CN115660231A

CN115660231A - Maintenance error prediction method and device and electronic equipment

Info

Publication number: CN115660231A
Application number: CN202211670893.1A
Authority: CN
Inventors: 陈奇; 王占海; 付鹏
Original assignee: China Academy of Civil Aviation Science and Technology
Current assignee: China Academy of Civil Aviation Science and Technology
Priority date: 2022-12-26
Filing date: 2022-12-26
Publication date: 2023-01-31
Anticipated expiration: 2042-12-26
Also published as: CN115660231B

Abstract

The application provides a repair error prediction method, a repair error prediction device and electronic equipment, wherein the method comprises the following steps: acquiring aviation operation data of a target object to be predicted, wherein the target object is any aviation device; and inputting the aviation operation data into a target prediction model to obtain a maintenance error prediction result corresponding to the target object, wherein the target prediction model is used for predicting the maintenance error of the target object. By the method and the device, the problem that the error type and reason possibly occurring in the aviation maintenance process cannot be accurately predicted in the related technology is solved.

Description

Maintenance error prediction method and device and electronic equipment

Technical Field

The invention relates to the technical field of aircraft maintenance, in particular to a method and a device for predicting maintenance errors and electronic equipment.

Background

In the aviation field, a maintenance error for an aircraft has become an important factor affecting the safe operation of the aircraft, and the maintenance error refers to a wrong behavior caused by the influence of various external and internal factors on an aviation maintenance worker, so that the aviation maintenance operation activity is deviated and wrong, the expected purpose cannot be achieved, and adverse consequences such as abnormal aviation equipment state, equipment damage or casualties are caused. Therefore, the force must be increased to prevent the airplane from being repaired mistakenly and ensure the safe flight of the airplane.

With the rapid development of civil aviation industry in China, the number of fleets is continuously increased, the workload of aviation maintenance personnel is continuously increased, and the safety risk of the civil aviation industry caused by aviation maintenance errors is increasingly obvious. In addition, in the process of maintaining the airplane, aviation maintenance personnel are easily affected by human or machinery, and the airplane is subjected to more serious faults in a wrong maintenance mode. Among the factors that require particular attention in the maintenance of an aircraft are: complex structure of the aircraft, the skill level of the airline maintenance personnel, and whether or not maintenance regulations are violated. By considering human factors, the aviation maintenance errors are researched and predicted, the flight accident rate can be further reduced, and the flight safety level can be improved. However, the prior art does not accurately predict the type and cause of errors that may occur during an aircraft repair.

Therefore, there is a need to develop a repair error prediction method to predict the type and cause of a repair error in advance and avoid the occurrence of a repair error during the aircraft repair process.

Disclosure of Invention

The application provides a maintenance error prediction method and device and electronic equipment, which are used for at least solving the problem that the type and reason of errors possibly occurring in the aviation maintenance process cannot be accurately predicted in the related technology.

According to an aspect of an embodiment of the present application, there is provided a repair error prediction method, including:

acquiring aviation operation data of a target object to be predicted, wherein the target object is any aviation device;

and inputting the aviation operation data into a target prediction model to obtain a maintenance error prediction result corresponding to the target object, wherein the target prediction model is used for predicting the maintenance error of the target object.

According to another aspect of the embodiments of the present application, there is also provided a repair error prediction apparatus including:

the system comprises an acquisition module, a prediction module and a prediction module, wherein the acquisition module is used for acquiring aviation operation data of a target object to be predicted, and the target object is any aviation equipment;

and the obtaining module is used for inputting the aviation operation data into a target prediction model to obtain a maintenance error prediction result corresponding to the target object, wherein the target prediction model is used for predicting the maintenance error of the target object.

Optionally, the obtaining module includes:

the system comprises an acquisition unit, a maintenance unit and a maintenance unit, wherein the acquisition unit is used for acquiring a first preset number of groups of historical aviation operation data and corresponding historical maintenance error types and reasons;

the preprocessing unit is used for preprocessing the historical aviation operation data to obtain preprocessed historical aviation operation data;

the obtaining unit is used for respectively taking each group of the preprocessed historical aviation operation data and the corresponding historical maintenance error type and reason as a historical data sample to obtain a first preset number of historical data samples;

the first generation unit is used for acquiring a second preset number of historical data samples from a first preset number of historical data samples to generate a training sample set;

a second generating unit, configured to generate an initial prediction model according to the training sample set;

and the optimization unit is used for optimizing the initial prediction model through a target optimization algorithm based on the training sample set to obtain the target prediction model.

Optionally, the optimization unit comprises:

the first generation submodule is used for acquiring a third preset number of historical data samples from the first preset number of historical data samples to generate a test sample set;

the first obtaining submodule is used for inputting the historical aviation operation data in the test sample set into the target prediction model to obtain a corresponding historical maintenance error prediction result;

the comparison submodule is used for comparing the historical maintenance error prediction result with the historical maintenance error type and reason corresponding to the test sample set to obtain a comparison result;

the second obtaining submodule is used for obtaining the prediction error of the target prediction model according to the comparison result;

and the judgment submodule is used for judging whether the prediction error is within a preset range, and if so, the target prediction model can predict the target object maintenance error.

A first initialization submodule, configured to initialize a model parameter in the initial prediction model;

the coding submodule is used for coding the model parameters to obtain coding parameters;

the second initialization submodule is used for initializing a population of the target optimization algorithm and corresponding population parameters, wherein the population comprises a fifth preset number of the coding parameters as individuals, and the population parameters comprise the evolution times of the population, threshold values of the evolution times, cross probabilities and variation probabilities;

the acquisition submodule is used for acquiring a fitness function of the target optimization algorithm;

the decoding submodule is used for decoding the coding parameters to obtain a fifth preset number of decoding parameters;

the third obtaining submodule is used for obtaining the fitness of the individual according to the fitness function, the decoding parameters, the training sample set, the testing sample set and the initial prediction model;

an updating submodule, configured to update the population and the evolution frequency according to the fitness, the cross probability, the mutation probability and a preset method, determine whether the evolution frequency reaches the threshold, if not, determine whether an individual with a fitness smaller than a preset threshold exists in the population, if so, take the individual with the fitness smaller than the preset threshold as a target individual and stop updating the population and the evolution frequency, and if so, take the individual with the minimum fitness in the current population as the target individual, where a coding parameter corresponding to the target individual is an optimal solution;

the fourth obtaining submodule is used for decoding the coding parameters corresponding to the target individual to obtain target model parameters;

and the fifth obtaining submodule is used for modifying the model parameters of the initial prediction model into the target model parameters to obtain the target prediction model.

Optionally, the third obtaining sub-module includes:

the first obtaining subunit is configured to modify the model parameters of the initial prediction model into the decoding parameters to obtain a fifth preset number of intermediate prediction models, where the intermediate prediction models are used to obtain the fitness;

the training subunit is used for sequentially training the intermediate prediction model through the training sample set to obtain a training difference value;

the testing subunit is used for sequentially testing the intermediate prediction model through the testing sample set to obtain a testing difference value;

and the second obtaining subunit is used for respectively obtaining the fitness of each individual according to the fitness function, the training difference and the corresponding test difference.

Optionally, the pre-processing unit comprises:

the removing submodule is used for removing noise points and abnormal data in the historical aviation operation data to obtain clean data;

and the processing submodule is used for carrying out normalization processing on the cleaning data through a preset formula to obtain the preprocessed historical aviation operation data.

Optionally, the second generating unit includes:

a determining submodule, configured to determine, based on the training sample set, an input layer neuron number and an output layer neuron number of a neural network included in the initial prediction model;

a sixth obtaining submodule, configured to obtain an initial hidden layer neuron number according to the input layer neuron number, the output layer neuron number, and a target formula;

a seventh obtaining submodule, configured to adjust the initial hidden layer neuron number and the initial transfer function pair according to a preset index, and obtain a target hidden layer neuron number and a target transfer function pair, respectively;

and the second generation submodule is used for generating the initial prediction model according to the input layer neuron number, the output layer neuron number, the target hidden layer neuron number and the target transfer function pair.

Optionally, the seventh obtaining sub-module comprises:

an obtaining subunit, configured to obtain a fourth preset number of combinations, where each of the combinations includes a hidden layer neuron number and a transfer function pair;

a third obtaining subunit, configured to obtain, according to a preset formula, the preset index of the initial prediction model corresponding to each combination;

and the sub-unit is configured to use the hidden layer neuron number and the transfer function pair corresponding to the initial prediction model with the minimum preset index as the target hidden layer neuron number and the target transfer function pair, respectively.

According to another aspect of the embodiments of the present application, there is also provided an electronic device, including a processor, a communication interface, a memory, and a communication bus, where the processor, the communication interface, and the memory communicate with each other through the communication bus; wherein the memory is used for storing the computer program; a processor for performing the method steps in any of the above embodiments by running the computer program stored on the memory.

According to a further aspect of an embodiment of the present application, there is also provided a computer-readable storage medium, in which a computer program is stored, wherein the computer program is configured to perform the method steps in any of the above embodiments when the computer program is executed.

In the embodiment of the application, the aviation operation data of a target object to be predicted is obtained, wherein the target object is any aviation device; and inputting the aviation operation data into a target prediction model to obtain a maintenance error prediction result corresponding to the target object, wherein the target prediction model is used for predicting the maintenance error of the target object. According to the method and the device, the obtained aviation operation data are input into the target prediction model, the possible maintenance error types and reasons are predicted, and the target prediction model is obtained by continuously adjusting and optimizing the model parameters of the initial prediction model through the genetic algorithm, so that the prediction accuracy of the target prediction model on various aviation operation data is improved. The problem that the error type and reason which may appear in the aviation maintenance process cannot be accurately predicted in the related technology is solved.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and together with the description, serve to explain the principles of the invention.

In order to more clearly illustrate the embodiments or technical solutions in the prior art of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious for those skilled in the art to obtain other drawings without inventive labor.

FIG. 1 is a schematic flow diagram of an alternative repair error prediction method according to an embodiment of the application;

FIG. 2 is a graph of mean square error after training of different hidden layers and activation functions of an alternative BP neural network according to an embodiment of the present application;

FIG. 3 is a graph A comparing the predicted results of an alternative optimized BP neural network based on genetic algorithm with a conventional BP neural network according to an embodiment of the present application;

FIG. 4 is a graph B comparing the predicted results of an alternative optimized BP neural network based on genetic algorithm with a conventional BP neural network according to an embodiment of the present application;

FIG. 5 is a flow chart of an alternative method for optimizing a BP neural network based on a genetic algorithm for repair error prediction according to an embodiment of the present application;

FIG. 6 is a block diagram of an alternative repair error prediction device according to an embodiment of the present application;

fig. 7 is a block diagram of an alternative electronic device according to an embodiment of the present application.

Detailed Description

In order to make the technical solutions better understood by those skilled in the art, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only partial embodiments of the present application, but not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.

It should be noted that the terms "first," "second," and the like in the description and claims of this application and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the application described herein are capable of operation in sequences other than those illustrated or described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.

Along with the rapid development of civil aviation industry in China, the size of a fleet is continuously enlarged, the workload of maintenance personnel is continuously increased, and the safety risk of the civil aviation industry caused by the personal errors of aviation maintenance is increasingly obvious. Therefore, the maintenance error and the reason which are possibly caused can be predicted in advance, and the maintenance error phenomenon can be effectively avoided.

Based on the above, according to an aspect of the embodiments of the present application, there is provided a repair error prediction method, as shown in fig. 1, a flow of the method may include the following steps:

step S101, acquiring aviation operation data of a target object to be predicted, wherein the target object is any aviation device.

Optionally, obtaining the aviation operation data of the aircraft needing to be repaired, namely the target object to be predicted, includes: passenger turnover, transportation turnover, cargo turnover, flight hours, airplane take-off number, sign number and the like.

And S102, inputting the aviation operation data into a target prediction model to obtain a maintenance error prediction result corresponding to the target object, wherein the target prediction model is used for predicting the maintenance error of the target object.

Optionally, the aviation operation data is used as an input of the target prediction model, and a maintenance error prediction result is output, wherein the maintenance error prediction result comprises a maintenance error type (damaged airplanes, injured personnel, improper installation and the like) and a maintenance error reason (violation of regulations, personal factors, lack of skill knowledge and the like).

In the embodiment of the application, the aviation operation data of a target object to be predicted is obtained, wherein the target object is any aviation device; and inputting the aviation operation data into a target prediction model to obtain a maintenance error prediction result corresponding to the target object, wherein the target prediction model is used for predicting the maintenance error of the target object. According to the method and the device, the obtained aviation operation data are input into the target prediction model, the types and reasons of possible maintenance errors are predicted, the target prediction model is obtained by continuously adjusting and optimizing model parameters of the initial prediction model through a genetic algorithm, and the prediction accuracy of the target prediction model on various aviation operation data is improved. The problem that the error type and reason which may appear in the aviation maintenance process cannot be accurately predicted in the related technology is solved.

As an alternative embodiment, before inputting the aviation operation data into the target prediction model, the method further comprises:

acquiring a first preset quantity of groups of historical aviation operation data and corresponding historical maintenance error types and reasons;

preprocessing historical aviation operation data to obtain preprocessed historical aviation operation data;

respectively taking each group of preprocessed historical aviation operation data and corresponding historical maintenance error types and reasons as a historical data sample to obtain a first preset number of historical data samples;

acquiring a second preset number of historical data samples from the first preset number of historical data samples to generate a training sample set;

generating an initial prediction model according to the training sample set;

and optimizing the initial prediction model through a target optimization algorithm based on the training sample set to obtain a target prediction model.

Alternatively, historical aviation operation data, maintenance error types and reasons can be obtained through an aviation official website and a ground aircraft maintenance station, and then the Excel table is used for summarizing the historical aviation operation data, the maintenance error types and the reasons. And then, loading the historical aviation operation data into Matlab software to preprocess the historical aviation operation data to obtain the preprocessed historical aviation operation data.

And classifying the historical aviation operation data, the maintenance error types and the reasons according to the years, and taking the historical aviation operation data of the same year and the corresponding historical maintenance error types and reasons as a historical data sample to obtain N historical data samples, namely a first preset number of historical data samples, wherein N can be 10 or other positive integers, and is not limited to the specific number of N.

Selecting M historical data samples, namely a second preset number of historical data samples, from the N historical data samples as training samples, and generating a training set, namely a training sample set, wherein M can be 8 or other positive integers less than or equal to N, and the specific number of M is not limited.

And determining the structure of the BP neural network according to training samples in a training set, determining the neuron number of a hidden layer of the BP neural network and a transfer function among an input layer, the hidden layer and an output layer, wherein the initial prediction model comprises the BP neural network.

And (3) taking historical aviation operation data after the training set is preprocessed: passenger turnover, transportation turnover, cargo turnover, flight hours, airplane takeoff number, sign times and the like are used as input of an initial prediction model, the type of maintenance errors (damaged airplanes, personnel injuries, improper installation and the like) and the reasons of the maintenance errors (violating regulations, personal factors, lack of technical knowledge and the like) are used as output of the model, and a BP neural network prediction model in the initial prediction model is trained through a Genetic Algorithm (GA) to finally obtain a target prediction model.

In the embodiment of the application, the historical aviation operation data is preprocessed by acquiring the historical aviation operation data, the maintenance error type and the reason, and N samples are generated. And selecting M samples from the N samples to generate a training set. The method has the advantages that the initial prediction model is built through the training set, the target prediction model is obtained through training the initial prediction model through the training set, the possible maintenance error types and reasons corresponding to different aviation operation data can be accurately and efficiently predicted through the target prediction model obtained through the method, and the problem that the error types and reasons possibly occurring in the aviation maintenance process cannot be accurately predicted in the related technology is solved.

As an alternative embodiment, after the initial prediction model is optimized by the target optimization algorithm based on the training sample set to obtain the target prediction model, the method further includes:

acquiring a third preset number of historical data samples from the first preset number of historical data samples to generate a test sample set;

inputting historical aviation operation data in a test sample set into a target prediction model to obtain a corresponding historical maintenance error prediction result;

comparing the historical maintenance error prediction result with the historical maintenance error type and reason corresponding to the test sample set to obtain a comparison result;

according to the comparison result, obtaining a prediction error of the target prediction model;

and judging whether the prediction error is within a preset range, and if so, predicting the maintenance error of the target object by using the target prediction model.

Optionally, Q, that is, a third preset number of historical data samples are selected from the N historical data samples as test samples, and a test set, that is, a test sample set, is generated, where Q may be 2 or another positive integer less than or equal to N, and no specific number of Q is defined here.

And (3) testing the historical aviation operation data after centralized preprocessing: the passenger turnover amount, the transportation turnover amount, the cargo turnover amount, the flight hours, the aircraft takeoff number, the sign times and the like are used as input of a target prediction model, and a historical maintenance error prediction result including the type and the reason of the prediction error is obtained through the target prediction model.

And comparing the predicted error type and reason with the real error type and reason corresponding to the test set to obtain a comparison result.

And according to the comparison result, calculating the root mean square error to obtain the prediction error of the target prediction model.

If the prediction error is within an acceptable range, i.e., a preset range, the target prediction model may be used to predict future airline maintenance errors.

Optionally, the test centralized historical aviation operation data can be respectively input into the prediction models based on the genetic algorithm optimized BP neural network and the traditional BP neural network to respectively obtain prediction results, and the prediction results can reflect that the prediction accuracy of the genetic algorithm optimized BP neural network is higher. When Q is equal to 2, fig. 3 is a comparison graph a of the prediction results of an optional genetic algorithm-based optimized BP neural network and a conventional BP neural network according to an embodiment of the present application, and fig. 4 is a comparison graph B of the prediction results of the optional genetic algorithm-based optimized BP neural network and the conventional BP neural network according to an embodiment of the present application. The above comparative graph a and comparative graph B correspond to the 1 st and 2 nd samples in the test set, respectively. In the comparison graph A and the comparison graph B, the maintenance error types and reasons are numbered, and the BP predicted value, the GA-BP predicted value and the frequency of occurrence of each maintenance error type and reason actually occurring in the test set are respectively counted. By comparing the graph A with the graph B, the prediction effect of the target prediction model based on the BP neural network (GA-BP) optimized by the genetic algorithm is better than that of the target prediction model based on the traditional BP neural network.

In the embodiment of the application, the target prediction model is tested through the test sample set, and if the prediction error obtained through the test is within an acceptable range, the target prediction model can be used. Through testing and selecting a target prediction model with acceptable test errors, the accuracy of the types and reasons of the predicted errors can be effectively improved. And the technical effect of the target prediction model is excellent by comparing the BP neural network optimized based on the genetic algorithm with the traditional BP neural network. The problem that the error type and reason which may appear in the aviation maintenance process cannot be accurately predicted in the related technology is solved.

As an optional embodiment, the preprocessing the historical aviation operation data to obtain the preprocessed historical aviation operation data includes:

removing noise and abnormal data in historical aviation operation data to obtain clean data;

and carrying out normalization processing on the cleaning data through a preset formula to obtain the preprocessed historical aviation operation data.

Optionally, loading historical aviation operation data into Matlab software, preprocessing the obtained historical aviation operation data, and firstly removing noise and abnormal data in the historical aviation operation data; then, normalizing the data without the noise and the abnormal data by a formula (1), wherein the normalization adopts a maximum and minimum normalization method, and the formula (1) of the data normalization is as follows:

（1）

wherein, the first and the second end of the pipe are connected with each other,

the experimental data after the normalization are shown,

the experimental data before the normalization is shown,

the maximum value of the experimental data is shown,

represents the minimum of experimental data.

In the embodiment of the application, by removing noise and abnormal data in historical aviation operation data and carrying out normalization processing on the data, the accuracy of initial prediction model training is improved, and the normalization processing data enables the prediction process to be more convenient and faster.

As an alternative embodiment, generating the initial prediction model from the training sample set comprises:

determining the number of input layer neurons and the number of output layer neurons of the neural network contained in the initial prediction model based on the training sample set;

obtaining the number of neurons of an initial hidden layer according to the number of neurons of an input layer, the number of neurons of an output layer and a target formula;

adjusting the number of neurons of the initial hidden layer and the initial transfer function pair according to a preset index to respectively obtain the number of neurons of the target hidden layer and the target transfer function pair;

and generating an initial prediction model according to the input layer neuron number, the output layer neuron number, the target hidden layer neuron number and the target transfer function pair.

Optionally, the number of neurons m of the input layer is equal to the number of input variables in the data to be processed, and the number of neurons n of the output layer is equal to the number of outputs associated with each input.

And obtaining the initial hidden layer neuron number k according to the input layer neuron number m, the output layer neuron number n and a target formula, namely formula (2).

（2）

Where a is a constant between 1 and 10, and when the value on the right side of equation (2) is not an integer, k is the closest integer to the value on the right side.

And adjusting the number k of neurons in the initial hidden layer and an initial transfer function pair according to a preset index, namely the Mean Square Error (MSE), wherein the initial transfer function pair comprises transfer functions between an input layer and the hidden layer and between the hidden layer and an output layer. And taking the hidden layer neuron number and transfer function pair with the minimum corresponding MSE value as a target hidden layer neuron number and target transfer function pair.

After the cyclic measurement of the BP neural network model, the obtained BP neural network has the structure that the number of neurons in an input layer is 12, the number of neurons in a target hidden layer is 7, the number of neurons in an output layer is 15, and a target transfer function pair is an activation function among the input layer, the hidden layer and the output layer and can be logsig and purelin.

And finally, generating an initial BP neural network according to the number of the neurons of the input layer, the number of the neurons of the output layer, the number of the neurons of the target hidden layer and the target transfer function pair, and further generating an initial prediction model.

In the embodiment of the application, the number of neurons in an input layer and the number of neurons in an output layer are obtained through a training sample set, so that the number of neurons in a target hidden layer and a target transfer function pair are obtained, and an initial prediction model is established according to the number of neurons in the input layer, the number of neurons in the output layer, the number of neurons in the target hidden layer and the target transfer function pair, so that the technical effect of predicting maintenance errors can be successfully achieved by the initial prediction model and a subsequent target prediction model.

As an optional embodiment, the adjusting the initial hidden layer neuron number and the initial transfer function pair according to the preset index to obtain the target hidden layer neuron number and the target transfer function pair respectively includes:

acquiring a fourth preset number of combinations, wherein each combination comprises a hidden layer neuron number and a transfer function pair;

respectively obtaining preset indexes of the initial prediction model corresponding to each combination according to a preset formula;

and respectively taking the hidden layer neuron number and the transfer function pair corresponding to the initial prediction model with the minimum preset index as a target hidden layer neuron number and a target transfer function pair.

Optionally, fig. 2 is a graph of mean square error after training of different hidden layers and activation functions of an optional BP neural network according to an embodiment of the present application, as shown in fig. 2, where the graph includes a plurality of combinations, that is, a fourth preset number, where the fourth preset number indicates a plurality, and no specific number is limited here. The figure has four transfer function pairs of tansig-purelin, logsig-purelin, tansig-logsig and logsig-tansig, and the value of the number of hidden layer neurons, namely the number of hidden layer neurons, is from 6 to 15. Different hidden layer neuron numbers and different transfer function pairs form a plurality of combinations.

And respectively obtaining the average Mean Square Error (MSE), namely a preset index, of the initial prediction model corresponding to each combination according to a preset formula, namely formula (3).

（3）

Wherein the content of the first and second substances,

is the number of the samples and is the number of the samples,

for the prediction of the number of types of errors,

is the actual value of the number of error types.

The combination of MSE minimum in fig. 2, i.e., MSE =0.040747, contains a hidden layer neuron number of 7, and the transfer function pair logsig-purelin is taken as the target hidden layer neuron number and target transfer function pair, respectively.

In the embodiment of the application, the number of the neurons of the plurality of hidden layers and the combination of the transfer function pairs are obtained by circularly measuring the BP neural network model, the average mean square error of each combination is referred to, the number of the neurons of the target hidden layer and the target transfer function pair are more accurately selected, and meanwhile, the established initial prediction model and the subsequent target prediction model can successfully realize the technical effect of the maintenance error prediction.

As an optional embodiment, optimizing the initial prediction model by using a target optimization algorithm based on the training sample set to obtain the target prediction model includes:

initializing model parameters in an initial prediction model;

coding the model parameters to obtain coding parameters;

initializing a population of a target optimization algorithm and corresponding population parameters, wherein the population comprises a fifth preset number of coding parameters as individuals, and the population parameters comprise the evolution times of the population, threshold values of the evolution times, cross probabilities and variation probabilities;

acquiring a fitness function of a target optimization algorithm;

decoding the coding parameters to obtain a fifth preset number of decoding parameters;

obtaining the fitness of the individual according to the fitness function, the decoding parameters, the training sample set, the testing sample set and the initial prediction model;

updating the population and the evolution times according to the fitness, the cross probability, the variation probability and a preset method, judging whether the evolution times reach a threshold value, if not, judging whether individuals with the fitness smaller than the preset threshold value exist in the population, if so, taking the individuals with the fitness smaller than the preset threshold value as target individuals and stopping updating the population and the evolution times, and if so, taking the individuals with the minimum fitness in the current population as the target individuals, wherein the coding parameters corresponding to the target individuals are optimal solutions;

decoding the coding parameters corresponding to the target individual to obtain target model parameters;

and modifying the model parameters of the initial prediction model into target model parameters to obtain a target prediction model.

Optionally, initializing model parameters corresponding to the BP neural network included in the initial prediction model, including a weight and a threshold length, and performing encoding processing on the model parameters to obtain encoding parameters. Setting population parameters of a target optimization algorithm, namely a genetic algorithm, setting a threshold value of the evolution times to be 100 times, taking the population scale to be 50, the cross probability to be 0.9 and the variation probability to be 0.05, generating 50 coding parameters as individuals to form a new population, setting the fifth preset number to be equal to the population scale number of 50, and setting the evolution times of the population to be 0 at the moment.

Obtaining fitness function of target optimization algorithm, namely formula (4)

（4）

Wherein the content of the first and second substances,

in order to be a degree of fitness,

and

the number of training sets, test sets,

，

the values of the training set and the test set of the neural network prediction respectively,

，

the true values of the training set and the test set, respectively.

Decoding the coding parameters of each individual to obtain 50 decoding parameters; and then obtaining the fitness of each individual according to a formula (4), the decoding parameters, the training sample set, the testing sample set and the initial prediction model.

In this embodiment, there are two conditions for stopping optimization, one is that the number of evolutionary times of the population reaches a threshold value, that is, 100, and the other is that an individual with a fitness smaller than a preset threshold value appears in the population, and the preset threshold value may be set according to a use requirement. Judging whether the evolution frequency of the population reaches 100, if not, further judging whether an individual with fitness smaller than a preset threshold exists in the current population, if so, stopping the optimization process, selecting the individual as a target individual, and if not, generating a new population through selection, intersection and variation operations in a genetic algorithm, wherein the evolution frequency of the population is added by 1. And if the evolution times reach 100, stopping the optimization process, and selecting the individual with the minimum fitness in the current population as the target individual.

And decoding the coding parameters of the target individual to obtain target model parameters, and then modifying the model parameters of the initial prediction model into the target model parameters to obtain the target prediction model.

In the embodiment of the application, the optimal weight and the threshold value in the BP neural network are determined through a genetic algorithm and assigned to the BP neural network, so that the prediction precision of the BP neural network is further improved, and the problem that the error type and the reason which possibly occur in the aviation maintenance process cannot be accurately predicted in the related technology is solved.

As an optional embodiment, obtaining the fitness of the individual according to the fitness function, the decoding parameters, the training sample set, the testing sample set, and the initial prediction model includes:

modifying the model parameters of the initial prediction model into decoding parameters to obtain a fifth preset number of intermediate prediction models, wherein the intermediate prediction models are used for obtaining the fitness;

sequentially training the intermediate prediction models through a training sample set to obtain a training difference value;

sequentially testing the intermediate prediction models through the test sample set to obtain a test difference value;

and respectively obtaining the fitness of each individual according to the fitness function, the training difference and the corresponding test difference.

Optionally, the model parameters of the initial prediction model are modified into decoding parameters obtained by decoding the encoding parameters of each individual, so as to obtain 50 intermediate prediction models.

The fitness can be found according to a fitness function, namely formula (4)

The MSE (mean square error) of a training set and a corresponding real value predicted by a BP (Back propagation) neural network contained in an intermediate prediction model is

MSE plus test set and corresponding real value, i.e.

And (4) obtaining the product. Therefore, the intermediate prediction models are sequentially trained through the training sample set to obtain the MSE (mean square error) of the predicted training set and the corresponding real value, namely the training difference, the intermediate prediction models are sequentially tested through the testing sample set to obtain the MSE of the testing set and the corresponding real value, namely the testing difference, the MSE and the testing difference are added to obtain the fitness of the corresponding individual, and the fitness of the 50 individuals is sequentially solved by adopting the same method.

In the embodiment of the application, the fitness of the individual is calculated by calculating the MSE of the predicted training set and the corresponding real value and the MSE of the test set and the corresponding real value, so that the obtained fitness is more accurate, and the optimization effect of the genetic algorithm is better.

As an alternative embodiment, fig. 5 is a flowchart of an alternative repair error prediction method for optimizing a BP neural network based on a genetic algorithm according to an embodiment of the present application, the method including the following steps:

inputting data; preprocessing data; determining a BP network topological structure; initializing a neural network weight and a threshold length; coding the weight and the threshold value of the neural network to obtain an initialized population; determining a fitness function; decoding to obtain a weight value and a threshold value; assigning the weight value and the threshold value to a BP neural network; training a BP neural network through a training sample, and testing the BP neural network through a testing sample; calculating the fitness; judging whether a termination condition is met, and if the termination condition is not met, performing selection operation; performing cross operation; performing mutation operation; generating a new generation of population, and then starting to execute subsequent operations from the weight value and the threshold value obtained by decoding; if the termination condition is met, decoding to obtain an optimal weight and a threshold; a GA-BP neural network; and (6) performing simulation prediction to obtain a result.

In the embodiment of the application, by combining the respective advantages of a genetic algorithm and a BP neural network, the advantages of the two methods are complemented, a complicated nonlinear mapping relation between aviation operation data such as passenger turnover amount, cargo turnover amount and the like and maintenance error types and reasons is established through the BP neural network, the maintenance error prediction model based on the genetic algorithm optimization BP neural network is established, the optimal weight and threshold in the BP neural network are determined through the genetic algorithm and assigned to the BP neural network, the prediction precision of the BP neural network is further improved, the maintenance error types and reasons are predicted in advance, and the prediction error is reduced. The problem that the error type and reason which may appear in the aviation maintenance process cannot be accurately predicted in the related technology is solved.

According to another aspect of the embodiment of the application, a repair error prediction device for implementing the repair error prediction method is also provided. Fig. 6 is a block diagram of an alternative repair error prediction apparatus according to an embodiment of the present application, which may include, as shown in fig. 6:

the acquiring module 601 is configured to acquire aviation operation data of a target object to be predicted, where the target object is any aviation device;

an obtaining module 602, configured to input the aviation operation data into a target prediction model, and obtain a repair error prediction result corresponding to a target object, where the target prediction model is used to predict a repair error of the target object.

Through the modules, the obtained aviation operation data are input into the target prediction model, the types and reasons of possible maintenance errors are predicted, and the target prediction model is obtained by continuously adjusting and optimizing model parameters of the initial prediction model by using a genetic algorithm, so that the prediction accuracy of the target prediction model on various aviation operation data is improved. The problem that the error type and reason which may appear in the aviation maintenance process cannot be accurately predicted in the related technology is solved.

As an alternative embodiment, the obtaining module includes:

the preprocessing unit is used for preprocessing historical aviation operation data to obtain preprocessed historical aviation operation data;

the obtaining unit is used for respectively taking each group of preprocessed historical aviation operation data and corresponding historical maintenance error types and reasons as a historical data sample to obtain a first preset number of historical data samples;

the device comprises a first generation unit, a second generation unit and a third generation unit, wherein the first generation unit is used for acquiring a second preset number of historical data samples from a first preset number of historical data samples and generating a training sample set;

the second generation unit is used for generating an initial prediction model according to the training sample set;

and the optimization unit is used for optimizing the initial prediction model through a target optimization algorithm based on the training sample set to obtain a target prediction model.

As an alternative embodiment, the optimization unit comprises:

the first obtaining submodule is used for inputting historical aviation operation data in the test sample set into the target prediction model to obtain a corresponding historical maintenance error prediction result;

The first initialization submodule is used for initializing model parameters in the initial prediction model;

the second initialization submodule is used for initializing a population of the target optimization algorithm and corresponding population parameters, wherein the population comprises a fifth preset number of encoding parameters as individuals, and the population parameters comprise the evolution times of the population, threshold values of the evolution times, cross probabilities and variation probabilities;

the updating submodule is used for updating the population and the evolution times according to the fitness, the cross probability, the variation probability and a preset method, judging whether the evolution times reach a threshold value, if not, judging whether individuals with the fitness smaller than the preset threshold value exist in the population, if so, taking the individuals with the fitness smaller than the preset threshold value as target individuals and stopping updating the population and the evolution times, and if so, taking the individuals with the minimum fitness in the current population as the target individuals, wherein the coding parameters corresponding to the target individuals are optimal solutions;

and the fifth obtaining submodule is used for modifying the model parameters of the initial prediction model into target model parameters to obtain the target prediction model.

As an alternative embodiment, the third deriving submodule includes:

the first obtaining subunit is configured to modify the model parameters of the initial prediction model into decoding parameters to obtain a fifth preset number of intermediate prediction models, where the intermediate prediction models are used to obtain a fitness;

the training subunit is used for sequentially training the intermediate prediction models through a training sample set to obtain training difference values;

the test subunit is used for sequentially testing the intermediate prediction models through the test sample set to obtain a test difference value;

As an alternative embodiment, the pre-processing unit comprises:

As an alternative embodiment, the second generating unit includes:

the determining submodule is used for determining the input layer neuron number and the output layer neuron number of the neural network contained in the initial prediction model based on the training sample set;

and the second generation submodule is used for generating an initial prediction model according to the input layer neuron number, the output layer neuron number, the target hidden layer neuron number and the target transfer function pair.

As an alternative embodiment, the seventh obtaining sub-module includes:

an obtaining subunit, configured to obtain a fourth preset number of combinations, where each combination includes a hidden layer neuron number and a transfer function pair;

a third obtaining subunit, configured to obtain, according to a preset formula, preset indexes of the initial prediction model corresponding to each combination respectively;

and the sub-unit is used for respectively taking the hidden layer neuron number and the transfer function pair corresponding to the initial prediction model with the minimum preset index as a target hidden layer neuron number and a target transfer function pair.

It should be noted here that the modules described above are the same as the examples and application scenarios implemented by the corresponding steps, but are not limited to the disclosure of the above embodiments.

Fig. 7 is a block diagram of an alternative electronic device according to an embodiment of the present application, as shown in fig. 7, including a processor 701, a communication interface 702, a memory 703 and a communication bus 704, where the processor 701, the communication interface 702 and the memory 703 complete communication with each other through the communication bus 704, where,

a memory 703 for storing a computer program;

the processor 701 is configured to implement the following steps when executing the computer program stored in the memory 703:

Alternatively, in this embodiment, the communication bus may be a PCI (Peripheral Component Interconnect) bus, an EISA (Extended Industry Standard Architecture) bus, or the like. The communication bus may be divided into an address bus, a data bus, a control bus, etc. For ease of illustration, only one thick line is shown in FIG. 7, but this is not intended to represent only one bus or type of bus.

The communication interface is used for communication between the electronic equipment and other equipment.

The memory may include RAM, and may also include non-volatile memory (non-volatile memory), such as at least one disk memory. Alternatively, the memory may be at least one memory device located remotely from the processor.

As an example, as shown in fig. 7, the memory 703 may include, but is not limited to, an obtaining module 601 and an obtaining module 602 in the repair error prediction apparatus. In addition, other module units in the repair error prediction apparatus may also be included, but are not limited to, and are not described in this example again.

The processor may be a general-purpose processor, and may include but is not limited to: a CPU (Central Processing Unit), NP (Network Processor), and the like; but also a DSP (Digital Signal Processing), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array) or other Programmable logic device, discrete Gate or transistor logic device, discrete hardware component.

Optionally, the specific examples in this embodiment may refer to the examples described in the above embodiments, and this embodiment is not described herein again.

It can be understood by those skilled in the art that the structure shown in fig. 7 is only an illustration, and the device implementing the repair error prediction method may be a terminal device, and the terminal device may be a terminal device such as a smart phone (e.g., an Android phone, an iOS phone, etc.), a tablet computer, a palm computer, a Mobile Internet Device (MID), a PAD, and the like. Fig. 7 does not limit the structure of the electronic device. For example, the terminal device may also include more or fewer components (e.g., network interfaces, display devices, etc.) than shown in FIG. 7, or have a different configuration than shown in FIG. 7.

Those skilled in the art will appreciate that all or part of the steps in the methods of the above embodiments may be implemented by a program instructing hardware associated with the terminal device, where the program may be stored in a computer-readable storage medium, and the storage medium may include: flash disk, ROM, RAM, magnetic or optical disk, and the like.

According to still another aspect of an embodiment of the present application, there is also provided a storage medium. Alternatively, in the present embodiment, the storage medium may be configured to store a program code for executing the repair error prediction method.

Optionally, in this embodiment, the storage medium may be located on at least one of a plurality of network devices in a network shown in the above embodiment.

Optionally, in this embodiment, the storage medium is configured to store program code for performing the following steps:

Optionally, the specific example in this embodiment may refer to the example described in the above embodiment, which is not described again in this embodiment.

Optionally, in this embodiment, the storage medium may include, but is not limited to: various media capable of storing program codes, such as a U disk, a ROM, a RAM, a removable hard disk, a magnetic disk, or an optical disk.

In the description of the present specification, reference to the description of the terms "this embodiment," "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., means that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the present disclosure. In this specification, the schematic representations of the terms used above are not necessarily intended to refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, various embodiments or examples and features of different embodiments or examples described in this specification can be combined and combined by one skilled in the art without contradiction. In the description of the present disclosure, "plurality" means at least two, e.g., two, three, etc., unless explicitly defined otherwise.

It should be understood that the above examples are only for clarity of illustration and are not intended to limit the embodiments. Other variations and modifications will be apparent to persons skilled in the art in light of the above description. And are neither required nor exhaustive of all embodiments. And obvious variations or modifications derived therefrom are intended to be within the scope of the invention.

Claims

1. A repair error prediction method, the method comprising:

2. The method of claim 1, wherein prior to said inputting said aviation operation data into a target prediction model, said method further comprises:

acquiring a first preset quantity group of historical aviation operation data and corresponding historical maintenance error types and reasons;

preprocessing the historical aviation operation data to obtain preprocessed historical aviation operation data;

generating an initial prediction model according to the training sample set;

and optimizing the initial prediction model through a target optimization algorithm based on the training sample set to obtain the target prediction model.

3. The method of claim 2, wherein after the optimizing the initial prediction model by a target optimization algorithm based on the training sample set to obtain the target prediction model, the method further comprises:

inputting the historical aviation operation data in the test sample set into the target prediction model to obtain a corresponding historical maintenance error prediction result;

obtaining a prediction error of the target prediction model according to the comparison result;

and judging whether the prediction error is within a preset range, if so, predicting the target object maintenance error by the target prediction model.

4. The method of claim 2, wherein the preprocessing the historical aerial operational data to obtain preprocessed historical aerial operational data comprises:

removing noise and abnormal data in the historical aviation operation data to obtain clean data;

5. The method of claim 2, wherein the generating an initial predictive model from the set of training samples comprises:

obtaining the number of the neurons of the initial hidden layer according to the number of the neurons of the input layer, the number of the neurons of the output layer and a target formula;

adjusting the initial hidden layer neuron number and the initial transfer function pair according to a preset index to respectively obtain a target hidden layer neuron number and a target transfer function pair;

and generating the initial prediction model according to the input layer neuron number, the output layer neuron number, the target hidden layer neuron number and the target transfer function pair.

6. The method of claim 5, wherein the adjusting the initial hidden layer neuron number and the initial transfer function pair according to a preset index to obtain a target hidden layer neuron number and a target transfer function pair respectively comprises:

obtaining a fourth preset number of combinations, wherein each combination comprises a hidden layer neuron number and a transfer function pair;

respectively obtaining the preset indexes of the initial prediction model corresponding to each combination according to a preset formula;

and respectively taking the hidden layer neuron number and the transfer function pair corresponding to the initial prediction model with the minimum preset index as the target hidden layer neuron number and the target transfer function pair.

7. The method of claim 3, wherein the optimizing the initial prediction model by a target optimization algorithm based on the training sample set to obtain the target prediction model comprises:

initializing model parameters in the initial prediction model;

coding the model parameters to obtain coding parameters;

initializing a population of the target optimization algorithm and corresponding population parameters, wherein the population comprises a fifth preset number of the coding parameters as individuals, and the population parameters comprise the evolution times of the population, threshold values of the evolution times, cross probabilities and variation probabilities;

acquiring a fitness function of the target optimization algorithm;

updating the population and the evolution times according to the fitness, the cross probability, the mutation probability and a preset method, judging whether the evolution times reach the threshold value, if not, judging whether individuals with fitness lower than the preset threshold value exist in the population, if so, taking the individuals with the fitness lower than the preset threshold value as target individuals and stopping updating the population and the evolution times, and if so, taking the individuals with the minimum fitness in the current population as the target individuals, wherein the coding parameters corresponding to the target individuals are optimal solutions;

and modifying the model parameters of the initial prediction model into the target model parameters to obtain the target prediction model.

8. The method of claim 7, wherein the deriving the fitness of the individual according to the fitness function, the decoding parameters, the training sample set, the testing sample set, and the initial prediction model comprises:

modifying the model parameters of the initial prediction model into the decoding parameters to obtain a fifth preset number of intermediate prediction models, wherein the intermediate prediction models are used for obtaining the fitness;

sequentially training the intermediate prediction models through the training sample set to obtain training difference values;

9. A repair error prediction apparatus, comprising:

the system comprises an acquisition module, a prediction module and a prediction module, wherein the acquisition module is used for acquiring aviation operation data of a target object to be predicted, and the target object is any aviation device;

10. An electronic device comprising a processor, a communication interface, a memory and a communication bus, wherein said processor, said communication interface and said memory communicate with each other via said communication bus,

the memory for storing a computer program;

the processor for performing the method steps of any one of claims 1 to 7 by running the computer program stored on the memory.