Flight guarantee vehicle fault prediction method and device, electronic equipment and storage medium

Abstract

The invention provides a flight support vehicle fault prediction method and apparatus, an electronic device and a storage medium. The method comprises the steps of obtaining historical vehicle data of a flight support vehicle and a prediction probability that a fault state obtained by inputting the historical vehicle data into each fault prediction model is an actual fault state corresponding to the historical vehicle data; for each fault prediction model, calculating log-likelihood function values of all prediction probabilities obtained by the fault prediction model; calculating a fault prediction weight value corresponding to each fault prediction model according to the log-likelihood function values corresponding to all the fault prediction models; and performing fault prediction on the flight support vehicle according to the fault prediction weight value corresponding to each fault prediction model. Through the mode of the invention, fault prediction can be carried out on the flight support vehicle, so that the short maintenance resources are targeted to serve the vehicle which is urgently needed to be maintained, and the fault probability of the vehicle in actual operation is effectively reduced.

Description

Technical Field

[0001] The present invention relates to the field of flight ground support service scheduling, and in particular to a flight support vehicle fault prediction method, device, electronic equipment and storage medium. Background Art

[0002] Flight support vehicles undertake many important support links in flight ground support services, and due to the professionalism and irreplaceability of the equipment, support vehicles directly affect the safety and efficiency of flight ground support, and thus affect the normal operation of flights. Because ground support vehicles are special vehicles with many types of vehicles, strong professionalism and high safety requirements, vehicle failures will increase the risk of flight operations and reduce support operation efficiency. In order to keep support vehicles in a high-availability state, the vehicle management department performs manual maintenance on vehicles at a higher frequency.

[0003] However, the different purchase times, workloads and working environments of different support vehicles lead to large differences in the time of vehicle failure, and the efficiency of manual maintenance is not high. Therefore, the traditional method of regular manual maintenance cannot provide targeted services for scarce maintenance resources to vehicles that are in urgent need of maintenance, nor can it effectively reduce the probability of failure in actual vehicle operations. Summary of the invention

[0004] In view of this, the purpose of the present application is to provide a flight support vehicle fault prediction method, device, electronic device and storage medium, which can predict the failure of flight support vehicles, and then target the scarce maintenance resources to serve the vehicles that are really in urgent need of maintenance, thereby effectively reducing the failure probability of the vehicle in actual operation.

[0005] In a first aspect, an embodiment of the present application provides a method for predicting a flight support vehicle failure, the method comprising:

[0006] Obtaining historical vehicle data of flight support vehicles and inputting the historical vehicle data into each fault prediction model to obtain a fault state that is a predicted probability of an actual fault state corresponding to the historical vehicle data; the fault prediction model is trained based on the historical vehicle data and the corresponding actual fault state;

[0007] For each fault prediction model, calculate the log-likelihood function values ​​of all prediction probabilities obtained by the fault prediction model;

[0008] According to the log-likelihood function values ​​corresponding to all fault prediction models, the fault prediction weight value corresponding to each fault prediction model is calculated;

[0009] Fault prediction is performed on flight support vehicles according to the fault prediction weight values ​​corresponding to each fault prediction model.

[0010] In a possible implementation, calculating the log-likelihood function values ​​of all predicted probabilities obtained by the fault prediction model includes:

[0011] The sum of the logarithms of all predicted probabilities obtained by the fault prediction model is determined as the log-likelihood function value of all predicted probabilities obtained by the fault prediction model.

[0012] In a possible implementation, calculating the fault prediction weight value corresponding to each fault prediction model according to the log-likelihood function values ​​corresponding to all fault prediction models includes:

[0013] Substitute the log-likelihood function values ​​corresponding to all fault prediction models into the following formula to calculate the fault prediction weight value corresponding to each fault prediction model;

[0014]

[0015] Among them, P(M i |X) is the fault prediction weight value corresponding to the i-th fault prediction model, M i is the i-th fault prediction model, X is the set of historical vehicle data, P(X|M i ) is the log-likelihood function value corresponding to the i-th fault prediction model, P(M i ) is the prior fault prediction weight value corresponding to the i-th fault prediction model, Q is the number of fault prediction models, P(X|M j ) is the log-likelihood function value corresponding to the j-th fault prediction model, P(M j ) is the prior fault prediction weight value corresponding to the j-th fault prediction model.

[0016] In a possible implementation, the priori prediction weight values ​​corresponding to each fault prediction model are calculated by the following steps:

[0017] Obtaining a first fault prediction weight value determined based on the actual prediction accuracy of the model and a second fault prediction weight value obtained based on expert scoring corresponding to each fault prediction model;

[0018] Calculating the sum of all first fault prediction weight values ​​and all second fault prediction weight values ​​to obtain a first sum value;

[0019] For each fault prediction model, calculating the sum of a first fault prediction weight value and a second fault prediction weight value corresponding to the fault prediction model to obtain a second sum value;

[0020] The ratio of the second sum value to the first sum value is determined as the priori prediction weight value corresponding to the fault prediction model.

[0021] In a possible implementation, fault prediction is performed on a flight support vehicle according to the fault prediction weight value corresponding to each fault prediction model, including:

[0022] Get the current vehicle data of flight support vehicles;

[0023] Input the current vehicle data into each fault prediction model to obtain the initial prediction probability corresponding to each fault state;

[0024] According to each fault state, all initial prediction probabilities corresponding to the fault state are weighted and summed according to the prediction weight values ​​corresponding to each fault prediction model to obtain the target prediction probability corresponding to the fault state;

[0025] The fault state with the highest target prediction probability is determined as the predicted fault state of the flight support vehicle.

[0026] In a possible implementation, obtaining historical vehicle data of flight support vehicles includes:

[0027] Obtain initial historical vehicle data corresponding to each vehicle data type within a preset time period of flight support vehicles;

[0028] For each vehicle data type, the initial historical vehicle data corresponding to the vehicle data type is window-slid, and the standard deviation value of the initial historical vehicle data in each window is calculated; and each window carries a sliding sequence number;

[0029] The set of standard deviation values ​​corresponding to windows with the same serial number is determined as the historical vehicle data of the flight support vehicle.

[0030] In a possible implementation, before performing window sliding on the initial historical vehicle data corresponding to the vehicle data type and calculating the standard deviation value of the initial historical vehicle data in each window, the method further includes:

[0031] Fill missing values ​​in the initial historical vehicle data corresponding to each vehicle data type;

[0032] The initial historical vehicle data corresponding to each vehicle data type after the missing values ​​are filled is standardized to obtain the final initial historical vehicle data.

[0033] In a second aspect, the embodiment of the present application further provides a flight support vehicle failure prediction device, the device comprising:

[0034] An acquisition module is used to acquire historical vehicle data of flight support vehicles and input the historical vehicle data into each fault prediction model to obtain a predicted probability that the fault state obtained is the actual fault state corresponding to the historical vehicle data; the fault prediction model is trained based on the historical vehicle data and the corresponding actual fault state;

[0035] A calculation module, used for calculating, for each fault prediction model, the log-likelihood function values ​​of all prediction probabilities obtained by the fault prediction model;

[0036] The calculation module is further used to calculate the fault prediction weight value corresponding to each fault prediction model according to the log-likelihood function values ​​corresponding to all fault prediction models;

[0037] The prediction module is used to predict the faults of flight support vehicles according to the fault prediction weight values ​​corresponding to each fault prediction model.

[0038] In a possible implementation, the calculation module is specifically configured to determine the sum of the logarithms of all predicted probabilities obtained by the fault prediction model as the log-likelihood function value of all predicted probabilities obtained by the fault prediction model.

[0039] In a possible implementation manner, the calculation module is specifically used to substitute the log-likelihood function values ​​corresponding to all fault prediction models into the following formula to calculate the fault prediction weight value corresponding to each fault prediction model;

[0040]

[0041] Among them, P(M i |X) is the fault prediction weight value corresponding to the i-th fault prediction model, M i is the i-th fault prediction model, X is the set of historical vehicle data, P(X|M i ) is the log-likelihood function value corresponding to the i-th fault prediction model, P(M i ) is the prior fault prediction weight value corresponding to the i-th fault prediction model, Q is the number of fault prediction models, P(X|M j ) is the log-likelihood function value corresponding to the j-th fault prediction model, P(M j ) is the prior fault prediction weight value corresponding to the j-th fault prediction model.

[0042] In a possible implementation, the computing module is further configured to:

[0043] Obtaining a first fault prediction weight value determined based on the actual prediction accuracy of the model and a second fault prediction weight value obtained based on expert scoring corresponding to each fault prediction model;

[0044] Calculating the sum of all first fault prediction weight values ​​and all second fault prediction weight values ​​to obtain a first sum value;

[0045] For each fault prediction model, calculating the sum of a first fault prediction weight value and a second fault prediction weight value corresponding to the fault prediction model to obtain a second sum value;

[0046] The ratio of the second sum value to the first sum value is determined as the priori prediction weight value corresponding to the fault prediction model.

[0047] In one possible implementation, the prediction module is specifically used to obtain current vehicle data of the flight support vehicle; input the current vehicle data into each fault prediction model to obtain the initial prediction probability corresponding to each fault state; according to each fault state, according to the prediction weight value corresponding to each fault prediction model, weighted sum all the initial prediction probabilities corresponding to the fault state to obtain the target prediction probability corresponding to the fault state; and determine the fault state with the largest target prediction probability as the predicted fault state of the flight support vehicle.

[0048] In one possible implementation, an acquisition module is specifically used to acquire initial historical vehicle data corresponding to each vehicle data type within a preset time period of a flight support vehicle; for each vehicle data type, a window is slid for the initial historical vehicle data corresponding to the vehicle data type, and a standard deviation value of the initial historical vehicle data in each window is calculated; and each window carries a sliding serial number; and a set of standard deviation values ​​corresponding to windows with the same serial number is determined as the historical vehicle data of the flight support vehicle.

[0049] In a possible implementation, before sliding the window of the initial historical vehicle data corresponding to the vehicle data type and calculating the standard deviation value of the initial historical vehicle data in each window, the acquisition module is further used to:

[0050] Fill missing values ​​in the initial historical vehicle data corresponding to each vehicle data type;

[0051] The initial historical vehicle data corresponding to each vehicle data type after the missing values ​​are filled is standardized to obtain the final initial historical vehicle data.

[0052] In a third aspect, an embodiment of the present application further provides an electronic device comprising: a processor, a storage medium and a bus, wherein the storage medium stores machine-readable instructions executable by the processor. When the electronic device is running, the processor and the storage medium communicate via the bus, and the processor executes the machine-readable instructions to perform the steps of any flight support vehicle fault prediction method of the first aspect.

[0053] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of any flight support vehicle failure prediction method of the first aspect are executed.

[0054] The embodiment of the present application provides a method, device, electronic device and storage medium for predicting faults of flight support vehicles. The method includes: obtaining historical vehicle data of flight support vehicles and inputting the historical vehicle data into each fault prediction model to obtain the predicted probability that the fault state is the actual fault state corresponding to the historical vehicle data; the fault prediction model is trained based on the historical vehicle data and the corresponding actual fault state. For each fault prediction model, the log-likelihood function value of all predicted probabilities obtained by the fault prediction model is calculated; according to the log-likelihood function values ​​corresponding to all fault prediction models, the fault prediction weight value corresponding to each fault prediction model is calculated; according to the fault prediction weight value corresponding to each fault prediction model, the flight support vehicle is predicted. The present application can predict faults of flight support vehicles by calculating the fault prediction weight value corresponding to each fault prediction model, and then use the scarce maintenance resources to serve the vehicles that are really in urgent need of maintenance, effectively reducing the probability of failure in the actual operation of the vehicle. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0056] Figure 1 A flowchart of a flight support vehicle failure prediction method provided by an embodiment of the present application is shown;

[0057] Figure 2 A flowchart of another method for predicting failures of flight support vehicles provided in an embodiment of the present application is shown;

[0058] Figure 3 A schematic diagram of the structure of a flight support vehicle fault prediction device provided in an embodiment of the present application is shown;

[0059] Figure 4 A schematic structural diagram of an electronic device provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0060] To make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. It should be understood that the drawings in the present application only serve the purpose of explanation and description and are not used to limit the scope of protection of the present application. In addition, it should be understood that the schematic drawings are not drawn in real proportion. The flowchart used in this application shows the operations implemented according to some embodiments of the present application. It should be understood that the operations of the flowchart can be implemented out of sequence, and the steps without logical context can be reversed in order or implemented simultaneously. In addition, those skilled in the art can add one or more other operations to the flowchart under the guidance of the content of the present application, or remove one or more operations from the flowchart.

[0061] In addition, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application.

[0062] In order to enable those skilled in the art to use the content of this application, the following implementation is provided in conjunction with a specific application scenario, "the field of flight ground support service scheduling". For those skilled in the art, the general principles defined herein can be applied to other embodiments and application scenarios without departing from the spirit and scope of this application. Although this application is mainly described around the "field of flight ground support service scheduling", it should be understood that this is only an exemplary embodiment.

[0063] It should be noted that the term "comprising" will be used in the embodiments of the present application to indicate the existence of the features declared thereafter, but does not exclude the addition of other features.

[0064] The following is a detailed description of a flight support vehicle failure prediction method provided in an embodiment of the present application.

[0065] Reference Figure 1 As shown, it is a flowchart of a method for predicting a flight support vehicle failure provided by an embodiment of the present application. The exemplary steps of the embodiment of the present application are described below:

[0066] S101, acquiring historical vehicle data of flight support vehicles and inputting the historical vehicle data into each fault prediction model to obtain a predicted probability that the fault state obtained is the actual fault state corresponding to the historical vehicle data.

[0067] In the implementation mode of the present application, the historical vehicle data corresponds to vehicle data values ​​corresponding to a variety of vehicle data types; vehicle data types include vehicle operation data (mileage, driving time, cumulative load, engine operating conditions, etc.), professional equipment parameters (temperature, pressure, vibration, damping, etc.) and environmental data (weather conditions, altitude, road conditions, etc.). The historical vehicle data is input into each fault prediction model to obtain the predicted probability that the fault state is the actual fault state corresponding to the historical vehicle data. Among them, the actual fault state corresponding to the historical vehicle data refers to the fault state when the vehicle data of the flight support vehicle is the historical vehicle data; the fault state includes a fault state and a no-fault state; the fault prediction model is trained based on the historical vehicle data and the corresponding actual fault state.

[0068] Here, the fault prediction model can adopt a depth-gated long short-term memory network (Depth-Gated LSTM), a Log-logistic survival function model, etc.

[0069] Furthermore, if the fault prediction model adopts a deep gated long short-term memory network, the fault prediction model includes multiple LSTM layers, each of which includes an input gate i l 、Forget gate l , output gate o l ; In addition, each LSTM layer is also used to calculate the cell state c l and the hidden state h l The hidden state h output by the last LSTM layer of the fault prediction model l That is, the initial prediction probability corresponding to each fault state.

[0070] Among them, (1) the input gate is as follows:

[0071]

[0072] in, is the output of the input gate of the lth layer at time step t, σ is the sigmoid activation function, is the weight matrix of the input gate of the lth layer, is the input of the lth layer at time step t, Hidden State to the weight matrix of the input gate, is the hidden state of the lth layer at time step t-1, Cell state to the weight matrix of the input gate, is the cell state of the lth layer at time step t-1, is the bias term of the input gate.

[0073] (2) Construct the forget gate as follows:

[0074]

[0075] in, is the output of the forget gate of the lth layer at time step t, is the weight matrix of the lth layer input to the forget gate, Hidden state To the weight matrix of the forget gate, is the bias term of the forget gate.

[0076] (3) Construct the cell state update function as follows:

[0077]

[0078] in, is the cell state of the lth layer at time step t, is the output of the forget gate of layer l at time step t, is the cell state of the lth layer at time step t-1, is the output of the input gate of layer l at time step t, is the weight matrix for updating the cell state input to the lth layer, Hidden state To the weight matrix of the cell state update, Bias term for cell state update.

[0079] (4) Construct the output gate as follows:

[0080]

[0081] in, is the output of the output gate of layer l at time step t, is the weight matrix from input to output gate, The weight matrix from hidden state to output gate, is the weight matrix from cell state to output gate, is the cell state of the lth layer at time step t, is the bias term of the output gate.

[0082] (5) Construct the hidden state update function as follows:

[0083]

[0084] in, is the hidden state of the lth layer at time step t, is the output of the output gate of layer l at time step t, is the cell state of the lth layer at time step t, and tanh is the hyperbolic tangent function.

[0085] In addition, when the fault prediction model uses a deep gated long short-term memory network, a mean square error loss function is defined to make the mean square error between the prediction result and the true label as small as possible. The details are as follows:

[0086]

[0087] in, represents the actual fault status of the i-th historical vehicle data (1 indicates a fault exists, 0 indicates no fault exists), It represents the predicted probability that the fault state predicted by the model of the i-th historical vehicle data is the actual fault state, and N is the number of historical vehicle data.

[0088] When the fault prediction model uses a deep gated long short-term memory network, the gradient of the model parameters is calculated according to the loss function, and the model parameters are updated using the stochastic gradient descent optimization algorithm. Finally, the model parameters are updated through the back propagation algorithm and optimizer to minimize the loss function. The specific formula is as follows:

[0089]

[0090] in, is the gradient of the model parameters, is the gradient of the loss function Loss to the i-th historical vehicle data, θ new is the updated model parameter, θ old is the model parameter before updating, lr is the learning rate for controlling the step length of model parameter updating, and the default value in the embodiment of the present application is 0.001. In addition, That is, the partial derivative of the loss function with respect to the model parameter θ, which reflects the contribution of each historical vehicle data to the overall loss of the model.

[0091] Furthermore, if the fault prediction model adopts the Log-logistic survival function model, the survival function of the model is established:

[0092]

[0093] Among them, S(t|X) is the survival function value of the historical vehicle data in the future time period t (that is, the probability of a fault-free state), X is the set of historical vehicle data, λ0 is the scale parameter that describes the fault occurrence rate, γ is the shape parameter that describes the distribution of the fault occurrence time, β is the regression coefficient vector of the model, and T is the time period corresponding to the set of historical vehicle data.

[0094] In addition, according to the preset time interval, the model parameters λ0, γ, and β are updated by the following formula:

[0095]

[0096] in, is the likelihood function value of the model, δ i is the actual fault status of the i-th historical vehicle data (1 indicates a fault exists, 0 indicates no fault exists), h(t|X i ) is the risk function value of the i-th historical vehicle data in the future time period t, S(t|X i ) is the survival function value of the i-th historical vehicle data in the future time period t.

[0097] Here, the maximum likelihood estimation is used for the Log-logistic model, combined with the gradient descent method for multiple iterations until the parameters that maximize the likelihood function value are found.

[0098] Furthermore, historical vehicle data of flight support vehicles is obtained, including: obtaining initial historical vehicle data corresponding to each vehicle data type within a preset time period of the flight support vehicle; for each vehicle data type, sliding a window of the initial historical vehicle data corresponding to the vehicle data type, and calculating the standard deviation value of the initial historical vehicle data in each window; and each window carries a sliding serial number; and the set of standard deviation values ​​corresponding to windows with the same serial number is determined as the historical vehicle data of the flight support vehicle.

[0099] Optionally, before sliding the window of the initial historical vehicle data corresponding to the vehicle data type and calculating the standard deviation value of the initial historical vehicle data in each window, the method also includes: filling missing values ​​for the initial historical vehicle data corresponding to each vehicle data type; and standardizing the initial historical vehicle data corresponding to each vehicle data type after the missing values ​​are filled, to obtain the final initial historical vehicle data.

[0100] S102: For each fault prediction model, calculate the log-likelihood function values ​​of all prediction probabilities obtained by the fault prediction model.

[0101] In an embodiment of the present application, the sum of the logarithms of all predicted probabilities obtained by the fault prediction model is determined as the log-likelihood function value of all predicted probabilities obtained by the fault prediction model.

[0102] For example, the predicted probabilities obtained by the fault prediction model include {0.3, 0.5, 0.5, 0.4}. Then the log-likelihood function value of all predicted probabilities obtained by the fault prediction model is log0.3+log0.5+log0.5+log0.4.

[0103] S103: Calculate the fault prediction weight value corresponding to each fault prediction model according to the log-likelihood function values ​​corresponding to all fault prediction models.

[0104] In the implementation manner of the present application, the log-likelihood function values ​​corresponding to all fault prediction models are substituted into the following formula to calculate the fault prediction weight value corresponding to each fault prediction model;

[0105]

[0106] Among them, P(M i |X) is the fault prediction weight value corresponding to the i-th fault prediction model, M i is the i-th fault prediction model, X is the set of historical vehicle data, P(X|M i ) is the log-likelihood function value corresponding to the i-th fault prediction model, P(M i ) is the prior fault prediction weight value corresponding to the i-th fault prediction model, Q is the number of fault prediction models, P(X|M j ) is the log-likelihood function value corresponding to the j-th fault prediction model, P(M j ) is the prior fault prediction weight value corresponding to the j-th fault prediction model.

[0107] Furthermore, the prior prediction weight values ​​corresponding to each fault prediction model are calculated through the following steps:

[0108] Step 1: Obtain a first fault prediction weight value corresponding to each fault prediction model, which is determined based on the actual prediction accuracy of the model, and a second fault prediction weight value obtained based on expert scoring.

[0109] In the implementation manner of the present application, the sum of the first fault prediction weight values ​​corresponding to all fault prediction models is 1, and the sum of the second fault prediction weight values ​​corresponding to all fault prediction models is 1. The higher the actual prediction accuracy of the model, the greater the first fault prediction weight value.

[0110] Step 2: Calculate the sum of all first fault prediction weight values ​​and all second fault prediction weight values ​​to obtain a first sum value.

[0111] For example, the first fault prediction weight value corresponding to the first fault prediction model is 0.7, and the second fault prediction weight value corresponding to the first fault prediction model is 0.3. The first fault prediction weight value corresponding to the second fault prediction model is 0.5, and the second fault prediction weight value corresponding to the second fault prediction model is 0.5. The sum of all first fault prediction weight values ​​and all second fault prediction weight values ​​is 0.7+0.3+0.5+0.5=2.

[0112] Step three: for each fault prediction model, calculate the sum of the first fault prediction weight value and the second fault prediction weight value corresponding to the fault prediction model to obtain a second sum value.

[0113] For example, the first fault prediction weight value corresponding to the first fault prediction model is 0.7, and the second fault prediction weight value corresponding to the first fault prediction model is 0.3. The first fault prediction weight value corresponding to the second fault prediction model is 0.5, and the second fault prediction weight value corresponding to the second fault prediction model is 0.5. The second sum value corresponding to the first fault prediction model is 0.7+0.5=1.2, and the second sum value corresponding to the second fault prediction model is 0.3+0.5=0.8.

[0114] Step 4: Determine the ratio of the second sum value to the first sum value as the prior prediction weight value corresponding to the fault prediction model.

[0115] For example, the sum of all first fault prediction weight values ​​and all second fault prediction weight values ​​is 2. The second sum value corresponding to the first fault prediction model is 0.7+0.5=1.2, and the second sum value corresponding to the second fault prediction model is 0.3+0.5=0.8. The prior prediction weight value corresponding to the first fault prediction model is 1.2 / 2=0.6, and the prior prediction weight value corresponding to the second fault prediction model is 0.8 / 2=0.4.

[0116] S104: Perform fault prediction on the flight support vehicle according to the fault prediction weight value corresponding to each fault prediction model.

[0117] Specifically, the following steps are used to predict the fault of the flight support vehicle according to the fault prediction weight value corresponding to each fault prediction model, including:

[0118] Step 1: Get the current vehicle data of the flight support vehicle.

[0119] In an implementation manner of the present application, the vehicle data type corresponding to the current vehicle data is the same as the vehicle data type corresponding to the historical vehicle data.

[0120] Step 2: Input the current vehicle data into each fault prediction model to obtain the initial prediction probability corresponding to each fault state.

[0121] In the implementation manner of the present application, the current vehicle data is input into each fault prediction model, and each fault prediction model can output the prediction probability corresponding to each fault state, that is, the initial prediction probability.

[0122] Step 3: According to each fault state, all initial prediction probabilities corresponding to the fault state are weighted summed according to the prediction weight values ​​corresponding to each fault prediction model to obtain the target prediction probability corresponding to the fault state.

[0123] For example, the prior prediction weight value corresponding to the first fault prediction model is 0.6, the prior prediction weight value corresponding to the second fault prediction model is 0.4, and the initial prediction probability corresponding to the fault-free state predicted by the first fault prediction model is 0.7, and the initial prediction probability corresponding to the fault-free state predicted by the second fault prediction model is. Then the target prediction probability corresponding to the fault-free state is 0.6×0.7+0.4×0.5=0.62.

[0124] Step 4: Determine the fault state with the highest target prediction probability as the predicted fault state of the flight support vehicle.

[0125] The embodiment of the present application provides a method for predicting faults of flight support vehicles, the method comprising: obtaining historical vehicle data of flight support vehicles and inputting the historical vehicle data into each fault prediction model to obtain a predicted probability that the fault state is the actual fault state corresponding to the historical vehicle data; the fault prediction model is trained based on the historical vehicle data and the corresponding actual fault state. For each fault prediction model, the log-likelihood function value of all predicted probabilities obtained by the fault prediction model is calculated; according to the log-likelihood function values ​​corresponding to all fault prediction models, the fault prediction weight value corresponding to each fault prediction model is calculated; and according to the fault prediction weight value corresponding to each fault prediction model, the flight support vehicle is predicted to have a fault. The present application can predict faults of flight support vehicles by calculating the fault prediction weight value corresponding to each fault prediction model, and then use the scarce maintenance resources to serve the vehicles that are really in urgent need of maintenance, effectively reducing the probability of failure in the actual operation of the vehicle.

[0126] Reference Figure 2 As shown, it is a flow chart of another method for predicting failures of flight support vehicles provided in an embodiment of the present application. The exemplary steps of the embodiment of the present application are described below:

[0127] S201. Acquire initial historical vehicle data corresponding to each vehicle data type within a preset time period of flight support vehicles.

[0128] S202. For each vehicle data type, a window is slid for the initial historical vehicle data corresponding to the vehicle data type, and a standard deviation value of the initial historical vehicle data in each window is calculated; and each window carries a sliding sequence number.

[0129] In the implementation of the present application, the window size is generally 15 natural days, and the sliding step is generally 1 day. The sequence number is the number of times the window is slid; for example, the sequence number of the first sliding window is 1.

[0130]

[0131] in, is the standard deviation value of the initial historical vehicle data in the i-th window corresponding to the k-th vehicle data type, w is the window size, x k,j is the value of the jth initial historical vehicle data corresponding to the kth vehicle data type, is the average value of the data value of the initial historical vehicle data in the i-th window corresponding to the k-th vehicle data type, X k is the set of initial historical vehicle data corresponding to the kth vehicle data type.

[0132] Optionally, missing values ​​are filled in for initial historical vehicle data corresponding to each vehicle data type.

[0133] In the implementation manner of the present application, for each vehicle data type, the missing value of the initial historical vehicle data corresponding to the vehicle data type is filled in by the following formula:

[0134]

[0135] Among them, x′ is the data value of the missing initial historical vehicle data (i.e., the data value of the filled initial historical vehicle data), x′1 is the data value of the previous known initial historical vehicle data of the missing initial historical vehicle data corresponding to the vehicle data type, x′2 is the data value of the next known initial historical vehicle data of the missing initial historical vehicle data corresponding to the vehicle data type, t1 is the time of the previous known initial historical vehicle data of the missing initial historical vehicle data corresponding to the vehicle data type, t1 is the time of the next known initial historical vehicle data of the missing initial historical vehicle data corresponding to the vehicle data type, and t is the time of the missing initial historical vehicle data corresponding to the vehicle data type.

[0136] Optionally, data standardization is performed on the initial historical vehicle data corresponding to each vehicle data type after the missing values ​​are filled to obtain the final initial historical vehicle data.

[0137] In the implementation manner of the present application, for each vehicle data type, the initial historical vehicle data corresponding to the vehicle data type after missing values ​​are filled is normalized by the following formula to obtain the final initial historical vehicle data:

[0138]

[0139] Among them, x zis the final initial historical vehicle data corresponding to the vehicle data type, x′ is the initial historical vehicle data corresponding to the vehicle data type after the missing values ​​are filled, min(x′) is the minimum value in the initial historical vehicle data corresponding to the vehicle data type after the missing values ​​are filled, and max(x′) is the maximum value in the initial historical vehicle data corresponding to the vehicle data type after the missing values ​​are filled.

[0140] S203: Determine a set of standard deviation values ​​corresponding to windows with the same serial number as historical vehicle data of flight support vehicles.

[0141] For example, if the standard deviation of the initial historical vehicle data in the first window corresponding to the first vehicle data type is 0.7, the standard deviation of the initial historical vehicle data in the first window corresponding to the second vehicle data type is 0.3, and the standard deviation of the initial historical vehicle data in the first window corresponding to the second vehicle data type is 0.5, then the historical vehicle data is {0.7, 0.3, 0.5}.

[0142] Here, through the implementation method of the present application, it is possible to ensure that the historical vehicle data can accurately reflect vehicle failures, and at the same time reduce the amount of historical vehicle data, thereby reducing the prediction workload.

[0143] Based on the same inventive concept, the embodiments of the present application also provide a flight support vehicle fault prediction device corresponding to the flight support vehicle fault prediction method. Since the principle of solving the problem by the device in the embodiments of the present application is similar to the above-mentioned flight support vehicle fault prediction method in the embodiments of the present application, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be repeated.

[0144] Reference Figure 3 FIG. 1 is a schematic diagram of a flight support vehicle fault prediction device provided in an embodiment of the present application, the device comprising:

[0145] The acquisition module 301 is used to acquire historical vehicle data of flight support vehicles and input the historical vehicle data into each fault prediction model to obtain the predicted probability that the fault state obtained is the actual fault state corresponding to the historical vehicle data; the fault prediction model is trained based on the historical vehicle data and the corresponding actual fault state;

[0146] A calculation module 302 is used to calculate the log-likelihood function values ​​of all prediction probabilities obtained by the fault prediction model for each fault prediction model;

[0147] The calculation module 302 is further used to calculate the fault prediction weight value corresponding to each fault prediction model according to the log-likelihood function values ​​corresponding to all fault prediction models;

[0148] The prediction module 303 is used to perform fault prediction on the flight support vehicle according to the fault prediction weight value corresponding to each fault prediction model.

[0149] In a possible implementation, the calculation module 302 is specifically configured to determine the sum of the logarithms of all predicted probabilities obtained by the fault prediction model as the log-likelihood function value of all predicted probabilities obtained by the fault prediction model.

[0150] In a possible implementation, the calculation module 302 is specifically used to substitute the log-likelihood function values ​​corresponding to all fault prediction models into the following formula to calculate the fault prediction weight value corresponding to each fault prediction model;

[0151]

[0152] Among them, P(M i |X) is the fault prediction weight value corresponding to the i-th fault prediction model, M i is the i-th fault prediction model, X is the set of historical vehicle data, P(X|M i ) is the log-likelihood function value corresponding to the i-th fault prediction model, P(M i ) is the prior fault prediction weight value corresponding to the i-th fault prediction model, Q is the number of fault prediction models, P(X|M j ) is the log-likelihood function value corresponding to the j-th fault prediction model, P(M j ) is the prior fault prediction weight value corresponding to the j-th fault prediction model.

[0153] In a possible implementation, the calculation module 302 is further configured to:

[0154] Obtaining a first fault prediction weight value determined based on the actual prediction accuracy of the model and a second fault prediction weight value obtained based on expert scoring corresponding to each fault prediction model;

[0155] Calculating the sum of all first fault prediction weight values ​​and all second fault prediction weight values ​​to obtain a first sum value;

[0156] For each fault prediction model, calculating the sum of a first fault prediction weight value and a second fault prediction weight value corresponding to the fault prediction model to obtain a second sum value;

[0157] The ratio of the second sum value to the first sum value is determined as the priori prediction weight value corresponding to the fault prediction model.

[0158] In one possible implementation, the prediction module 303 is specifically used to obtain current vehicle data of the flight support vehicle; input the current vehicle data into each fault prediction model to obtain the initial prediction probability corresponding to each fault state; according to each fault state, according to the prediction weight value corresponding to each fault prediction model, weighted sum all the initial prediction probabilities corresponding to the fault state to obtain the target prediction probability corresponding to the fault state; and determine the fault state with the largest target prediction probability as the predicted fault state of the flight support vehicle.

[0159] In one possible implementation, the acquisition module 301 is specifically used to obtain initial historical vehicle data corresponding to each vehicle data type within a preset time period of the flight support vehicle; for each vehicle data type, a window is slid for the initial historical vehicle data corresponding to the vehicle data type, and the standard deviation value of the initial historical vehicle data in each window is calculated; and each window carries a sliding serial number; and the set of standard deviation values ​​corresponding to the windows with the same serial number is determined as the historical vehicle data of the flight support vehicle.

[0160] In a possible implementation, before sliding the window of the initial historical vehicle data corresponding to the vehicle data type and calculating the standard deviation value of the initial historical vehicle data in each window, the acquisition module 301 is further used to:

[0161] Fill missing values ​​in the initial historical vehicle data corresponding to each vehicle data type;

[0162] The initial historical vehicle data corresponding to each vehicle data type after the missing values ​​are filled is standardized to obtain the final initial historical vehicle data.

[0163] The embodiment of the present application provides a flight support vehicle fault prediction device, which includes: an acquisition module 301, which is used to obtain historical vehicle data of the flight support vehicle and input the historical vehicle data into each fault prediction model to obtain the predicted probability of the actual fault state corresponding to the historical vehicle data; the fault prediction model is obtained by training based on the historical vehicle data and the corresponding actual fault state; a calculation module 302, which is used to calculate the logarithmic likelihood function value of all predicted probabilities obtained by the fault prediction model for each fault prediction model; the calculation module 302 is also used to calculate the fault prediction weight value corresponding to each fault prediction model according to the logarithmic likelihood function value corresponding to all fault prediction models; a prediction module 303, which is used to predict the fault of the flight support vehicle according to the fault prediction weight value corresponding to each fault prediction model. The present application can predict the fault of the flight support vehicle by calculating the fault prediction weight value corresponding to each fault prediction model, and then use the scarce maintenance resources to serve the vehicles that are really in urgent need of maintenance, effectively reducing the probability of failure in the actual operation of the vehicle.

[0164] like Figure 4 As shown, an electronic device 400 provided in an embodiment of the present application includes: a processor 401, a memory 402 and a bus, the memory 402 stores machine-readable instructions executable by the processor 401, when the electronic device is running, the processor 401 and the memory 402 communicate through the bus, and the processor 401 executes the machine-readable instructions to perform the steps of the above-mentioned flight support vehicle fault prediction method.

[0165] Specifically, the memory 402 and the processor 401 can be general-purpose memories and processors, which are not specifically limited here. When the processor 401 runs the computer program stored in the memory 402, the flight support vehicle fault prediction method can be executed.

[0166] Corresponding to the above-mentioned flight support vehicle failure prediction method, an embodiment of the present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned flight support vehicle failure prediction method are executed.

[0167] Those skilled in the art can clearly understand that, for the convenience and simplicity of description, the specific working process of the system and device described above can refer to the corresponding process in the method embodiment, and will not be repeated in this application. In the several embodiments provided in this application, it should be understood that the disclosed system, device and method can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the modules is only a logical function division. There may be other division methods in actual implementation. For example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0168] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0169] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0170] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium that is executable by a processor. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the information processing method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard drives, ROM, RAM, magnetic disks, or optical disks.

[0171] The above are only specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for predicting failure of flight support vehicles, characterized in that: The method comprises: Acquire historical vehicle data of flight support vehicles and input the historical vehicle data into each fault prediction model to obtain a predicted probability that the fault state is the actual fault state corresponding to the historical vehicle data; the fault prediction model is trained based on the historical vehicle data and the corresponding actual fault state; For each fault prediction model, calculating the log-likelihood function values ​​of all prediction probabilities obtained by the fault prediction model; According to the log-likelihood function values ​​corresponding to all fault prediction models, the fault prediction weight value corresponding to each fault prediction model is calculated; Fault prediction is performed on the flight support vehicle according to the fault prediction weight value corresponding to each fault prediction model.

2. The flight support vehicle failure prediction method according to claim 1, characterized in that: The calculating the log-likelihood function values ​​of all predicted probabilities obtained by the fault prediction model includes: The sum of the logarithms of all predicted probabilities obtained by the fault prediction model is determined as the log-likelihood function value of all predicted probabilities obtained by the fault prediction model.

3. The flight support vehicle failure prediction method according to claim 1, characterized in that: The method of calculating the fault prediction weight value corresponding to each fault prediction model according to the log-likelihood function values ​​corresponding to all fault prediction models includes: Substitute the log-likelihood function values ​​corresponding to all fault prediction models into the following formula to calculate the fault prediction weight value corresponding to each fault prediction model; Among them, P(M i |X) is the fault prediction weight value corresponding to the i-th fault prediction model, M i is the i-th fault prediction model, X is the set of historical vehicle data, P(X|M i ) is the log-likelihood function value corresponding to the i-th fault prediction model, P(M i ) is the prior fault prediction weight value corresponding to the i-th fault prediction model, Q is the number of fault prediction models, P(X|M j ) is the log-likelihood function value corresponding to the j-th fault prediction model, P(M j ) is the prior fault prediction weight value corresponding to the j-th fault prediction model.

4. The flight support vehicle failure prediction method according to claim 3, characterized in that: The prior prediction weight values ​​corresponding to each fault prediction model are calculated through the following steps: Obtaining a first fault prediction weight value determined based on the actual prediction accuracy of the model and a second fault prediction weight value obtained based on expert scoring corresponding to each fault prediction model; Calculating the sum of all first fault prediction weight values ​​and all second fault prediction weight values ​​to obtain a first sum value; For each fault prediction model, calculating the sum of a first fault prediction weight value and a second fault prediction weight value corresponding to the fault prediction model to obtain a second sum value; The ratio of the second sum value to the first sum value is determined as the priori prediction weight value corresponding to the fault prediction model.

5. The flight support vehicle failure prediction method according to claim 1, characterized in that: The performing fault prediction on the flight support vehicle according to the fault prediction weight value corresponding to each fault prediction model includes: Obtaining current vehicle data of the flight support vehicle; Inputting the current vehicle data into each fault prediction model to obtain an initial prediction probability corresponding to each fault state; According to each fault state, according to the prediction weight value corresponding to each fault prediction model, all initial prediction probabilities corresponding to the fault state are weighted summed to obtain the target prediction probability corresponding to the fault state; The fault state with the highest target predicted probability is determined as the predicted fault state of the flight support vehicle.

6. The flight support vehicle failure prediction method according to claim 5, characterized in that: Obtain the historical vehicle data of the flight support vehicle, including: Acquire initial historical vehicle data corresponding to each vehicle data type within a preset time period of the flight support vehicle; For each vehicle data type, a window is slid for the initial historical vehicle data corresponding to the vehicle data type, and the standard deviation value of the initial historical vehicle data in each window is calculated; and each window carries a sliding sequence number; A set of standard deviation values ​​corresponding to windows with the same serial number is determined as the historical vehicle data of the flight support vehicle.

7. The flight support vehicle failure prediction method according to claim 6, characterized in that: Before sliding the window of the initial historical vehicle data corresponding to the vehicle data type and calculating the standard deviation value of the initial historical vehicle data in each window, the method further includes: Fill missing values ​​in the initial historical vehicle data corresponding to each vehicle data type; The initial historical vehicle data corresponding to each vehicle data type after the missing values ​​are filled is standardized to obtain the final initial historical vehicle data.

8. A flight support vehicle failure prediction device, characterized in that: The device comprises: An acquisition module, used for acquiring historical vehicle data of flight support vehicles and inputting the historical vehicle data into each fault prediction model to obtain a predicted probability that the fault state obtained is the actual fault state corresponding to the historical vehicle data; the fault prediction model is trained based on the historical vehicle data and the corresponding actual fault state; A calculation module, used for calculating, for each fault prediction model, the log-likelihood function values ​​of all prediction probabilities obtained by the fault prediction model; The calculation module is further used to calculate the fault prediction weight value corresponding to each fault prediction model according to the log-likelihood function values ​​corresponding to all fault prediction models; The prediction module is used to predict the faults of the flight support vehicles according to the fault prediction weight values ​​corresponding to the fault prediction models.

9. An electronic device, characterized in that: include: A processor, a storage medium and a bus, wherein the storage medium stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor and the storage medium communicate via the bus, and the processor executes the machine-readable instructions to perform the steps of the flight support vehicle fault prediction method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the flight support vehicle failure prediction method according to any one of claims 1 to 7 are executed.

Images