CN116644673A - State data time sequence change prediction method of aircraft thermal management system - Google Patents
State data time sequence change prediction method of aircraft thermal management system Download PDFInfo
- Publication number
- CN116644673A CN116644673A CN202310926491.1A CN202310926491A CN116644673A CN 116644673 A CN116644673 A CN 116644673A CN 202310926491 A CN202310926491 A CN 202310926491A CN 116644673 A CN116644673 A CN 116644673A
- Authority
- CN
- China
- Prior art keywords
- thermal management
- management system
- aircraft
- neural network
- fuel
- 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
- 238000000034 method Methods 0.000 title claims abstract description 46
- 230000008859 change Effects 0.000 title claims abstract description 33
- 238000004088 simulation Methods 0.000 claims abstract description 27
- 238000012549 training Methods 0.000 claims abstract description 23
- 238000003062 neural network model Methods 0.000 claims abstract description 16
- 238000012216 screening Methods 0.000 claims abstract description 6
- 239000000446 fuel Substances 0.000 claims description 28
- 238000013528 artificial neural network Methods 0.000 claims description 17
- 230000004044 response Effects 0.000 claims description 14
- 238000005070 sampling Methods 0.000 claims description 11
- 230000002123 temporal effect Effects 0.000 claims description 10
- 239000000295 fuel oil Substances 0.000 claims description 7
- 238000010219 correlation analysis Methods 0.000 claims description 6
- 238000005057 refrigeration Methods 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 5
- 238000004458 analytical method Methods 0.000 claims description 4
- 230000007613 environmental effect Effects 0.000 claims description 4
- 230000001050 lubricating effect Effects 0.000 claims description 4
- 238000012795 verification Methods 0.000 claims description 4
- 238000012897 Levenberg–Marquardt algorithm Methods 0.000 claims description 3
- 238000001914 filtration Methods 0.000 claims description 2
- 238000005461 lubrication Methods 0.000 claims description 2
- 238000013461 design Methods 0.000 abstract description 8
- 238000002485 combustion reaction Methods 0.000 abstract description 4
- 238000012360 testing method Methods 0.000 description 10
- 238000000540 analysis of variance Methods 0.000 description 9
- 230000035945 sensitivity Effects 0.000 description 6
- 238000010276 construction Methods 0.000 description 5
- 238000010206 sensitivity analysis Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 238000012546 transfer Methods 0.000 description 4
- 238000013400 design of experiment Methods 0.000 description 3
- 239000002828 fuel tank Substances 0.000 description 3
- 230000002596 correlated effect Effects 0.000 description 2
- 238000001704 evaporation Methods 0.000 description 2
- 230000008020 evaporation Effects 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000013178 mathematical model Methods 0.000 description 2
- 238000005381 potential energy Methods 0.000 description 2
- 230000000306 recurrent effect Effects 0.000 description 2
- 238000000611 regression analysis Methods 0.000 description 2
- 238000010200 validation analysis Methods 0.000 description 2
- 238000001134 F-test Methods 0.000 description 1
- 238000004378 air conditioning Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 239000003921 oil Substances 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000036962 time dependent Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/26—Discovering frequent patterns
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/27—Regression, e.g. linear or logistic regression
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
- G06F30/15—Vehicle, aircraft or watercraft design
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/08—Thermal analysis or thermal optimisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2123/00—Data types
- G06F2123/02—Data types in the time domain, e.g. time-series data
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Geometry (AREA)
- Software Systems (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Computer Hardware Design (AREA)
- Health & Medical Sciences (AREA)
- Aviation & Aerospace Engineering (AREA)
- Medical Informatics (AREA)
- Automation & Control Theory (AREA)
- Mathematical Optimization (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Pure & Applied Mathematics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The application provides a state data time sequence change prediction method of an aircraft thermal management system, which comprises the following steps: constructing a simulation model of an aircraft thermal management system; screening input factors based on the simulation model, and constructing a data set; training a neural network model NARX based on the data set; based on the trained neural network model NARX, predicting the state data time sequence change of the mechanical thermal management system. The method can realize the real-time dynamic prediction of the internal combustion oil temperature of the aircraft oil tank under the unknown flight mission section, provides reference for aircraft design, and has engineering practical value and theoretical significance.
Description
Technical Field
The application belongs to the technical field of aircraft thermal management systems, and particularly relates to a state data time sequence change prediction method of an aircraft thermal management system.
Background
The aircraft system heat management is to determine mass transfer and heat transfer boundary conditions from the system angle, and on the basis of maintaining the heat balance of a heat source and a heat sink, the system efficiency is improved to the greatest extent possible and useful work is output, so that the requirements of flight tasks, flight safety, life cycle cost and the like are met.
The thermal management system involves integration between different systems, thermal balance between different heat sources and heat sinks, covering a large number of timing parameters, typically requiring millions of different system cycles to run to cover the full design space. Traditional modeling simulation methods (including one-dimensional simulation and three-dimensional simulation) based on physics are widely applied to the simulation problem of a thermal management system.
Based on the physical traditional modeling simulation method, the existing defects are mainly represented in the following 2 aspects: (1) Aircraft thermal management system simulations often require consideration of predictions of thermal data from different unknown flight profile states, and physics-based modeling simulations are difficult to generalize. (2) Forward prediction requires modeling simulations to run faster than real-time events, otherwise the meaning of prediction is lost. However, the operation speed of the physical-based modeling is usually at the cost of solving the thermodynamic state equation, and the operation speed of a system which is more accurate and complex is slower, so that the physical-based modeling is difficult to meet the requirement.
Disclosure of Invention
In order to solve the technical problems, the application provides a state data time sequence change prediction method of an aircraft thermal management system, which can realize real-time dynamic prediction of the fuel temperature of an aircraft fuel tank under an unknown flight mission section, provides references for aircraft design, and has engineering practical value and theoretical significance.
In order to achieve the above object, the present application provides a method for predicting time sequence change of state data of an aircraft thermal management system, which is characterized by comprising:
constructing a simulation model of an aircraft thermal management system;
screening input factors based on the simulation model, and constructing a data set;
training a neural network model NARX based on the data set;
and predicting the time sequence change of the state data of the aircraft thermal management system based on the trained neural network model NARX.
Optionally, the aircraft thermal management system comprises: a fuel system, an engine system, an environmental control system, and a refrigeration system;
the hydraulic system and the electric system are used as boundary conditions of the simulation model in the process of constructing the simulation model.
Optionally, constructing the dataset includes:
based on the simulation model, acquiring an initial input factor and an initial output response; wherein the initial input factors include: the method comprises the steps of (1) determining the height of a flight envelope, mach number of the flight envelope, thermal load of a hydraulic system, thermal load of an air system, thermal load of an electronic equipment system, thermal load of a lubricating system, fuel consumption rate of a fuel system, mass flow rate of the hydraulic system, mass flow rate of the electronic equipment system and mass flow rate of the lubricating system, and outputting response to be average temperature of fuel;
screening the initial input factors to obtain final input factors; wherein the final input factors include: the altitude of the flight envelope, the Mach number of the flight envelope, the thermal load of the hydraulic system, the thermal load of the air system, the thermal load of the electronic equipment system, the thermal load of the lubrication system and the fuel consumption rate of the engine system;
sampling the final input factor;
the data set is constructed based on the sampled final input factor and the output response.
Optionally, filtering the initial input factor includes:
and performing correlation analysis on each two initial input factors, performing variance analysis on the output response and each initial input factor, and removing weak correlation parameters in the initial input factors.
Optionally, the correlation analysis is: measuring the linear relation between the two initial input factors based on a correlation coefficient r;
the correlation coefficient r is:
wherein n is the number of verification points, X i As the true value of the variable X,is the mean value of the variable X, Y i For the true value of the variable Y, +.>Is the mean of the variable Y.
Optionally, the final input factor includes: maximum and minimum values of the factor;
the final input factor is a representation form of a multi-scale parameter space; wherein each point in the multi-scale parameter space represents a flight state, i.e., a data point for neural network training.
Optionally, sampling includes:
and generating sampling points for the final input factors by using a space filling Latin hypercube method, and adopting a step function as a time mode between the sampling points.
Optionally, training the neural network model NARX includes:
training a neural network model NARX by adopting a Levenberg-Marquardt algorithm, and evaluating the trained model by adopting a mean square error and a decision coefficient.
Optionally, the mathematical expression of the neural network model NARX is:
where y (t) is the output at time t, f is the function to which the neural network is fitted, x (t) is the input at time t, and t is time.
Optionally, obtaining the initial input factor and output response includes:
based on the simulation model, acquiring an initial input factor and an initial output response by using an energy balance equation;
the energy balance equation is:
wherein dT is the temperature change of the fuel,for the total energy change in the tank +.>Is the specific heat of the fuel oil,for the fuel mass in the tank>For heat change, ++>For energy change, +.>For fuel quality variation, h is enthalpy and subscripts 1 and 2 represent inlet and outlet, respectively.
Compared with the prior art, the application has the following advantages and technical effects:
1. the forward prediction method for the thermal management of the aircraft can realize the real-time dynamic prediction of the temperature of the fuel in the fuel tank of the aircraft under the unknown flight mission section, provides references for the design of the aircraft, and has engineering practical value and theoretical significance.
2. According to the forward prediction method for the thermal management of the aircraft, based on the state data of the thermal management system of the aircraft with high fidelity, the error between the obtained fuel temperature prediction result and the actual result is within 1%, the simulation speed is faster than that of the actual event, the calculated amount of the simulation process is saved, and the efficiency is greatly improved.
3. The forward prediction method for the aircraft thermal management provides a modeling thought based on high-fidelity aircraft thermal management system data instead of actual flight data, greatly avoids the problems that the flight data sample size is small and the acquisition is difficult, and can obtain a high-precision prediction model.
Drawings
The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification, illustrate embodiments of the application and together with the description serve to explain the application. In the drawings:
FIG. 1 is a schematic diagram of a thermal management system according to an embodiment of the present application;
FIG. 2 is a schematic diagram of an input parameter sequence according to an embodiment of the present application;
FIG. 3 is a histogram of error distribution according to an embodiment of the present application;
FIG. 4 is a graph of temperature prediction results according to an embodiment of the present application; fig. 4a is a schematic diagram of a part of predicted values and target values, and fig. 4b is a schematic diagram of an error variation in the whole flight profile;
fig. 5 is a flow chart of a method for predicting time sequence change of state data of an aircraft thermal management system according to an embodiment of the application.
Detailed Description
It should be noted that, without conflict, the embodiments of the present application and features of the embodiments may be combined with each other. The application will be described in detail below with reference to the drawings in connection with embodiments.
It should be noted that the steps illustrated in the flowcharts of the figures may be performed in a computer system such as a set of computer executable instructions, and that although a logical order is illustrated in the flowcharts, in some cases the steps illustrated or described may be performed in an order other than that illustrated herein.
As shown in FIG. 5, the present embodiment provides a method for predicting time sequence variation of status data of an aircraft thermal management system, which generally includes constructing a simulation model of the aircraft thermal management system; screening input factors based on a simulation model, and constructing a data set; training a neural network model NARX based on the data set; based on the trained neural network model NARX, predicting the state data time sequence change of the aircraft thermal management system.
The embodiment can be divided into four steps of high-fidelity model construction, relevant factor selection, test creation design and neural network construction. The first three steps are to prepare the data for the appropriate training at a later time. Finally, the NARX model was trained, tested and compared for predictive performance.
The first part, construction of a high-fidelity model: the high-fidelity model is based on a thermal management system consisting of four systems, namely a fuel oil system, an engine system, an environment control system and a refrigeration system. However, other systems interact with thermal management systems, such as hydraulic, electrical systems, and the like. These systems do affect the associated thermal management system, but their effect is negligible in terms of their complex internal structure. They are therefore considered boundary conditions.
And a second part for selecting input parameters: based on the established high-fidelity model, a relevant input (factor) and a relevant output (response) are selected. To remove input parameters that may not have significant impact, two cases need to be considered. The influence of the parameters is likely to be weak or replaceable. Thus, the sensitivity and correlation of the input parameters should be analyzed. Sensitivity analysis is a method of studying and analyzing the sensitivity of a mathematical model or system to output changes under system parameters or ambient conditions. It is an effective means of removing weakly correlated parameters. There are several more popular sensitivity analysis methods such as regression analysis, analysis of variance (ANOVA) and fourier amplitude sensitivity tests. Current work chooses analysis of variance. By doing so, a multi-scale parameter space is created. Each point in this space represents a flight state, i.e., a data point for neural network training. Accordingly, a test design method is presented herein.
Third part, test design: a design of experiment (DoE) is a set of experiments created to gather the maximum amount of information about a system while reducing the total number of experiments for a given system. The application provides a feasible method of a forward prediction model based on NARX. Since this is a time-dependent problem, two methods of processing time can be used. One is to take time as an additional input, the other is to take time as a hidden variable, represented by other parameters, called space filling. The first method directly relates time to response, possibly resulting in over-fitting problems. Thus, the space filling method is selected. For other inputs, the 2-level full factorial and space-filling Latin hypercube approach is selected. The temporal pattern between data points may be selected as a step function, linear or other type. The present application selects a step time mode.
Fourth part, construction of neural network: the neural network model NARX (nonlinear autoregressive exogenous model) is trained using the sample points obtained in the previous section. NARX is a Recurrent Neural Network (RNN) that predicts as a "sequence-to-sequence" type for a data structure inherent to time-sequence information. NARX is a relatively simple model. It uses a feedback loop to insert the previous output into the input, building the relationship between the current output y (t) and the outputs y before the inputs u and t, namely:
however, if each output is used as an input for future predictions, this structure is very bulky. Thus, the time delay is used to determine the number of backoff steps predicted for the next time step. The method is based on the markov assumption that thermal trajectory trends can be predicted by considering only neighboring data. Thus, the model is very effective when the data has suitable markov properties.
The first part, establishment of a high simulation model:
the high-fidelity model is based on a thermal management system consisting of four systems, namely a fuel oil system, an engine system, an environment control system and a refrigeration system. However, other systems interact with thermal management systems, such as hydraulic, electrical systems, and the like. These systems do affect the associated thermal management system, but their effect is negligible in terms of their complex internal structure. They are therefore considered boundary conditions. A top-level view of a thermal management integrated system is shown in fig. 1.
Wherein the fuel tank acts as a heat sink and interacts with other systems. The fuel is pumped from the tank and in turn absorbs heat from the engine, environmental control and refrigeration systems. And one part of fuel oil is cooled by ram air and then returns to the oil tank, and the other part of fuel oil is combusted in the combustion chamber. The refrigeration system consists of three cycles, a hot PAO cycle, an evaporation cycle and a cold PAO cycle. The thermal load from the electronics and cabin is added to the PAO cold cycle, which is transported to the fuel by the evaporation cycle, the hot PAO cycle. Engine systems play a critical role in providing thrust that requires fuel to reach a specific flow, pressure, and temperature before combustion in the combustion chamber. The environmental control system is a simplified aircraft air conditioning system. The system is intended to provide air supply, thermal control and pressurization for passengers and crewmembers.
In order to select relevant parameters from the system perspective, the mass and heat transfer processes of the thermal management system are analyzed, and the tank is an open system according to the first law of thermodynamics. The energy balance equation can be expressed as:
（1）
here, theIs a change of heat->Is the shaft work, the->Is the change of energy, +.>For fuel quality variation, h is enthalpy, +.>Is kinetic energy>Is gravitational potential energy. Subscripts 1 and 2 represent the inlet and outlet, respectively; for the thermal management system in the figure, the kinetic energy variation and for negligible velocity, the gravitational potential energy variation and the position fluctuation can be ignored. The shaft work is zero, and the formula (1) can be simplified as follows:
（2）
（3）
where dT is the change in temperature of the fuel,for the total energy change in the tank +.>Specific heat of fuel oil>Is the mass of fuel in the tank. From the formulas (2) and (3), it is possible to obtain:
。
second, selection of relevant factors:
based on the mass and heat transfer process analysis of the previous section, relevant inputs (factors) were selected, and the original impact parameters are shown in Table 1.
Table 1 original parameter table
To remove input parameters that may not have significant impact, two cases need to be considered. The influence of the parameters is likely to be weak or replaceable. Thus, sensitivity and correlation should be analyzed. Sensitivity analysis is a method of studying and analyzing the sensitivity of a mathematical model or system to output changes under system parameters or ambient conditions. It is an effective means of removing weakly correlated parameters. There are several commonly used sensitivity analysis methods, such as regression analysis, analysis of variance (ANOVA), fourier amplitude sensitivity test, etc. The present application selects an analysis of variance for sensitivity analysis characterized by F-test or concomitance probability p, with the criteria for significance level as shown in table 2.
Table 2 significance level criteria table
The results of the analysis of variance are shown in table 3:
TABLE 3 analysis of variance table
To screen out independent factors, correlation analysis was performed on each parameter in the table. The correlation coefficient r measures the linear relation between two factors, and the calculation formula is as follows:
the results of the correlation analysis are shown in table 4:
table 4 related parameter table
Based on the correlation matrix and analysis of variance results, the final inputs are filtered as shown in Table 5, with the maximum and minimum values listed in the table. By doing so, a multi-scale parameter space is created. Each point in this space represents a flight state, i.e., a data point for neural network training.
TABLE 5 influence parameter Table
Third, design created by trial:
in the table, the input parameters involved have been given a maximum value and a minimum value. By doing so, a multi-scale parameter space is created. Each point in space represents experimental data for one flight state, i.e., one neural network training. To construct the next neural network training dataset, sampling points are generated using a space-filling Latin hypercube method.
Up to now, a training set of discrete sample points has been created. For predictive problems, the data structure of the input parameters should be a time-varying sequence. The sampling points are connected by a certain time mode. Several temporal patterns between sampling points may be selected, such as step functions, linear or other types. A step function is selected as the temporal pattern between the data points. The input schedule is shown in fig. 2.
Fourth part, construction of neural network:
the neural network model NARX (nonlinear autoregressive exogenous model) is trained using the DoE method. NARX is a Recurrent Neural Network (RNN) that predicts as a "sequence-to-sequence" type for a data structure inherent to time-sequence information. The NRAX neural network uses a feedback loop to insert the previous output into the input, building the relationship between the current output y (t) and the outputs y before the inputs u and t, namely:
however, if each output is used as an input for future predictions, this structure is very bulky. Thus, the time delay is used to determine the number of backoff steps predicted for the next time step. The method is based on the markov assumption that only neighboring data is considered to interpret the history effect. Thus, it is very useful when the data has suitable markov properties.
Fifth section, result analysis:
the time series prediction problem employs a NARX neural network. The fuel system is used for evaluation. The training set is generated by a high-fidelity model comprising a 200 x 1000s time series. 200 sample points are selected between the maximum and minimum values, and each state is assumed to last 1000 seconds. The pattern of change between two state points is designated as a step change. The test set is generated according to the flight profile in the figure, and the fuel temperature is used as the only output parameter.
For NARX, the number of neurons per layer was 10, and the delay was set to 2. The neural network gradient descent algorithm adopts a Levenberg-Marquardt algorithm. The training set is divided into a training subset, a validation subset, and a test subset of 70:15:15. If one of the target values is reached, training is stopped. In this case, training is stopped when the verification check has been performed six times, which means that the verification subset error rate continues to increase over 6 cycles.
NARX can be evaluated by Mean Square Error (MSE) and a decision coefficient R, where R and MSE are calculated as
Where y is the target temperature and where,for NRAX output, < >>And N is the number of test points. The sum of the mean square errors for the three data sets is shown in table 6, indicating a higher model accuracy.
TABLE 6 training results Table
To measure the NARX fitting accuracy, the R value is calculated in the table. The R values on three different data sets can be seen, training set, validation set, test set and total data set. The closer the R value is to 1, the better the model fit. The error distribution is shown in fig. 3, and in fig. 3, the error concentration is seen to be distributed around zero, indicating high accuracy over three data sets.
Figure 3 shows the test errors based on the flight profile in the figure. FIG. 4 shows target and output fuel temperatures. In fig. 4a, the square block represents the predicted value output by the neural network, the circle represents the actual target value of the output parameter, the region between the predicted value and the target value represents the error, and it can be seen that the dashed line represents the temperature change curve formed by all the predicted values in the enlarged view. Since the error is relatively small, only part of the predicted and target values are shown in fig. 4a, and the error variation over the entire flight profile is shown in fig. 4 b. The result shows that the NARX model has better prediction capability, and the error in the whole flight section is within 0.03 ℃.
The present application is not limited to the above-mentioned embodiments, and any changes or substitutions that can be easily understood by those skilled in the art within the technical scope of the present application are intended to be included in the scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.
Claims (10)
1. A method for predicting a time-series change of state data of an aircraft thermal management system, comprising:
constructing a simulation model of an aircraft thermal management system;
screening input factors based on the simulation model, and constructing a data set;
training a neural network model NARX based on the data set;
and predicting the time sequence change of the state data of the aircraft thermal management system based on the trained neural network model NARX.
2. The method for predicting a temporal change in status data of an aircraft thermal management system of claim 1, wherein the aircraft thermal management system comprises: a fuel system, an engine system, an environmental control system, and a refrigeration system;
the hydraulic system and the electric system are used as boundary conditions of the simulation model in the process of constructing the simulation model.
3. The method of claim 1, wherein constructing the data set comprises:
based on the simulation model, acquiring an initial input factor and an initial output response; wherein the initial input factors include: the method comprises the steps of (1) determining the height of a flight envelope, mach number of the flight envelope, thermal load of a hydraulic system, thermal load of an air system, thermal load of an electronic equipment system, thermal load of a lubricating system, fuel consumption rate of a fuel system, mass flow rate of the hydraulic system, mass flow rate of the electronic equipment system and mass flow rate of the lubricating system, and outputting response to be average temperature of fuel;
screening the initial input factors to obtain final input factors; wherein the final input factors include: the altitude of the flight envelope, the Mach number of the flight envelope, the thermal load of the hydraulic system, the thermal load of the air system, the thermal load of the electronic equipment system, the thermal load of the lubrication system and the fuel consumption rate of the engine system;
sampling the final input factor;
the data set is constructed based on the sampled final input factor and the output response.
4. A method of predicting a temporal change in status data of an aircraft thermal management system as claimed in claim 3, wherein the filtering of the initial input factors comprises:
and performing correlation analysis on each two initial input factors, performing variance analysis on the output response and each initial input factor, and removing weak correlation parameters in the initial input factors.
5. The method for predicting a temporal change in state data of an aircraft thermal management system of claim 4,
the correlation analysis is as follows: measuring the linear relation between the two initial input factors based on a correlation coefficient r;
the correlation coefficient r is:
wherein n is the number of verification points, X i Is the true value of the variable X, +.>Is the mean value of the variable X, Y i For the true value of the variable Y, +.>Is the mean of the variable Y.
6. A method of predicting a temporal change in state data of an aircraft thermal management system according to claim 3, wherein the final input factor comprises: maximum and minimum values of the factor;
the final input factor is a representation form of a multi-scale parameter space; wherein each point in the multi-scale parameter space represents a flight state, i.e., a data point for neural network training.
7. A method of predicting a temporal change in status data of an aircraft thermal management system as claimed in claim 3, wherein sampling comprises:
and generating sampling points for the final input factors by using a space filling Latin hypercube method, and adopting a step function as a time mode between the sampling points.
8. The method of claim 1, wherein training the neural network model NARX comprises:
training a neural network model NARX by adopting a Levenberg-Marquardt algorithm, and evaluating the trained model by adopting a mean square error and a decision coefficient.
9. The method for predicting a temporal change in status data of an aircraft thermal management system according to claim 1, wherein the mathematical expression of the neural network model NARX is:
where y (t) is the output at time t, f is the function to which the neural network is fitted, x (t) is the input at time t, and t is time.
10. A method of predicting a temporal change in status data of an aircraft thermal management system as in claim 3, wherein obtaining initial input factors and output responses comprises:
based on the simulation model, acquiring an initial input factor and an initial output response by using an energy balance equation;
the energy balance equation is:
wherein dT is the fuel temperature change, +.>For the total energy change in the tank +.>Specific heat of fuel oil>For the fuel mass in the tank>For heat change, ++>As a function of energyTransform (alleviate) the wind>For fuel quality variation, h is enthalpy and subscripts 1 and 2 represent inlet and outlet, respectively.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310926491.1A CN116644673B (en) | 2023-07-27 | 2023-07-27 | State data time sequence change prediction method of aircraft thermal management system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310926491.1A CN116644673B (en) | 2023-07-27 | 2023-07-27 | State data time sequence change prediction method of aircraft thermal management system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116644673A true CN116644673A (en) | 2023-08-25 |
| CN116644673B CN116644673B (en) | 2023-10-20 |
Family
ID=87619211
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310926491.1A Active CN116644673B (en) | 2023-07-27 | 2023-07-27 | State data time sequence change prediction method of aircraft thermal management system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116644673B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111815056A (en) * | 2020-07-10 | 2020-10-23 | 中国人民解放军空军工程大学 | Aircraft external field aircraft fuel system fault prediction method based on flight parameter data |
| CN112862164A (en) * | 2021-01-22 | 2021-05-28 | 桂林电子科技大学 | Dry clutch temperature prediction method based on dynamic neural network time sequence prediction |
| CN114154234A (en) * | 2021-11-04 | 2022-03-08 | 中国人民解放军海军航空大学青岛校区 | Modeling method, system and storage medium for aircraft engine |
| US20220414283A1 (en) * | 2021-06-23 | 2022-12-29 | The Boeing Company | Predictive Modeling of Aircraft Dynamics |
-
2023
- 2023-07-27 CN CN202310926491.1A patent/CN116644673B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111815056A (en) * | 2020-07-10 | 2020-10-23 | 中国人民解放军空军工程大学 | Aircraft external field aircraft fuel system fault prediction method based on flight parameter data |
| CN112862164A (en) * | 2021-01-22 | 2021-05-28 | 桂林电子科技大学 | Dry clutch temperature prediction method based on dynamic neural network time sequence prediction |
| US20220414283A1 (en) * | 2021-06-23 | 2022-12-29 | The Boeing Company | Predictive Modeling of Aircraft Dynamics |
| CN114154234A (en) * | 2021-11-04 | 2022-03-08 | 中国人民解放军海军航空大学青岛校区 | Modeling method, system and storage medium for aircraft engine |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116644673B (en) | 2023-10-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Cortés et al. | Optimization of operating conditions for compressor performance by means of neural network inverse | |
| CN113591215B (en) | Abnormal satellite component layout detection method based on uncertainty | |
| CN107491840B (en) | Flow wear characteristic prediction and service life evaluation method based on ELM neural network model | |
| Ledesma et al. | Analysis and modeling of a variable speed reciprocating compressor using ANN | |
| CN112084701A (en) | System transient temperature prediction method based on data driving | |
| Kumar et al. | Calibrating transient models with multiple responses using Bayesian inverse techniques | |
| CN116644673B (en) | State data time sequence change prediction method of aircraft thermal management system | |
| CN107061032B (en) | A kind of prediction technique and forecasting system of engine operating state | |
| Dilay et al. | The experimental validation of a transient power electronic building block (PEBB) mathematical model | |
| CN115762653B (en) | Fuel combustion mechanism optimization method based on evolutionary algorithm and deep learning | |
| Weiss et al. | Probabilistic finite-element analyses on turbine blades | |
| CN115982854A (en) | Design method applied to ablation type thermal protection system | |
| CN115470726A (en) | Hypersonic inlet channel flow field rapid prediction method based on deep learning | |
| US9245067B2 (en) | Probabilistic method and system for testing a material | |
| Kamtsiuris et al. | A health index framework for condition monitoring and health prediction | |
| Chen et al. | Design of intelligent acceleration schedules for extending the life of aircraft engines | |
| Abharian et al. | Power probability density function control and performance assessment of a nuclear research reactor | |
| Guo et al. | Grey self-memory combined model for complex equipment cost estimation | |
| Zhang et al. | Probabilistic invertible neural network for inverse design space exploration and reasoning | |
| Enright et al. | Micromechanics-based fracture risk assessment using integrated probabilistic damage tolerance analysis and manufacturing process models | |
| Pehlivanoglu et al. | Inverse design of 2-D airfoil via vibrational genetic algorithm | |
| Tang et al. | Dynamic Analysis and Prediction of Performance on the Thermal Management System of an Aero-Engine Based on Theneural Network of Long Short-Term Memory | |
| Kuznetsova et al. | Aeroengine NOx-emissions automatic control based on neural network model | |
| Chen et al. | Monte Carlo simulation for system damage prediction: an example from thermo-mechanical fatigue (TMF) damage for a turbine engine | |
| Omekanda et al. | Optimal Parameter Calibration for Physics Based Multi-Mass Engine Model |
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 |