient stability evaluation system and method with high reliability for power system | |
CN115048874B (en) | Aircraft design parameter estimation method based on machine learning | |
CN112557826A (en) | Ship electric power system fault diagnosis method | |
CN111523778A (en) | Power grid operation safety assessment method based on particle swarm algorithm and gradient lifting tree | |
CN111190349A (en) | Method, system and medium for monitoring state and diagnosing fault of ship engine room equipment | |
CN110737976A (en) | mechanical equipment health assessment method based on multi-dimensional information fusion | |
CN111273125A (en) | RST-CNN-based power cable channel fault diagnosis method | |
CN113225346A (en) | Network operation and maintenance situation assessment method based on machine learning | |
CN112257914A (en) | Aviation safety cause and effect prediction method based on random forest | |
Li et al. | Intelligent fault diagnosis of aeroengine sensors using improved pattern gradient spectrum entropy | |
CN115587290A (en) | Aero-engine fault diagnosis method based on variational self-coding generation countermeasure network | |
CN115600088A (en) | Distribution transformer fault diagnosis method based on vibration signals | |
CN115526258A (en) | Power system transient stability evaluation method based on Spearman correlation coefficient feature extraction | |
CN116956215A (en) | Fault diagnosis method and system for transmission system | |
Li et al. | Aero-engine exhaust gas temperature prediction based on LightGBM optimized by improved bat algorithm |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |