A method for predicting the health status of combat systems

By combining the hybrid prediction method of autoregressive moving average and artificial neural network, the problem of health status prediction of modern combat systems is solved, efficient and accurate health status prediction is achieved, the maintenance needs of modern warfare is met, and the overall support level of combat systems is improved. .

CN116523120BActive Publication Date: 2024-11-12CHENGDU AIRCRAFT INDUSTRY GROUP
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202310405347.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-04-14
Publication Date
2024-11-12
Estimated Expiration
2043-04-14

AI Technical Summary

Technical Problem

现代作战系统的传统维修模式无法满足现代战争对任务出勤率和战备完好率的要求,缺乏高效准确的健康状态预测手段以确定最佳维修时机和方案。

Method used

A hybrid prediction method combined with autoregressive moving average (ARIMA) model and artificial neural network (ANN) is used to establish linear and nonlinear submodels of time series to combine prediction results to improve the prediction accuracy of combat system health status.

Benefits of technology

It realizes efficient and accurate prediction of the health status of the combat system, provides the best maintenance time and plan, and improves the overall support level of the combat system.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116523120B_ABST
    Figure CN116523120B_ABST
Patent Text Reader

Abstract

The present invention belongs to the field of aircraft flight information technology, and in particular, relates to a method for predicting the health status of a combat system, comprising the following steps: Step 1: Establishing an autoregressive moving average model; Step 2: Establishing an artificial neural network model; Step 3: Establishing a hybrid prediction mathematical model for the combat system; Step 4: Using the hybrid model to achieve system health status prediction. The present invention uses actual data and prediction cases to show that the time series prediction based on the autoregressive sum moving average model and the combined data-driven prediction method based on the artificial neural network meet the important needs of upgrading the modern combat system support technology. The algorithm provided can meet the actual needs of combat system health status prediction to a considerable extent, with high prediction efficiency and strong guiding value for the prediction results. It has strong technical promotion and engineering reference significance for improving the overall support level of the combat system.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of aircraft flight informationization, and in particular relates to a method for predicting the health status of a combat system. Background Art

[0002] The modern combat system is the command center of weapons and a typical "information-intensive" system. As the combat system develops towards modularization, intelligence, large-scale, complexity and integration, the traditional maintenance and support mode of preventive maintenance and post-maintenance can no longer meet the requirements of modern warfare for the mission attendance rate and combat readiness rate of weapons and equipment. In this context, the situation-based maintenance strategy based on fault prediction and health management (PHM) technology has been highly valued by equipment users. Researching an efficient and accurate health status prediction method for modern combat systems in order to determine the best maintenance time and the best maintenance plan based on the deterioration trend of their health status has become an important requirement for the upgrading of combat system support technology. Summary of the invention

[0003] The purpose of the present invention is to overcome the shortcomings of the above-mentioned background technology and provide a method for predicting the health status of a combat system.

[0004] In order to achieve the above technical effects, the present invention is implemented by the following technical solutions:

[0005] A method for predicting the health status of a combat system comprises the following steps:

[0006] Step 1: Establish an autoregressive moving average (ARIMA) model:

[0007] Step 2: Establish an artificial neural network (ANN) model;

[0008] Step 3: Establish a hybrid prediction mathematical model for the combat system;

[0009] Step 4: Use the hybrid model to predict the system health status.

[0010] The first step includes: establishing a mathematical model based on the rules obtained from time series analysis, observing and measuring a certain variable x(t) or a group of variables, and converting the variables x(t) into a series of variables at time t. 1 , t 1 ,…,t n , (t is the time independent variable) are arranged in time order and used to explain the mathematical expressions of variables and their relationships; the resulting discrete numbers form a sequence set x(t 1 ), x(t 2 ), …, x(t n), called time series. By analyzing these data arranged in time sequence, we can find the laws of their internal changes;

[0011] The step 2 includes: the types of processing units are divided into three categories: input units, output units and hidden units; the input units receive signals and data from the outside world; the output units realize the output of system processing results; the hidden units are units between the input and output units and cannot be observed from outside the system. The connection weights between neurons reflect the connection strength between units. The representation and processing of information are reflected in the connection relationship of the network processing units.

[0012] The step three includes: using an artificial neural network to model the nonlinear component of the time series, and then combining the two parts of the results to obtain a higher prediction accuracy;

[0013] The specific steps are as follows:

[0014] Step 1: The prediction steps based on time series are roughly divided into four steps:

[0015] (1) Collect and organize historical data, form time series, and draw statistical charts.

[0016] (2) Analyze time series: In a time series, each set of data of the forecast object is the result of many different factors acting simultaneously. These data can be discrete or regular. By analyzing the time series and mining the hidden and valuable information, it plays a vital role in the selection of forecast models.

[0017] (3) Identification model and curve fitting: By analyzing the time series, valuable information is extracted. According to its characteristics and rules, combined with the advantages and disadvantages of the prediction model, a suitable prediction model is selected to fit a curve that can reflect the change trend of the predicted object in a certain period of time in the future;

[0018] (4) Predict the future based on the curve: By analyzing and reasoning the fitted curve, the fault of the predicted object can be predicted at a certain moment in the future as time passes.

[0019] Step 2: For the combat system, the artificial neural network model and its prediction process for predicting its health status can be illustrated by a three-layer forward neural network. 1 , x 2 , …x n is the n input of the network, y 1 ,y 2 ,…y m is the m output of the network, ω ij is the connection weight between the i-th node in the input layer and the j-th node in the intermediate layer, ω jkis the connection weight between the jth node in the middle layer and the mth node in the output layer. When it is used to predict the health status of the combat system, the input vector X in the input layer is (x 1 , x 2 , …x n ) is the status monitoring data of the combat system or its lower-level unit equipment. After certain preprocessing (such as noise reduction, normalization, etc.), these data are used as input parameters of the prediction model. The middle layer of the prediction model is also called the hidden layer, which can be a single layer or a multi-layer structure. ij and ω jk is to connect the input layer and the output layer. The output value of the output layer is the predicted value, and the number of nodes in the output layer m is the total number of predicted results. 1 ,y 2 ,…y m ) is the output result of this prediction. Neural networks mainly realize prediction functions through two methods. The first method uses neural networks as function approximators to fit and predict parameters. The second method considers the dynamic relationship between input and output and uses dynamic neural networks with feedback to establish dynamic models for parameters for prediction. In the process of prediction based on neural network models, firstly, the state monitoring data is used as samples to select reasonable training, test and analysis samples; then the training model is set through network parameters; then the trained network model is tested with test samples to verify the network performance; finally, the model and analysis samples are used for prediction.

[0020] Step 3: The mixed model consists of two parts: the ARIMA linear sub-model and the ANN nonlinear sub-model.

[0021] (1) ARIMA linear sub-model: Assume that the time series input is X = (x1, x2, ... xn), and its true expected output is Y = (y1, y2, ... ym). First, use the ARIMA model to predict the time series, and then you can get the ARIMA prediction value

[0022]

[0023] (2) ANN nonlinear sub-model: According to the basic principle of artificial neural network model, the prediction result of nonlinear sub-model is assumed to be

[0024]

[0025] Among them, w j (j=1, 2, 3, ..., q) and w ij (i=1, 2, 3, ..., p; j=1, 2, 3, ..., q) are the connection weights of the neural network. p and q represent the number of nodes in the input layer and the middle layer of the network respectively. Usually, the output layer is 1 for one-step forward prediction. 0With b 0j is the bias term, εt is the prediction error at time t, and g is the activation function of the network.

[0026] (3) Hybrid model: Combining equations (1) and (2), the hybrid model expression is:

[0027]

[0028] Among them, Y t is the final prediction result, Y L (t) is the forecast result of the autoregressive integrated moving average model ARIMA, Y N (t) is the residual of ARIMA prediction. N (t) As the input of the artificial neural network ANN model, the ANN model is trained to obtain the ANN model prediction results.

[0029] Since the autoregressive sum moving average model cannot capture the nonlinear part of the time series, the residual term obtained by formula (3) contains the nonlinear component of the time series. Using artificial neural network to model the residual term and then combining the two parts of the results will result in a higher prediction accuracy.

[0030] Step 4: Use the ARIMA and ANN hybrid model to implement the process steps of combat system health status prediction. The following is a detailed description of the hybrid model prediction algorithm:

[0031] The basic scenario of prediction is to predict the time series X = (x 1 , x 2 , …, x t )(t is the current time) to perform hybrid dynamic L-step prediction. The initial time k=t, j=1 (j is the number of prediction cycles), and N is the number of prediction data.

[0032] 1: Stabilize and preprocess the data according to the characteristics of the sample data. Suppose the processed data sequence is That is, the training data.

[0033] 2: Model identification, that is, determining the model structure, and estimating the autoregressive order n and the moving average order m according to the properties of the autocorrelation coefficient (ACF) and partial correlation coefficient (PACF) of the data series and the AIC criterion.

[0034] 3: Estimate model parameters based on the least squares method to determine the relationship between the current observation value and the historical observation value and the white noise sequence.

[0035] 4: Use the training data to check whether the model has reached the accuracy. If it meets the requirement, a reasonable ARIMA model is obtained, and thus the static multi-step prediction error etrain(t) (i.e., the training data of the ANN) is obtained, then go to step 5; otherwise, go to step 2.

[0036] 5: Substitute the historical data into the prediction equation to obtain the data at time k + 1

[0037] 6: If k + 1 - n < L, then k = k + 1, add the predicted data to the sequence, go to step 3, and re-estimate the parameters; otherwise, go to step 7.

[0038] 7: If L j < N, go to step 8; otherwise, obtain the prediction result Go to step 9.

[0039] 8: Replace the previous L-step predicted values with the actual observed values as the time series, go to step 3, and perform the L-step prediction for the (j = j + 1)-th cycle.

[0040] 9: Use etrain(t) in step 4 to train the ANN model, and obtain the prediction residual Y (t) from the prediction result obtained by ARIMA in step 7 as the time series data of the ANN. Similar to steps 5, 6, 7, and 8, obtain the prediction result of the non-linear part N (t), as the time series data of the ANN. Similar to steps 5, 6, 7, and 8, obtain the prediction result of the non-linear part

[0041] 10: Obtain the prediction result of the hybrid model from and End. End.

[0042] The advantages of the present invention are as follows:

[0043] The present invention deeply analyzes the business and fault characteristics of the combat system, gives a data-driven prediction method combining time series prediction based on autoregressive integrated moving average model (ARIMA) and artificial neural network (ANN), gives a hybrid prediction mathematical model for their combination, and specific prediction steps based on this method. Taking a subsystem of the combat system as an example, actual monitoring data is introduced, and an application case of the health state prediction algorithm is given.

[0044] Through actual data and prediction cases, it is shown that the time series prediction based on the autoregressive integrated moving average model (ARIMA) and the combined data-driven prediction method based on the artificial neural network (ANN) meet the important needs of the modern combat system support technology upgrade. The algorithm given can meet the actual needs of the combat system health status prediction to a considerable extent. The prediction efficiency is high, and the prediction results have strong guiding value. It has strong technical promotion and engineering reference significance for improving the overall support level of the combat system. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 It is a schematic diagram of the process of the present invention.

[0046] Figure 2 It is the basic process step of time series based forecasting.

[0047] Figure 3 It is a schematic diagram of a neural network used to predict the health status of a combat system.

[0048] Figure 4 It is a flowchart for using hybrid model to predict the health status of combat system. DETAILED DESCRIPTION

[0049] In order to more clearly illustrate the technical solution provided by the present invention, the present invention will be further described below in conjunction with the accompanying drawings and embodiments. It should be noted that the embodiments provided are only part of the embodiments of the present invention, rather than all the embodiments, and therefore should not be regarded as limiting the scope of protection. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technical personnel in this field without making creative work are within the scope of protection of the present invention.

[0050] A method for predicting the health status of a combat system comprises the following steps:

[0051] Step 1: Establish an autoregressive moving average (ARIMA) model:

[0052] The data sequence of the predicted object system that changes over time is regarded as a random sequence. After establishing a suitable mathematical model for the sequence, the future value is predicted through the past and present values ​​of the time series. The time series prediction method compiles, analyzes and infers the time series formed by the system equipment data, and makes comparisons, analogies and extensions based on the development process, change direction and formation trend reflected in the time series to make predictions for the next period of time.

[0053] Step 2: Build an Artificial Neural Network (ANN) model

[0054] A neural network is a computational model that consists of a large number of interconnected nodes (or neurons). Each node represents a specific output function, called an activation function. The connection between each two nodes represents a weighted value for the signal passing through the connection, called a weight, which is equivalent to the memory of the artificial neural network. The output of the network varies depending on the network connection method, weight value and activation function. The larger the weight, the greater the impact of the input signal on the neuron. The weight can be negative, which means that the input signal is suppressed. Different weights will result in different calculations for neurons. By adjusting the weights, the required output value can be obtained under fixed input. However, when the ANN is composed of hundreds or thousands of neurons, it becomes extremely complicated to manually calculate these weights. At this time, some algorithmic skills are needed. The process of adjusting the weights is called the "learning" or "training" process of the ANN.

[0055] Step 3: Establish a hybrid forecasting mathematical model for combat systems: Since the autoregressive moving average model is suitable for capturing the linear part of the time series, the error is often large when solving complex nonlinear problems. Neural networks work better in predicting nonlinear time series, but they perform poorly in predicting linear time series. Therefore, this study combines the ARIMA and ANN methods to establish a hybrid model that has the advantages of both models, so that it can better predict time series and improve prediction accuracy.

[0056] Step 4: Use hybrid models to predict system health status: Aiming at the engineering goal of predicting the health status of the combat system, fully consider the impact of real-time data on model parameters during the prediction process, combine the prediction process of ARIMA and ANN, and establish a real-time dynamic prediction model.

[0057] Step 1 includes: building a mathematical model based on the rules obtained from time series analysis, observing and measuring a certain variable or a group of variables x(t), and converting the data into 1 , t 1 ,…,t n , (t is the time independent variable) are arranged in time order and used to explain the mathematical expressions of variables and their relationships; the resulting discrete numbers form a sequence set x(t 1 ), x(t 2 ), …, x(t n ), called time series, this kind of sequence with time significance is also called dynamic data. By analyzing these data arranged in time order, we can find the law of their internal changes;

[0058] Step 2 includes: In artificial neural networks, neuron processing units can represent different objects, such as features, letters, concepts, or some meaningful abstract patterns. The types of processing units in the network are divided into three categories: input units, output units, and hidden units; input units receive signals and data from the outside world; output units realize the output of system processing results; hidden units are units between input and output units and cannot be observed from outside the system. The connection weights between neurons reflect the connection strength between units. The representation and processing of information are reflected in the connection relationship of network processing units; artificial neural networks are a non-programmed, adaptive, brain-style information processing. Its essence is to obtain a parallel distributed information processing function through the transformation and dynamic behavior of the network, and imitate the information processing function of the human brain nervous system to varying degrees and levels;

[0059] Step three includes: since the autoregressive sum moving average model cannot capture the nonlinear part of the time series, the nonlinear component of the time series is modeled using an artificial neural network, and then the two parts of the results are combined to obtain a higher prediction accuracy;

[0060] Step four includes: The main research content of the platform is to realize the integration, analysis and sharing of data generated in various heterogeneous environments during the production process of high-tech product development laboratories, including test data, data models, business processes, facility resources, knowledge and standards, and provide multi-faceted support such as process assurance, environmental assurance, personnel assurance and condition assurance in scientific research and production.

[0061] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0062] Reference Figure 1 As shown, a method for predicting the health status of a combat system of the present invention comprises the following steps:

[0063] Step 1: If Figure 2 As shown, the prediction steps based on time series are roughly divided into four steps:

[0064] (1) Collect and organize historical data, form time series, and draw statistical charts.

[0065] (2) Analyze time series: In a time series, each set of data of the forecast object is the result of many different factors acting simultaneously. These data can be discrete or regular. By analyzing the time series and mining the hidden and valuable information, it plays a vital role in the selection of forecast models.

[0066] (3) Identification model and curve fitting: By analyzing the time series, valuable information is extracted. According to its characteristics and rules, combined with the advantages and disadvantages of the prediction model, a suitable prediction model is selected to fit a curve that can reflect the change trend of the predicted object in a certain period of time in the future;

[0067] (4) Predict the future based on the curve: By analyzing and reasoning the fitted curve, the fault of the predicted object can be predicted at a certain moment in the future as time passes.

[0068] Step 2: If Figure 3 As shown, for the combat system, the artificial neural network model and its prediction process for predicting its health status can be schematically illustrated by a three-layer forward neural network. 1 , x 2 , …x n is the n input of the network, y 1 ,y 2 ,…y m is the m output of the network, ω ij is the connection weight between the i-th node in the input layer and the j-th node in the intermediate layer, ω jk is the connection weight between the jth node in the middle layer and the mth node in the output layer. When it is used to predict the health status of the combat system, the input vector X in the input layer is (x 1 , x 2 , …x n ) is the status monitoring data of the combat system or its lower-level unit equipment. After certain preprocessing (such as noise reduction, normalization, etc.), these data are used as input parameters of the prediction model. The middle layer of the prediction model is also called the hidden layer, which can be a single layer or a multi-layer structure. ij and ω jk is to connect the input layer and the output layer. The output value of the output layer is the predicted value, and the number of nodes in the output layer m is the total number of predicted results. 1 ,y 2 ,…y m ) is the output result of this prediction. Neural networks mainly realize prediction functions through two methods. The first method uses neural networks as function approximators to fit and predict parameters. The second method considers the dynamic relationship between input and output and uses dynamic neural networks with feedback to establish dynamic models for parameters for prediction. In the process of prediction based on neural network models, firstly, the state monitoring data is used as samples to select reasonable training, test and analysis samples; then the training model is set through network parameters; then the trained network model is tested with test samples to verify the network performance; finally, the model and analysis samples are used for prediction.

[0069] Step 3: The mixed model consists of two parts: the ARIMA linear sub-model and the ANN nonlinear sub-model.

[0070] (1) ARIMA linear sub-model: Assume that the time series input is X = (x1, x2, ... xn), and its true expected output is Y = (y1, y2, ... ym). First, use the ARIMA model to predict the time series, and then you can get the ARIMA prediction value

[0071]

[0072] (2) ANN nonlinear sub-model: According to the basic principle of artificial neural network model, the prediction result of nonlinear sub-model is assumed to be

[0073]

[0074] Among them, w j (j=1, 2, 3, ..., q) and w ij (i=1, 2, 3, ..., p; j=1, 2, 3, ..., q) are the connection weights of the neural network. p and q represent the number of nodes in the input layer and the middle layer of the network respectively. Usually, the output layer is 1 for one-step forward prediction. 0 With b 0j is the bias term, εt is the prediction error at time t, and g is the activation function of the network.

[0075] (3) Hybrid model: Combining equations (1) and (2), the hybrid model expression is:

[0076]

[0077] Among them, Y t is the final prediction result, Y L (t) is the forecast result of the autoregressive integrated moving average model ARIMA, Y N (t) is the residual of ARIMA prediction. N (t) As the input of the artificial neural network ANN model, the ANN model is trained to obtain the ANN model prediction results.

[0078] Since the autoregressive sum moving average model cannot capture the nonlinear part of the time series, the residual term obtained by formula (3) contains the nonlinear component of the time series. Using artificial neural network to model the residual term and then combining the two parts of the results will result in a higher prediction accuracy.

[0079] Step 4: If Figure 4 As shown in the figure, the process steps of using the ARIMA and ANN hybrid model to realize the health status prediction of the combat system are described below:

[0080] The basic predicted scenario is to perform a hybrid dynamic L-step prediction on the time series X = (x 1 , x 2 , …, x t )(where t is the current moment). The initial moment is k = t, j = 1 (j is the prediction loop count), and N is the number of prediction data points.

[0081] Step 1: Perform stationarization and data preprocessing according to the characteristics of the sample data. Let the processed data sequence be That is, the training data.

[0082] Step 2: Model identification, that is, determine the model structure, and estimate the autoregressive order n and the moving average order m according to the properties of the autocorrelation function (ACF) and the partial autocorrelation function (PACF) of the data sequence and the AIC criterion.

[0083] Step 3: Estimate the model parameters based on the least squares method to determine the relationship between the observed value at the current moment and the observed values at historical moments and the white noise sequence.

[0084] Step 4: Use the training data to check whether the model reaches the accuracy. If it is satisfied, that is, a reasonable ARIMA model is obtained, and thus the static multi-step prediction error etrain(t) (which is also the training data of the ANN) is obtained, then go to Step 5; otherwise, go to Step 2.

[0085] Step 5: Substitute the historical data into the prediction equation to obtain the data at the k + 1 moment

[0086] Step 6: If k + 1 - n < L, then k = k + 1, add the predicted data to the sequence, go to Step 3, and re-estimate the parameters; otherwise, go to Step 7.

[0087] Step 7: If L j < N, go to Step 8; otherwise, obtain the prediction result Go to Step 9.

[0088] Step 8: Replace the previous L-step predicted values with the actual observed values as the time series, go to Step 3, and perform the L-step prediction for the j = j + 1th loop.

[0089] Step 9: Use the etrain(t) in Step 4 to train the ANN model, and obtain the prediction residuals Y (t) from the prediction result obtained by ARIMA in Step 7 as the time series data of the ANN. Follow the same steps as Step 5, Step 6, Step 7, and Step 8 to obtain the prediction result of the non-linear part N (t), as the time series data of the ANN. Follow the same steps as Step 5, Step 6, Step 7, and Step 8 to obtain the prediction result of the non-linear part

[0090] Step 10: By and Get the prediction results of the hybrid model Finish.

Claims

1. A method for predicting the health status of a combat system, characterized in that: It includes the following steps: Step 1: Establish an autoregressive integrated moving average (ARIMA) model; Step 2: Establish an artificial neural network (ANN) model; Step 3: Establish a hybrid prediction mathematical model for the combat system; Step 4: Use the hybrid model to achieve the prediction of the system health state; In Step 3, the non-linear components of the time series are modeled using an artificial neural network and then combined with the results of the autoregressive integrated moving average model to obtain a higher prediction accuracy; The hybrid model in Step 3 consists of two parts: an autoregressive integrated moving average linear sub-model and an artificial neural network non-linear sub-model. The specific steps are as follows: (1) Autoregressive integrated moving average linear sub-model: Assume that for the time series input X = (x1, x2,... xn), its true expected output is Y = (y1, y2,... ym); First, use the autoregressive integrated moving average model to predict the time series, and then obtain the predicted value of the autoregressive integrated moving average (2) Artificial neural network non-linear sub-model: According to the basic principle of the artificial neural network model, assume that the predicted result of the non-linear sub-model is Among them, w j (j=1, 2, 3, ..., q) and w ij (i=1, 2, 3, ..., p; j=1, 2, 3, ..., q) are the connection weights of the neural network. p and q represent the number of nodes in the input layer and the middle layer of the network respectively. The output layer is 1 for one-step forward prediction. b0 and b 0j is the bias term, εt is the prediction error at time t, and g is the activation function of the network; (3) Hybrid model: Combining equations (1) and (2), the expression of the hybrid model is obtained as: Among them, Y t is the final prediction result, Y L (t) is the prediction result of the autoregressive sum moving average model, Y N (t) is the residual of the autoregressive moving average prediction; Y N (t) serving as an input of the artificial neural network model, training the artificial neural network model, and obtaining the prediction result of the artificial neural network model; The process steps of using the autoregressive integrated moving average and artificial neural network hybrid model to achieve the prediction of the combat system health state in Step 4 are as follows. The detailed process of the hybrid model prediction algorithm is as follows: The basic scenario of prediction is to predict the time series X = (x1, x2, ..., x t ) performs hybrid dynamic L-step prediction, where t is the current time, the initial time k=t, j=1, where j is the number of prediction cycles, and N is the number of prediction data; Step 1. Perform stabilization and data preprocessing according to the characteristics of the sample data; suppose the processed data sequence is That is, training data; Step 2. Model identification, determine the model structure, and estimate the autoregressive order n and the moving average order m according to the properties of the autocorrelation function (ACF) and partial autocorrelation function (PACF) of the data sequence and the AIC criterion; Step 3. Estimate the model parameters based on the least squares method to determine the relationship between the observed value at the current moment and the observed values at historical moments and the white noise sequence; Step 4. Use the training data to check whether the model reaches the accuracy. If it is satisfied, a reasonable autoregressive integrated moving average model is obtained, and thus the static multi-step prediction error etrain(t) is obtained, which is the training data of the artificial neural network. Go to Step 5; otherwise, go to Step 2; Step 5. Historical data Substitute into the prediction equation and get the k+1 time data Step 6. If k + 1 - n < L, then k = k + 1, add the predicted data to the sequence, go to Step 3, and re-estimate the parameters; otherwise, go to Step 7; Step 7. If L j < N, go to Step 8; otherwise, obtain the prediction result Go to Step 9; Step 8. Replace the previous L-step predicted values with the actual observed values as the time series, go to Step 3, and perform the L-step prediction for the j = j + 1th cycle; Step 9. Use etrain(t) from step 4 to train the artificial neural network model, and the prediction results obtained by the autoregressive moving average in step 7 Get the predicted residual Y N (t), as the time series data of the artificial neural network, the same as steps 5, 6, 7, and 8, to obtain the prediction results of the nonlinear part Step 10. and Get the prediction results of the mixed model Finish.

2. A method for predicting the health status of a combat system according to claim 1, characterized in that: In step 1, a mathematical model is established based on the rules obtained from time series analysis, and a certain variable x(t) or a group of variables is observed and measured. n , t is the time independent variable, n is a natural number, arranged in time order, and used to explain the mathematical expression of variables and their relationships; the obtained discrete numbers form a sequence set x(t1), x(t2), ..., x(t n ), called time series, and analyze these data arranged in time order.

3. A method for predicting the health status of a combat system according to claim 2, characterized in that: In Step 2, the types of processing units in the artificial neural network are divided into input units, output units, and hidden units; The input units receive signals and data from the external world; The output units implement the output of the system processing results; The hidden units are located between the input and output units, and the connection weights between neurons represent the connection strength between units. The representation and processing of information are reflected in the connection relationship of the network processing units.

4. A method for predicting the health status of a combat system according to claim 2, characterized in that: The prediction steps based on the time series in Step 1 include the following steps: Step 1. Collect and organize historical data to form a time series and draw a statistical chart; Step 2. Analyze the time series; Step 3. Identify model and curve fitting: By analyzing the time series, extract valuable information, select a suitable prediction model based on its characteristics and rules, combined with the advantages and disadvantages of the prediction model, and fit a curve that can reflect the change trend of the predicted object in a certain period of time in the future; Step 4. Predict the future based on the curve: By analyzing and reasoning the fitted curve, the fault of the predicted object can be predicted at a certain moment in the future as time passes.

5. A method for predicting the health status of a combat system according to claim 3, characterized in that: The artificial neural network model and its prediction process used for health status prediction in step 2 are illustrated by a three-layer feedforward neural network, where x1, x2, …x n are the n inputs of the network, y1, y2, ...y m is the m output of the network, ω ij is the connection weight between the i-th node in the input layer and the j-th node in the intermediate layer, ω jk is the connection weight between the jth node in the middle layer and the mth node in the output layer; When it is used to predict the health status of a combat system, the input vector X in the input layer is (x1, x2, ... x n ) is the status monitoring data of the combat system or its lower-level unit equipment; After preprocessing, it is used as the input parameter of the prediction model; the middle layer of the prediction model is the hidden layer, which is a one-layer or multi-layer structure. ij and ω jk is to connect the input layer and the output layer; the output value of the output layer is the predicted value, and the number of output layer nodes m is the total number of predicted results, Y = (y1, y2, ...y m ) is the output result of this prediction.

6. A method for predicting the health status of a combat system according to claim 5, characterized in that: In the prediction process based on the neural network model, the state monitoring data is first used as a sample to select training, test and analysis samples; then the training model is set through the network parameters; the trained network model is tested with the test sample to verify the network performance; finally, the model and analysis samples are used for prediction.

7. A method for predicting the health status of a combat system according to claim 5, characterized in that: The data preprocessing includes noise reduction and normalization.