Special vehicle quality performance prediction method and device, electronic equipment and medium

Abstract

The invention provides a special vehicle quality performance prediction method and device, an electronic device and a medium, dynamic tracking of model parameters is considered on the basis of an autoregression model (namely an AR model), a state-space equation for vehicle quality performance prediction is established, a determination method of related parameters is given, and the quality performance of a special vehicle is predicted. Dynamic prediction of quality performance indexes is realized according to Kalman filtering; on the basis, an analytic hierarchy process prediction method of the comprehensive quality performance indexes is established, and a fusion evaluation method of time sequence prediction and analytic hierarchy process prediction results is given based on error analysis, so that prediction propagation and error analysis from bottom to top can be reasonably realized; and effectively fusing prediction results based on the time sequence and the analytic hierarchy process, and dynamically predicting single-machine-level, sub-system-level and whole-vehicle-level quality performance indexes of the vehicle.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of vehicle performance prediction, in particular to a special vehicle quality performance prediction method and device, electronic equipment and computer readable storage medium. BACKGROUND

[0002] Accurately grasping the vehicle quality performance level, especially grasping the vehicle quality performance level trend, and dynamically predicting the future based on the quality performance historical evaluation results, can effectively support the task planning and training optimization of special vehicles. At present, performance parameter prediction based on time series analysis is a relatively mature method, however, in the actual application of special vehicles, two problems are mainly found. The first problem is that due to the change of user usage intensity, the change of product performance evolution law, the influence of sudden events and other factors, the evolution law of vehicle quality performance indicators may change over time, resulting in dynamic change of time series model parameters. The second problem is that special vehicles are typical complex systems, and the whole vehicle quality performance indicators are obtained by evaluating single machine level and subsystem level quality performance indicators. In comprehensive quality performance prediction, how to reasonably realize the bottom-up prediction propagation and error analysis, and how to effectively fuse the prediction results based on time series and based on hierarchical analysis are problems to be solved. SUMMARY

[0003] The present application aims to provide a special vehicle quality performance prediction method, device, electronic equipment and computer readable storage medium which can overcome the above problems or at least partially solve the above problems.

[0004] To achieve the above purpose, the technical scheme of the present application is as follows:

[0005] The first aspect of the present application provides a special vehicle quality performance prediction method, comprising:

[0006] establishing a state space equation of special vehicle quality performance prediction according to an AR model;

[0007] dynamically recursively estimating the AR model parameters according to Kalman filtering method, and predicting single quality performance indicators by time series method;

[0008] performing comprehensive quality performance prediction and error evaluation of special vehicles according to hierarchical analysis method;

[0009] fusing and evaluating the quality performance indicator prediction results obtained by time series method and the quality performance prediction results obtained by hierarchical analysis method.

[0010] Optionally, the state space equation of special vehicle quality performance prediction according to the AR model comprises:

[0011] An evaluation value of a vehicle mass performance index is acquired, and an m-order autoregressive model AR(m) is used to model an index evolution law:

[0012]

[0013] wherein Z t is an index evaluation value at the tth moment, β = [a1…a m b] T , H t = [Z t-1 Z t-2 …Z t-m 1], a1,...,a m is an autoregressive coefficient, b is an intercept, and ε is a Gaussian white noise with a mean of 0 and a standard deviation of σ;

[0014] A quality performance index historical data matrix is established and Maximum likelihood estimates of β and σ are calculated:

[0015]

[0016] The value of m is determined;

[0017] A covariance matrix of H is constructed:

[0018]

[0019] A state space model is constructed:

[0020]

[0021] wherein β t = [a 1,t a 2,t …a m,t b t ] T , w t-1 is state noise, v t is measurement noise, is a state noise covariance matrix, is a measurement noise variance, wherein E(·) is a mathematical expectation function, Q t-1 and are the same as the main diagonal elements of H

[0022] Optionally, the AR model parameters are dynamically recursively estimated according to the Kalman filtering method, and a single quality performance index is predicted by a time series method, including:

[0023] An optimal recursive estimation of the parameter β is given by using the state space model and Kalman filtering method.

[0024] After obtaining the latest state estimation of the parameter β, a corresponding prediction variance evaluation is performed according to a Monte Carlo method.

[0025] Optionally, the step of giving an optimal recursive estimation of the parameter β by using the state space model and Kalman filtering method comprises:

[0026] calculating a state prediction is: a state prediction error covariance matrix is P t / (t-1) = P t-1 + Q t-1 ; wherein an initial state value is a maximum likelihood estimation result an initial state covariance P0 is taken as Q t-1 ;

[0027] calculating a state estimation at time t is: a state estimation error covariance matrix is P t = (I - K t H t )P t / t-1 ; wherein K t is a filtering gain matrix, K t = P t / t-1 H t T (H t P t / t-1 H t T + R t ) -1 ;

[0028] Let t = t + 1, and repeat the steps of calculating a state prediction, a state prediction error covariance matrix, and calculating a state estimation at time t, a state estimation error covariance matrix to perform an optimal recursive estimation of the parameter β.

[0029] Optionally, the step of performing a corresponding prediction variance evaluation according to a Monte Carlo method after obtaining the latest state estimation of the parameter β comprises:

[0030] generating k random numbers from a normal distribution , denoted as ε1, ε2, …, ε k ;

[0031] calculating a prediction result

[0032]

[0033] Repeat the above step M times, respectively calculate the sample variance of M times , denoted as Var t+1 , Var t+2 , …, Var t+k , the variance of the prediction result ;

[0034] Calculate the quality performance index prediction interval with probability p:

[0035]

[0036] Optionally, the comprehensive quality performance prediction and error evaluation of special vehicles according to the analytic hierarchy process comprises:

[0037] According to the quality performance index prediction value of each single machine at t+1-t+k time And the corresponding prediction variance The weight of each component single machine in the system quality performance evaluation is obtained by using the analytic hierarchy process:

[0038] Calculate the quality performance index prediction value and prediction variance of the subsystem A at t+1-t+k time:

[0039]

[0040] Calculate the upper and lower limits of the quality performance index prediction with probability p:

[0041]

[0042] Optionally, the fusion evaluation of the comprehensive quality performance prediction based on time series and analytic hierarchy process comprises:

[0043] The quality performance index prediction value at t+1-t+k time obtained by the analytic hierarchy process is The corresponding prediction variance is Var AHP,t+1 , Var AHP,t+2 , …, Var AHP,t+k , and the quality performance index prediction value at t+1-t+k time obtained by the time series method is The corresponding prediction variance is Var AR-KM,t+1 , Var AR-KM,t+2 , …, Var AR-KM,t+k Fusion prediction is carried out, and the fusion prediction value is calculated:

[0044]

[0045] And the fusion prediction variance

[0046]

[0047] Where, ω AHP,t+k and ω AR-KM,t+k The weights are respectively obtained using the following formula:

[0048]

[0049] The upper and lower bounds for the prediction of probability p after fusion of comprehensive performance indicators are:

[0050]

[0051] A second aspect of the present invention provides a special vehicle quality performance prediction device, comprising:

[0052] A module is established to build the state-space equations for predicting the quality and performance of special vehicles based on the AR model.

[0053] The first prediction module is used to dynamically recursively estimate the parameters of the AR model based on the Kalman filter method and predict single quality performance indicators using the time series method.

[0054] The second prediction module is used to predict the comprehensive quality performance and error assessment of special vehicles based on the analytic hierarchy process.

[0055] The integrated prediction module is used to integrate and evaluate the quality performance index prediction results obtained from the time series method and the quality performance prediction results obtained from the analytic hierarchy process.

[0056] A third aspect of the present invention provides an electronic device, comprising: a processor and a memory;

[0057] The memory is used to store computer programs;

[0058] The processor is configured to execute the special vehicle quality performance prediction method as described above by invoking the computer program.

[0059] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the special vehicle quality performance prediction method as described above.

[0060] Therefore, the special vehicle quality performance prediction method, device, electronic equipment, and computer-readable storage medium provided by this invention establish a state-space equation for vehicle quality performance prediction based on the autoregressive model (AR model) and consider the dynamic tracking of model parameters. A method for determining relevant parameters is also provided, and dynamic prediction of quality performance indicators is achieved using Kalman filtering. Furthermore, a hierarchical analysis prediction method for comprehensive quality performance indicators is established, and a method for fusing and evaluating the results of time-series prediction and hierarchical analysis prediction based on error analysis is provided. This method can reasonably achieve bottom-up prediction propagation and error analysis, effectively fusing the prediction results based on time series and hierarchical analysis, and dynamically predicting vehicle quality performance indicators at the single-machine, subsystem, and whole-vehicle levels. Attached Figure Description

[0061] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0062] Figure 1 A flowchart of a special vehicle quality performance prediction method provided in an embodiment of the present invention;

[0063] Figure 2 A flowchart illustrating a specific example of the special vehicle quality performance prediction method provided in this embodiment of the invention;

[0064] Figure 3 This is a hierarchical diagram of a special vehicle quality performance evaluation structure provided in an embodiment of the present invention;

[0065] Figure 4 This is a schematic diagram illustrating the state recursive estimation of parameters in the autoregressive model of power supply and distribution system quality performance provided in an embodiment of the present invention.

[0066] Figure 5 This is a schematic diagram of the power supply and distribution system quality performance prediction results provided in an embodiment of the present invention;

[0067] Figure 6 This is a schematic diagram of the chassis powertrain quality performance prediction results provided in an embodiment of the present invention;

[0068] Figure 7 This is a schematic diagram of the vehicle quality performance prediction results provided in an embodiment of the present invention;

[0069] Figure 8 This is a schematic diagram of the structure of the special vehicle quality performance prediction device provided in an embodiment of the present invention;

[0070] Figure 9 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0071] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0072] Figure 1 A flowchart of the special vehicle quality performance prediction method provided in an embodiment of the present invention is shown. Figure 2 A flowchart illustrating a specific example of the special vehicle quality performance prediction method provided in this embodiment of the invention is shown below. Figure 1 and Figure 2 The special vehicle quality performance prediction method provided in this embodiment of the invention includes:

[0073] S1. Establish the state-space equation for predicting the quality performance of special vehicles based on the AR model;

[0074] S2, The AR model parameters are dynamically recursively estimated based on the Kalman filter method, and the single-mass performance index is predicted by the time series method.

[0075] S3, predict and assess the comprehensive quality performance of special vehicles based on the analytic hierarchy process;

[0076] S4. The quality performance index prediction results obtained from the time series method and the quality performance prediction results obtained from the analytic hierarchy process are integrated and evaluated.

[0077] Specifically, the purpose of this invention is to establish a quality performance time series and state space model based on the historical evaluation results of special vehicle quality performance, and combine it with the analytic hierarchy process to achieve multi-level quality performance prediction of special vehicles at the single-machine level, subsystem level, and whole-vehicle level.

[0078] This invention first establishes a state-space model and dynamic prediction method for the quality performance indicators of special vehicles, and further presents a multi-level comprehensive quality performance fusion prediction method based on this. The specific steps are as follows:

[0079] Step 1: Establish the state-space equation for predicting the quality performance of special vehicles based on the AR model;

[0080] Step 2: Dynamically recursively estimate the AR model parameters using the Kalman filter method to predict single quality performance indicators;

[0081] Step 3: Predict and assess the overall quality performance and errors of special vehicles using the Analytic Hierarchy Process (AHP).

[0082] Step 4: Conduct a comprehensive quality performance prediction and fusion assessment based on time series and hierarchical analysis.

[0083] As an optional implementation of this invention, the step of establishing the state-space equation for predicting the quality performance of special vehicles based on the AR model includes:

[0084] The evaluation values ​​of vehicle quality performance indicators are obtained, and the evolution law of the indicators is modeled using an m-order autoregressive model AR(m):

[0085]

[0086] Among them, Z t Let β be the index evaluation value at time t, where β = [a1…a2] m b] T H t =[Z t-1 Z t-2 …Z t-m 1],a1,...,a m is the autoregressive coefficient, b is the intercept, ε is Gaussian white noise with a mean of 0 and a standard deviation of σ;

[0087] Establish a historical data matrix of quality performance indicators and Calculate the maximum likelihood estimates of β and σ:

[0088]

[0089] Determine the value of m;

[0090] Build Covariance matrix:

[0091]

[0092] Constructing a state-space model:

[0093]

[0094] Where, β t =[a 1,t a 2,t …a m,t b t ] T w t-1 For state noise, v t To measure noise, Here is the state noise covariance matrix. To measure the noise variance, where E(·) is the mathematical expectation function.

[0095] Q t-1 and The main diagonal elements are the same.

[0096] In specific implementation, the "establishment of the state-space equation for predicting the quality performance of special vehicles based on the AR model" described in step one of this invention is as follows:

[0097] Let Z be a certain quality performance index of the vehicle. t Given the evaluation value of the indicator at time t, it is now necessary to calculate the result based on the historical evaluation results Z1, Z2, ..., Z... t This step aims to establish a state-space equation for predicting the quality performance of special vehicles by forecasting future indicator values.

[0098] First, an m-order autoregressive model AR(m) is used to model the evolution of the index, i.e.,

[0099]

[0100] In the formula: a1,...,a m Let be the autoregressive coefficient, b be the intercept, and ε be Gaussian white noise with a mean of 0 and a standard deviation of σ. For ease of representation, let β = [a1…a2]. m b] T H t =[Z t-1 Z t-2 …Z t-m 1], then equation (1) can be simplified to:

[0101]

[0102] Let a set of historical evaluation results data for this indicator be (H) i Z i Let the data matrix be ), i = 1, 2, ..., n.

[0103]

[0104] Then the likelihood function can be constructed as follows:

[0105]

[0106] The log-likelihood function is

[0107]

[0108] According to the maximum likelihood estimation method

[0109]

[0110] The maximum likelihood estimates of β and σ can be obtained as follows:

[0111]

[0112] Furthermore, the model order m of the autoregressive model AR(m) can be given according to the AIC information criterion, that is, the value of m should minimize the result of the following formula.

[0113]

[0114] The Fisher information matrix is ​​used to measure the amount of information about model parameters contained in historical data on quality performance. The Fisher information matrix for β is...

[0115]

[0116] The greater the amount of information, the smaller the parameter estimation error, which leads to further parameter estimation. The covariance matrix is

[0117]

[0118] Based on the aforementioned AR model, considering factors such as changes in user usage intensity, changes in product performance evolution patterns, and the impact of unforeseen events, the evolution pattern of vehicle quality performance may gradually change, and the corresponding AR model parameters will also exhibit dynamic variation characteristics. To better adapt to and track these changes, a1,...,a m Let b be considered as a latent state variable that changes with time, that is, the model parameters of AR(m) are considered to be related to time t, denoted as...

[0119] β t =[a 1,t a 2,t …a m,t b t ] T (11)

[0120] And construct the state-space equation as follows

[0121]

[0122] In the formula: w t-1 For state noise, v t To measure noise, Q t-1 State noise covariance matrix, R t To measure the noise variance. Among them, Q t-1 and The main diagonal elements are the same, and the other elements are 0.

[0123] As an optional implementation of this invention, the step of dynamically recursively estimating the AR model parameters according to the Kalman filtering method and predicting the single-mass performance index using the time series method includes:

[0124] Using the state-space model, the optimal recursive estimate of parameter β is given by the Kalman filter method;

[0125] After obtaining the latest state estimate of parameter β through recursion, the corresponding prediction variance is evaluated according to the Monte Carlo method.

[0126] As an optional implementation of this invention, the step of using the state-space model and employing the Kalman filter method to provide the optimal recursive estimate of parameter β includes:

[0127] Computational state prediction for: The state prediction error covariance matrix is ​​P t / (t-1) =P t-1 +Q t-1 ; where the initial state value For the maximum likelihood estimation results The initial state covariance P0 is taken as Q. t-1 ;

[0128] Calculate the state estimate at time t for: The state estimation error covariance matrix is ​​P t =(IK t H t )P t / t-1 Among them, K t K is the filter gain matrix. t =P t / t-1 H t T (H t P t / t-1 H t T +R t ) -1 ;

[0129] Let t = t + 1, and repeat the steps of calculating the state prediction, the state prediction error covariance matrix, the state estimation at time t, and the state estimation error covariance matrix to perform the optimal recursive estimation of parameter β.

[0130] As an optional implementation of this invention, the step of performing a corresponding prediction variance evaluation according to the Monte Carlo method after recursively obtaining the latest state estimate of parameter β includes:

[0131] From normal distribution Generate k random numbers, denoted as ε1, ε2, ..., ε3. k ;

[0132] Calculate the prediction results

[0133]

[0134] Repeat the above steps M times, and calculate the results in each of the M times. The sample variance, denoted as Var t+1 Var t+2 ,…,Var t+k For the prediction results The variance;

[0135] Calculate the prediction interval for quality performance indicators with probability p:

[0136]

[0137] In specific implementation, the method described in step two of this invention, "dynamically and recursively estimating AR model parameters based on the Kalman filtering method to achieve prediction of single quality performance indicators," is as follows:

[0138] First, based on the state-space model in step one, the optimal recursive estimate of parameter β is given using the Kalman filter method, and the calculation formula is as follows:

[0139] ① Prediction Phase: State Prediction for

[0140]

[0141] The state prediction error covariance matrix is

[0142] P t / (t-1) =P t-1 +Q t-1 (14)

[0143] Among them, the initial state value Take the estimated result from step one. The initial state covariance P0 is taken as Q. t-1 .

[0144] ② Update phase: State estimation at time t for

[0145]

[0146] The state estimation error covariance matrix is

[0147] P t =(IK t H t )Pt / t-1 (16)

[0148] In the formula, K t The filter gain matrix is ​​calculated using the following formula:

[0149] K t =P t / t-1 H t T (H t P t / t-1 H t T +R t ) -1 (17)

[0150] ③ Let t = t + 1, and repeat steps ①-② above to achieve the optimal recursive estimate of parameter β.

[0151] After recursively obtaining the latest state estimate of parameter β, future quality performance predictions can be carried out, as follows. Based on the historical evaluation results Z1, Z2, ..., Z of the special vehicle quality performance indicators... t The predicted results of the quality performance index at times t+1 to t+k are as follows:

[0152]

[0153] The prediction variance was evaluated using the Monte Carlo method described below:

[0154] (1) From the normal distribution Generate k random numbers, denoted as ε1, ε2, ..., ε3. k ;

[0155] (2) Calculate according to the following formula

[0156]

[0157] (3) Repeat the above steps M times (usually 10⁵), and count the results of each of the M times. The sample variance, denoted as Var t+1 Var t+2 ,…,Var t+k This is the prediction result. The variance.

[0158] Furthermore, based on the above results, the prediction interval for the quality performance index with probability p can be given as follows:

[0159]

[0160] In predicting the quality and performance of special vehicles, the first step should be based on performance indicators Z1, Z2, ..., Z... tUsing a portion of the data (usually requiring more than 10 data points), an autoregressive model is constructed and model parameters are estimated according to the method in step one. Then, the Kalman filtering method in this step and other data are used to perform state-space recursive estimation of AR model parameters and to achieve future prediction of the quality performance index Z.

[0161] As an optional embodiment of the present invention, the prediction and error assessment of the comprehensive quality performance of special vehicles based on the analytic hierarchy process includes:

[0162] Based on the predicted quality performance indicators of each individual machine at times t+1 to t+k and the corresponding prediction variance The weights of each component unit in the subsystem quality performance evaluation are obtained using the analytic hierarchy process (AHP):

[0163] Calculate the predicted values ​​and variances of the quality performance indicators of subsystem A at times t+1 to t+k:

[0164]

[0165] Calculate the upper and lower limits of the prediction for a quality performance index with probability p:

[0166]

[0167] In specific implementation, the "prediction and error assessment of the comprehensive quality performance of special vehicles based on the analytic hierarchy process" described in step three of this invention is performed as follows:

[0168] See the hierarchical diagram of the quality performance evaluation structure for special vehicles. Figure 3 The quality performance indicators of a subsystem are obtained by weighted analytic hierarchy process (AHP) of the quality performance evaluation results of its constituent individual units, and the overall vehicle quality performance indicators are obtained by weighted analytic hierarchy process (AHP) of the quality performance evaluation results of its constituent subsystems. For the aforementioned comprehensive indicators, the same approach can be used to predict the future based on the quality performance prediction results of the next level. This invention will be illustrated below using subsystem A as an example.

[0169] First, the predicted quality performance indicators of each individual machine at times t+1 to t+k are obtained using the methods in steps one and two. and the corresponding prediction variance

[0170] Then, using the analytic hierarchy process (or other methods), the weight values ​​of each component unit in the quality performance evaluation of subsystem A are obtained.

[0171]

[0172] Finally, the predicted values ​​and prediction variances of the quality performance indicators of subsystem A at times t+1 to t+k are given by the following formula.

[0173]

[0174] Based on the above results, the upper and lower limits for predicting quality performance indicators with probability p can also be given, respectively.

[0175]

[0176] This step uses subsystem A as an example to demonstrate the hierarchical analysis prediction method for quality performance evaluation. Similar methods can be used to predict the quality performance of other subsystems and the whole vehicle, thereby enabling multi-level quality performance prediction for special vehicles.

[0177] As an optional implementation of this invention, the fusion evaluation of the comprehensive quality performance prediction based on time series and hierarchical analysis includes:

[0178] Predicted quality performance index values ​​from time t+1 to t+k obtained using the analytic hierarchy process. The corresponding prediction variance is Var AHP,t+1 Var AHP,t+2 ,…,Var AHP,t+k And the predicted quality performance index values ​​for times t+1 to t+k obtained by the time series method. The corresponding prediction variance is Var AR-KM,t+1 Var AR-KM,t+2 ,…,Var AR-KM,t+k Perform fusion prediction and calculate the fusion prediction value:

[0179]

[0180] And calculate the fusion prediction variance

[0181]

[0182] Where, ω AHP,t+k and ω AR-KM,t+k The weights are respectively obtained using the following formula:

[0183]

[0184] The upper and lower bounds for the prediction of probability p after fusion of comprehensive performance indicators are:

[0185]

[0186] In specific implementation, the "conducting a comprehensive quality performance prediction and fusion evaluation based on time series and hierarchical analysis" described in step four of this invention is carried out as follows:

[0187] For the problem of predicting the overall quality performance at the subsystem or vehicle level, it can be obtained either by weighting the prediction results of the next level of quality performance using the analytic hierarchy process in step three, or directly by using the time series method in steps one and two based on the historical evaluation results of the overall quality performance at the subsystem or vehicle level. This step performs a fusion evaluation of the two methods based on the principle of minimizing error.

[0188] Taking a certain comprehensive performance index (a subsystem or the whole vehicle) as an example, let's record the predicted quality performance index values ​​at times t+1 to t+k obtained by the analytic hierarchy process in step three. The corresponding prediction variance is Var AHP,t+1 Var AHP,t+2 ,…,Var AHP,t+k Additionally, record the predicted quality performance index values ​​for times t+1 to t+k obtained from the time series method in steps one and two. The corresponding prediction variance is Var AR-KM,t+1 Var AR-KM,t+2 ,…,Var AR-KM,t+k The fusion prediction value of this indicator is then...

[0189]

[0190] The fusion prediction variance is obtained by the following formula.

[0191]

[0192] In the formula: ω AHP,t+k and ω AR-KM,t+k The weights of the two methods are obtained from the following formula.

[0193]

[0194] The upper and lower bounds for the prediction of this comprehensive performance index with probability p after fusion are:

[0195]

[0196] The special vehicle quality performance prediction method provided in this invention establishes a method for predicting special vehicle quality performance indicators by combining AR models and Kalman filtering. This method can track the dynamic time-varying characteristics of AR model parameters, enhancing prediction adaptability. This method is suitable for predicting the quality performance indicators of a single special vehicle, as well as for predicting the comprehensive quality performance at the subsystem and vehicle levels. Simultaneously, a special vehicle quality performance indicator prediction method based on hierarchical analysis is established. This method can, based on the quality performance prediction results of a single special vehicle, achieve multi-level comprehensive quality performance prediction at the vehicle subsystem and vehicle levels from the bottom up, and realize error propagation evaluation in the hierarchical analysis process. For the comprehensive quality performance indicators at the subsystem and vehicle levels, a performance indicator fusion prediction method based on time series prediction and hierarchical analysis prediction is established on the above basis. This fusion can reduce the prediction error band of the quality performance indicators.

[0197] Therefore, the special vehicle quality performance prediction method provided in this embodiment of the invention considers the dynamic tracking of model parameters based on the autoregressive model (i.e., AR model), establishes the state space equation for vehicle quality performance prediction, and provides a method for determining relevant parameters. Dynamic prediction of quality performance indicators is achieved based on Kalman filtering. Furthermore, a hierarchical analysis prediction method for comprehensive quality performance indicators is established, and a method for fusing and evaluating the results of time series prediction and hierarchical analysis prediction based on error analysis is provided. This method can reasonably achieve bottom-up prediction propagation and error analysis, effectively fusing the prediction results based on time series and hierarchical analysis, and dynamically predicting vehicle quality performance indicators at the single-machine level, subsystem level, and whole-vehicle level.

[0198] The following section uses a vehicle quality performance prediction problem as an example to further illustrate the method of the present invention.

[0199] The quality performance evaluation results of two subsystems of a certain vehicle—the power supply and distribution system and the chassis powertrain system—are listed in Table 1. The overall vehicle quality performance is obtained by weighted calculation using the following analytic hierarchy process formula.

[0200] Z 整车 =ω 供配电 Z 供配电 +ω 动力 Z 动力 =0.3Z 供配电 +0.7Z 动力 (27)

[0201] The following describes the process for predicting the quality and performance of the power supply and distribution system, chassis power system, and the whole vehicle.

[0202] Table 1. Quality performance evaluation results of a certain vehicle over 36 months.

[0203]

[0204]

[0205] 1. Performance prediction of a vehicle's power supply and distribution system

[0206] First, the AR model parameters are estimated using the method in step one and the quality performance evaluation results from January 2018 to October 2018. According to the AIC information criterion, the order of the autoregressive model is taken as m = 2. Based on equation (7), the maximum likelihood estimation result of the AR(2) model parameters is obtained as follows:

[0207]

[0208] Then, the state-space equations are constructed according to equation (12) as follows:

[0209]

[0210] Where: state noise w t-1 Covariance matrix Q t-1 =diag(0.0644,0.0816,69.437) and measurement noise v t Variance R t =0.0194 is given by the maximum likelihood method. Using the Kalman filter formula from step two and the quality performance evaluation results from November 2018 to June 2020, the AR model parameters are recursively estimated, and the estimation results are shown below. Figure 4 .

[0211] Finally, using the latest AR model parameter recursive estimation results and the method in step two, the power supply and distribution system quality performance for the next six months (July 2020 - December 2020) is predicted. The prediction results are shown below. Figure 5 .

[0212] 2. Performance prediction of a vehicle chassis powertrain

[0213] First, the AR model parameters are estimated using the method in step one and the quality performance evaluation results from January 2018 to October 2018. According to the AIC information criterion, the order of the autoregressive model is taken as m=1. Based on equation (7), the maximum likelihood estimation result of the AR(1) model parameters is obtained as follows:

[0214]

[0215] Then, the state-space equations are constructed according to equation (12) as follows:

[0216]

[0217] Where: state noise w t-1 Covariance matrix Qt-1 =diag(0.012,118.2433) and measurement noise v t Variance R t =0.0307 is given by the maximum likelihood method. The state recursive estimation of AR model parameters is performed using the Kalman filter formula from step two and the quality performance evaluation results from November 2018 to June 2020.

[0218] Finally, using the latest AR model parameter state recursive estimation results and the method in step two, the quality performance of the chassis powertrain system for the next six months (i.e., July 2020 - December 2020) is predicted. The prediction results are shown in [the table below]. Figure 6 .

[0219] 3. Vehicle quality performance prediction

[0220] For the problem of predicting overall vehicle quality performance, on the one hand, similar to a subsystem, the overall vehicle quality performance assessment results of the first 30 months are used, along with the methods in steps one and two, to perform time series analysis and prediction of the quality performance for the last 6 months; on the other hand, based on the hierarchical analysis method in step three, and... Figure 5 and Figure 6 The subsystem prediction results can also be used to predict the overall vehicle quality performance; finally, according to step four, the prediction results obtained from the two methods are fused. The prediction results are shown below. Figure 7 Therefore, it can be seen that the prediction error band is the smallest after fusion evaluation.

[0221] Figure 8 This diagram illustrates the structure of a special vehicle quality performance prediction device provided in an embodiment of the present invention. This device applies the aforementioned method. The following is only a brief description of the structure of the special vehicle quality performance prediction device; for other matters not covered herein, please refer to the relevant descriptions in the aforementioned special vehicle quality performance prediction method. Figure 8 The special vehicle quality performance prediction device provided in this embodiment of the invention includes:

[0222] A module is established to build the state-space equations for predicting the quality and performance of special vehicles based on the AR model.

[0223] The first prediction module is used to dynamically recursively estimate the parameters of the AR model based on the Kalman filter method and predict single quality performance indicators using the time series method.

[0224] The second prediction module is used to predict the comprehensive quality performance and error assessment of special vehicles based on the analytic hierarchy process.

[0225] The integrated prediction module is used to integrate and evaluate the quality performance index prediction results obtained from the time series method and the quality performance prediction results obtained from the analytic hierarchy process.

[0226] As an optional implementation of this invention, the establishment module establishes the state-space equation for predicting the quality performance of special vehicles based on the AR model in the following manner:

[0227] The evaluation values ​​of vehicle quality performance indicators are obtained, and the evolution law of the indicators is modeled using an m-order autoregressive model AR(m):

[0228]

[0229] Among them, Z t Let β be the index evaluation value at time t, where β = [a1…a2] m b] T H t =[Z t-1 Z t-2 …Z t-m 1],a1,...,a m is the autoregressive coefficient, b is the intercept, ε is Gaussian white noise with a mean of 0 and a standard deviation of σ;

[0230] Establish a historical data matrix of quality performance indicators and Calculate the maximum likelihood estimates of β and σ:

[0231]

[0232] Determine the value of m;

[0233] Build Covariance matrix:

[0234]

[0235] Constructing a state-space model:

[0236]

[0237] Where, β t =[a 1,t a 2,t …a m,t b t ] T w t-1 For state noise, v t To measure noise, Here is the state noise covariance matrix. To measure the noise variance, where E(·) is the mathematical expectation function. Q t-1 and The main diagonal elements are the same.

[0238] As an optional implementation of this invention, the first prediction module dynamically recursively estimates the AR model parameters using the Kalman filter method and predicts single-mass performance indicators using the time series method in the following manner:

[0239] Using the state-space model, the optimal recursive estimate of parameter β is given by the Kalman filter method;

[0240] After obtaining the latest state estimate of parameter β through recursion, the corresponding prediction variance is evaluated according to the Monte Carlo method.

[0241] As an optional implementation of this invention, the first prediction module utilizes the state-space model and employs the Kalman filter method to provide the optimal recursive estimate of parameter β in the following manner:

[0242] Computational state prediction for: The state prediction error covariance matrix is ​​P t / (t-1) =P t-1 +Q t-1 ; where the initial state value For the maximum likelihood estimation results The initial state covariance P0 is taken as Q. t-1 ;

[0243] Calculate the state estimate at time t for: The state estimation error covariance matrix is ​​P t =(IK t H t )P t / t-1 Among them, K t K is the filter gain matrix. t =P t / t-1 H t T (H t P t / t-1 H t T +R t ) -1 ;

[0244] Let t = t + 1, and repeat the steps of calculating the state prediction, the state prediction error covariance matrix, the state estimation at time t, and the state estimation error covariance matrix to perform the optimal recursive estimation of parameter β.

[0245] As an optional implementation of this invention, the first prediction module, after recursively obtaining the latest state estimate of parameter β, performs a corresponding prediction variance evaluation using the Monte Carlo method as follows:

[0246] From normal distribution Generate k random numbers, denoted as ε1, ε2, ..., ε3. k ;

[0247] Calculate the prediction results

[0248]

[0249] Repeat the above steps M times, and calculate the results in each of the M times. The sample variance, denoted as Var t+1 Var t+2 ,…,Var t+k For the prediction results The variance;

[0250] Calculate the prediction interval for quality performance indicators with probability p:

[0251]

[0252] As an optional embodiment of the present invention, the second prediction module performs comprehensive quality performance prediction and error assessment of special vehicles based on the analytic hierarchy process in the following manner:

[0253] Based on the predicted quality performance indicators of each individual machine at times t+1 to t+k and the corresponding prediction variance The weights of each component unit in the subsystem quality performance evaluation are obtained using the analytic hierarchy process (AHP):

[0254] Calculate the predicted values ​​and variances of the quality performance indicators of subsystem A at times t+1 to t+k:

[0255]

[0256] Calculate the upper and lower limits of the prediction for a quality performance index with probability p:

[0257]

[0258] As an optional implementation of this invention, the comprehensive prediction module performs a fusion evaluation based on comprehensive quality performance prediction using time series and hierarchical analysis in the following manner:

[0259] Predicted quality performance index values ​​from time t+1 to t+k obtained using the analytic hierarchy process. The corresponding prediction variance is Var AHP,t+1 Var AHP,t+2 ,…,Var AHP,t+k And the predicted quality performance index values ​​for times t+1 to t+k obtained by the time series method. The corresponding prediction variance is Var AR-KM,t+1 Var AR-KM,t+2 ,…,Var AR-KM,t+k Perform fusion prediction and calculate the fusion prediction value:

[0260]

[0261] And calculate the fusion prediction variance

[0262]

[0263] Where, ω AHP,t+k and ω AR-KM,t+k The weights are respectively obtained using the following formula:

[0264]

[0265] The upper and lower bounds for the prediction of probability p after fusion of comprehensive performance indicators are:

[0266]

[0267] Therefore, the special vehicle quality performance prediction device provided in this embodiment of the invention considers the dynamic tracking of model parameters based on the autoregressive model (i.e., AR model), establishes the state space equation for vehicle quality performance prediction, and provides a method for determining relevant parameters. It achieves dynamic prediction of quality performance indicators based on Kalman filtering. Furthermore, it establishes a hierarchical analysis prediction method for comprehensive quality performance indicators and provides a method for fusing and evaluating the results of time series prediction and hierarchical analysis prediction based on error analysis. This method can reasonably achieve bottom-up prediction propagation and error analysis, effectively fusing the prediction results based on time series and hierarchical analysis, and dynamically predicting vehicle quality performance indicators at the single-machine level, subsystem level, and whole-vehicle level.

[0268] Figure 9 A schematic diagram of an electronic device provided in an embodiment of the present invention, see below. Figure 9 The electronic device provided in this embodiment of the invention includes: a processor and a memory;

[0269] The memory is used to store computer programs;

[0270] The processor is configured to execute the special vehicle quality performance prediction method as described above by invoking the computer program.

[0271] Electronic device 9 can be a desktop computer, laptop, handheld computer, cloud server, or other electronic device. Electronic device 9 may include, but is not limited to, processor 901 and memory 902. Those skilled in the art will understand that... Figure 9This is merely an example of electronic device 9 and does not constitute a limitation on electronic device 9. It may include more or fewer components than shown, or different components.

[0272] The processor 901 can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. In this embodiment of the invention, the processor can be the controller described in the above embodiments.

[0273] The memory 902 can be an internal storage unit of the electronic device 9, such as a hard disk or RAM. The memory 902 can also be an external storage device of the electronic device 9, such as a plug-in hard disk, SmartMedia Card (SMC), Secure Digital (SD) card, or Flash Card. The memory 902 can also include both internal and external storage units of the electronic device 9. The memory 902 is used to store the computer program 903 and other programs and data required by the electronic device 9.

[0274] Therefore, the electronic device provided in this embodiment of the invention, based on the autoregressive model (i.e., AR model), considers the dynamic tracking of model parameters, establishes the state-space equation for predicting vehicle quality performance, and provides a method for determining relevant parameters. It achieves dynamic prediction of quality performance indicators based on Kalman filtering. Furthermore, it establishes a hierarchical analysis prediction method for comprehensive quality performance indicators and provides a method for fusing and evaluating the results of time series prediction and hierarchical analysis prediction based on error analysis. This method can reasonably achieve bottom-up prediction propagation and error analysis, effectively fusing the prediction results based on time series and hierarchical analysis, and dynamically predicting vehicle quality performance indicators at the single-machine level, subsystem level, and whole-vehicle level.

[0275] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the special vehicle quality performance prediction method as described above.

[0276] Therefore, the computer-readable storage medium provided in this embodiment of the invention, based on the autoregressive model (i.e., AR model), considers the dynamic tracking of model parameters, establishes the state-space equation for vehicle quality performance prediction, and provides a method for determining relevant parameters. It achieves dynamic prediction of quality performance indicators based on Kalman filtering. Furthermore, it establishes a hierarchical analysis prediction method for comprehensive quality performance indicators and provides a method for fusing and evaluating the results of time series prediction and hierarchical analysis prediction based on error analysis. This method can reasonably achieve bottom-up prediction propagation and error analysis, effectively fusing the prediction results based on time series and hierarchical analysis, and dynamically predicting vehicle quality performance indicators at the single-machine level, subsystem level, and whole-vehicle level.

[0277] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0278] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for predicting the quality performance of special vehicles, characterized in that, include: Establish a state-space equation for predicting the quality and performance of special vehicles based on the AR model; The AR model parameters are dynamically recursively estimated using the Kalman filter method, and single-mass performance indicators are predicted using the time series method. Special vehicle comprehensive quality performance prediction and error assessment based on the analytic hierarchy process; The quality performance prediction results obtained from the time series method and the quality performance prediction results obtained from the analytic hierarchy process are integrated and evaluated.

2. The method according to claim 1, characterized in that, The state-space equation for predicting the quality performance of special vehicles based on the AR model includes: The evaluation values ​​of vehicle quality performance indicators are obtained, and the evolution law of the indicators is modeled using an m-order autoregressive model AR(m): Among them, Z t Let β be the evaluation value of the index at time t, where β = [a1 … a m b] T H t =[Z t-1 Z t-2 … Z t-m 1],a1,...,a m is the autoregressive coefficient, b is the intercept, ε is Gaussian white noise with a mean of 0 and a standard deviation of σ; Establish a historical data matrix of quality performance indicators and Calculate the maximum likelihood estimates of β and σ: Determine the value of m; Build Covariance matrix: Constructing a state-space model: Where, β t =[a 1,t a 2,t …a m,t b t ] T w t-1 For state noise, v t To measure noise, Here is the state noise covariance matrix. To measure the noise variance, where E(·) is the mathematical expectation function. Q t-1 and The main diagonal elements are the same.

3. The method according to claim 2, characterized in that, The step of dynamically recursively estimating AR model parameters using the Kalman filter method and predicting single-quality performance indicators using time series analysis includes: Using the state-space model, the optimal recursive estimate of parameter β is given by the Kalman filter method; After obtaining the latest state estimate of parameter β through recursion, the corresponding prediction variance is evaluated according to the Monte Carlo method.

4. The method according to claim 3, characterized in that, The step of using the state-space model and employing the Kalman filter method to provide the optimal recursive estimate of parameter β includes: Computational state prediction for: The state prediction error covariance matrix is ​​P t / (t-1) =P t-1 +Q t-1 ; where the initial state value For the maximum likelihood estimation results The initial state covariance P0 is taken as Q. t-1 ; Calculate the state estimate at time t for: The state estimation error covariance matrix is ​​P t =(IK t H t )P t / t-1 Among them, K t K is the filter gain matrix. t =P t / t-1 H t T (H t P t / t-1 H t T +R t ) -1 ; Let t = t + 1, and repeat the steps of calculating the state prediction, the state prediction error covariance matrix, the state estimation at time t, and the state estimation error covariance matrix to perform the optimal recursive estimation of parameter β.

5. The method according to claim 4, characterized in that, After obtaining the latest state estimate of parameter β through recursion, the corresponding prediction variance evaluation according to the Monte Carlo method includes: From normal distribution Generate k random numbers, denoted as ε1, ε2, ..., ε3. k ; Calculate the prediction results Repeat the above steps M times, and calculate the results in each of the M times. The sample variance, denoted as Var t+1 Var t+2 ,…,Var t+k For the prediction results The variance; Calculate the prediction interval for quality performance indicators with probability p:

6. The method according to claim 5, characterized in that, The method of predicting and evaluating the overall quality performance and errors of special vehicles based on the analytic hierarchy process includes: Based on the predicted quality performance indicators of each individual machine at times t+1 to t+k and the corresponding prediction variance The weights of each component unit in the subsystem quality performance evaluation are obtained using the analytic hierarchy process (AHP): Calculate the predicted values ​​and variances of the quality performance indicators of subsystem A at times t+1 to t+k: Calculate the upper and lower limits of the prediction for a quality performance index with probability p:

7. The method according to claim 6, characterized in that, The integrated quality performance prediction based on time series and hierarchical analysis is used for fusion evaluation, which includes: Predicted quality performance index values ​​from time t+1 to t+k obtained using the analytic hierarchy process. The corresponding prediction variance is Var AHP,t+1 Var AHP,t+2 ,…,Var AHP,t+k And the predicted quality performance index values ​​for times t+1 to t+k obtained by the time series method. The corresponding prediction variance is Var AR-KM,t+1 Var AR-KM,t+2 ,…,Var AR-KM,t+k Perform fusion prediction and calculate the fusion prediction value: And calculate the fusion prediction variance Where, ω AHP,t+k and ω AR-KM,t+k The weights are respectively obtained using the following formula: The upper and lower bounds for the prediction of probability p after fusion of comprehensive performance indicators are:

8. A special vehicle quality performance prediction device, characterized in that, include: A module is established to build the state-space equations for predicting the quality and performance of special vehicles based on the AR model. The first prediction module is used to dynamically recursively estimate the parameters of the AR model based on the Kalman filter method and predict single quality performance indicators using the time series method. The second prediction module is used to predict the comprehensive quality performance and error assessment of special vehicles based on the analytic hierarchy process. The integrated prediction module is used to integrate and evaluate the quality performance index prediction results obtained from the time series method and the quality performance prediction results obtained from the analytic hierarchy process.

9. An electronic device, characterized in that, include: Processor, memory; The memory is used to store computer programs; The processor is configured to execute the special vehicle quality performance prediction method as described in any one of claims 1 to 7 by invoking the computer program.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the special vehicle quality performance prediction method according to any one of claims 1 to 7.

Images