Performance prediction method of electromechanical equipment based on probabilistic deep neural network model

Abstract

This invention discloses a method for predicting the performance of electromechanical equipment based on a probabilistic deep neural network model. The method comprises: constructing a probabilistic deep neural network prediction model; obtaining a performance parameter vector of the electromechanical equipment and training the prediction model using the performance parameter vector; and predicting the performance data and reliability of the electromechanical equipment based on the trained prediction model. This method can simultaneously predict the evolutionary trends of electromechanical equipment performance and reliability, addressing the shortcomings of traditional methods in reliability prediction and risk control, and enabling accurate and rapid prediction of the health status of electromechanical equipment.

Description

Technical Field

[0001] The present invention belongs to the field of health management and fault prediction, and in particular relates to a method for predicting the performance of electromechanical equipment based on a probabilistic deep neural network model. Background Art

[0002] Performance prediction for electromechanical equipment is a crucial component of prognostics and health management (PHM). The performance of electromechanical equipment directly impacts the effective range and accuracy of remaining life prediction. Traditional deep neural network prediction methods can only learn and predict the performance parameters of a specific electromechanical system and are unable to calculate the reliability of electromechanical equipment, making it difficult to meet the risk control requirements for predictive maintenance. Summary of the Invention

[0003] The purpose of the present invention is to provide a method for predicting the performance of electromechanical equipment based on a probabilistic deep neural network model to solve the problems existing in the above-mentioned prior art.

[0004] To achieve the above objectives, the present invention provides a method for predicting the performance of electromechanical equipment based on a probabilistic deep neural network model, comprising:

[0005] Build a probabilistic deep neural network prediction model;

[0006] Obtaining a performance parameter vector of the electromechanical device, and training the prediction model using the performance parameter vector;

[0007] Based on the trained prediction model, the performance data and reliability of electromechanical equipment are predicted.

[0008] Preferably, the prediction model includes: an input module, a calculation module, an integrated decision module and an output module, wherein the input module, the calculation module, the integrated decision module and the output module are connected in series.

[0009] Preferably, the process of constructing the operation module includes:

[0010] Several recurrent neural networks are connected in parallel to construct the operation module.

[0011] Preferably, the process of constructing the integrated decision module includes:

[0012] The prediction results of the recurrent neural network are statistically analyzed based on the Monte Carlo algorithm, and the statistical results constitute the integrated decision module.

[0013] Preferably, the process of obtaining the performance parameter vector of the electromechanical device includes:

[0014] Install vibration sensors at different locations on electromechanical equipment;

[0015] A performance parameter vector of the electromechanical equipment is obtained through the vibration sensor.

[0016] Preferably, the process of training the prediction model using the performance parameter vector includes:

[0017] Calculating an eigenvector of the performance parameter vector;

[0018] Based on the time unit monitored by the vibration sensor, the eigenvector is matrix constructed to obtain a eigenmatrix;

[0019] The feature matrix is ​​input into the prediction model, and the prediction model is trained until the output error is reduced to an expected value, thereby obtaining a trained prediction model.

[0020] Preferably, the process of predicting the performance data of the electromechanical equipment includes:

[0021] Input the performance data of the electromechanical equipment into the trained prediction model to predict the performance data of the electromechanical equipment;

[0022] Based on the predicted value output by the trained prediction model, the confidence interval of the performance of the electromechanical device under the probability 1-α is calculated, and the performance prediction result of the electromechanical device is obtained through the confidence interval.

[0023] Preferably, the process of predicting the reliability of electromechanical equipment includes:

[0024] Obtain performance data of electromechanical equipment;

[0025] The performance data is input into a trained prediction model to predict the reliability of the electromechanical equipment and obtain a performance failure threshold of the electromechanical equipment. The confidence interval probability boundary of the performance failure threshold is the reliability of the electromechanical equipment.

[0026] The technical effects of the present invention are:

[0027] This method constructs a probabilistic deep neural network prediction model; trains the prediction model using performance parameter vectors of electromechanical equipment; and uses the trained prediction model to predict the performance data and reliability of electromechanical equipment. This method can simultaneously predict the evolutionary trends of electromechanical equipment performance and reliability, addressing the shortcomings of traditional methods in reliability prediction and risk control, and enabling accurate and rapid prediction of the health status of electromechanical equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The accompanying drawings, which constitute part of this application, are intended to provide a further understanding of this application. The exemplary embodiments and descriptions of this application are intended to explain this application and do not constitute an improper limitation on this application. In the accompanying drawings:

[0029] Figure 1 Schematic diagram of a novel probabilistic deep neural network model in an embodiment of the present invention;

[0030] Figure 2 is a flow chart of a method in an embodiment of the present invention;

[0031] Figure 3 is a schematic diagram of stress data in an embodiment of the present invention;

[0032] Figure 4 is a damage conversion histogram in an embodiment of the present invention;

[0033] Figure 5 Schematic diagram of a loss curve in an embodiment of the present invention;

[0034] Figure 6 Schematic diagram of damage prediction results in an embodiment of the present invention;

[0035] Figure 7 Schematic diagram of reliability prediction results in an embodiment of the present invention. DETAILED DESCRIPTION

[0036] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0037] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0038] Example 1

[0039] like Figure 1-2 As shown, this embodiment provides a method for predicting the performance of electromechanical equipment based on a probabilistic deep neural network model, including:

[0040] Build a probabilistic deep neural network prediction model;

[0041] Obtaining a performance parameter vector of the electromechanical device, and training the prediction model using the performance parameter vector;

[0042] Based on the trained prediction model, the performance data and reliability of electromechanical equipment are predicted.

[0043] In some embodiments, the prediction model includes: an input module, a calculation module, an integrated decision module, and an output module, wherein the input module, the calculation module, the integrated decision module, and the output module are connected in series.

[0044] In some embodiments, the process of constructing the operation module includes: connecting a plurality of recurrent neural networks in parallel to construct the operation module.

[0045] In some embodiments, the process of constructing the integrated decision module includes: performing statistics on the prediction results of the recurrent neural network based on the Monte Carlo algorithm, and the statistical results constitute the integrated decision module.

[0046] In some embodiments, the process of obtaining the performance parameter vector of the electromechanical device includes: installing vibration sensors at different positions of the electromechanical device; and obtaining the performance parameter vector of the electromechanical device through the vibration sensors.

[0047] In some embodiments, the process of training the prediction model using the performance parameter vector includes: calculating the eigenvector of the performance parameter vector; constructing a matrix of the eigenvector based on the time unit monitored by the vibration sensor to obtain a feature matrix; inputting the feature matrix into the prediction model, and training the prediction model until the output error is reduced to the expected value, thereby obtaining a trained prediction model.

[0048] In some embodiments, the process of predicting the performance data of the electromechanical equipment includes: inputting the performance data of the electromechanical equipment into a trained prediction model to predict the performance data of the electromechanical equipment; based on the predicted value output by the trained prediction model, calculating the confidence interval of the performance of the electromechanical equipment under probability 1-α, and obtaining the performance prediction result of the electromechanical equipment through the confidence interval.

[0049] In some embodiments, the process of predicting the reliability of electromechanical equipment includes: obtaining performance data of the electromechanical equipment at a certain moment; inputting the performance data into a trained prediction model to predict the reliability of the electromechanical equipment to obtain a performance failure threshold of the electromechanical equipment, and the confidence interval probability boundary of the performance failure threshold is the reliability of the electromechanical equipment at a certain moment.

[0050] In this embodiment, the new probabilistic deep neural network model is as follows Figure 1As shown in Figure 1, the model consists of four main components: the first is the output module, which primarily receives the feature matrix input; the second is the Multi-RNN unit, consisting of the first, second, …, and nth RNNs connected in parallel, which performs predictions based on training data from different electromechanical equipment; the third is the integrated decision module, which uses the Monte Carlo algorithm to perform statistical inference on the prediction results of each RNN; and the fourth is the output module, which outputs the prediction results. These four components of the model are designed in series, ultimately outputting the prediction results.

[0051] The new probabilistic deep neural network model is designed to be composed of four parts: collecting performance data, building feature matrix, training model, and predicting performance data and reliability. Specifically, Figure 2 As shown, the prediction method steps are as follows:

[0052] 1) Performance data collection: The vibration acceleration, stress, voltage, current and other performance parameter vectors x of the electromechanical equipment are collected by sensors such as vibration acceleration, stress, voltage, current, etc. arranged on the electromechanical equipment. For n devices, the performance parameter vectors obtained in each same monitoring time unit are x1, x2, ..., x n Therefore, after monitoring m time units, the performance parameter obtained is X1 = [x 1,1 ,x 2,1 ,…,x n,1 ],X2=[x 1,2 ,x 2,2 ,…,x n,2 ],……,X m =[x 1,m ,x 2,m ,…,x n,m ].

[0053] 2) Construct the feature matrix: For each time unit of n devices, calculate its feature vector θ. Therefore, for m time units, we can obtain a feature matrix A consisting of n×m units.

[0054]

[0055] 3) Training model: Input the characteristic matrix A of performance parameters into the probabilistic deep neural network model and train the Multi-RNN unit of the model. The total loss function L is expressed as

[0056]

[0057] For the above loss function, the model training can be completed by using the supervised learning training method.

[0058] 4) Predict performance data and reliability: Based on the trained model, the performance degradation trend and reliability of electromechanical equipment are analyzed. Finally, the performance degradation trend of electromechanical equipment throughout its life cycle is plotted.

[0059]

[0060] Where, The predicted value for each RNN output.

[0061] At the same time, the confidence interval of the estimated performance under the probability (1-α) It is calculated as follows:

[0062]

[0063]

[0064] In the formula, rank(·) represents the ranking of features.

[0065] For the performance status at a certain time t, determine whether it reaches the performance failure threshold θ m The confidence interval probability boundary β m =(1-α m ), then it is the reliability R(t) of the electromechanical equipment at that moment, that is, R(t) = β m .

[0066] The technical effects of this embodiment are:

[0067] This embodiment constructs a probabilistic deep neural network prediction model; trains the prediction model using performance parameter vectors of electromechanical equipment; and uses the trained prediction model to predict the performance data and reliability of electromechanical equipment. This invention can simultaneously predict the evolutionary trends of electromechanical equipment performance and reliability, addressing the shortcomings of traditional methods in reliability prediction and risk control, and enabling accurate and rapid prediction of the health status of electromechanical equipment.

[0068] This embodiment can accurately predict the current performance evolution trend of electromechanical equipment, and can also estimate the variation range of the performance status within a certain confidence interval, thereby estimating the reliability of the electromechanical equipment.

[0069] Example 2

[0070] This embodiment provides a case study on damage prediction of a rail vehicle running gear beam, as follows:

[0071] 1) Arrange strain gauges and rail vehicle crossbeams, and use data acquisition instruments to obtain stress data during vehicle operation. The test collected stress data from 5 rail vehicles. The collected stress data is as follows: Figure 3The strain gauge resistance used for acquisition is 120 ohms and the sampling frequency is 500 Hz.

[0072] 2) According to stress fatigue theory, the stress is converted into dimensionless structural damage value to obtain the characteristic matrix A. Among them, the damage conversion histogram is as follows: Figure 4 shown.

[0073] 3) Based on the above feature matrix, the prediction model is trained using the Adam algorithm. The loss curve of the training process is as follows: Figure 5 shown.

[0074] 4) Use the trained prediction model to predict the damage and reliability of the beam. The results are as follows: Figure 6 and Figure 7 shown.

[0075] It can be seen from the case that the model and method disclosed in the patent of this invention can simultaneously predict the evolution trend of damage performance and the change trend of reliability.

[0076] The above description is merely a preferred embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A method for predicting the performance of electromechanical equipment based on a probabilistic deep neural network model, characterized in that: The following steps are involved: Build a probabilistic deep neural network prediction model; Obtaining a performance parameter vector of the electromechanical device, and training the prediction model using the performance parameter vector; Predict the performance data and reliability of electromechanical equipment based on the trained prediction model; The process of obtaining the performance parameter vector of electromechanical equipment includes: Install vibration sensors at different locations on electromechanical equipment; Obtaining a performance parameter vector of the electromechanical device through the vibration sensor; The process of training the prediction model using the performance parameter vector includes: Calculating an eigenvector of the performance parameter vector; Based on the time unit monitored by the vibration sensor, the eigenvector is matrix constructed to obtain a eigenmatrix; Inputting the feature matrix into the prediction model, training the prediction model until the output error is reduced to an expected value, thereby obtaining a trained prediction model; The process of predicting the performance data of electromechanical equipment includes: Input the performance data of the electromechanical equipment into the trained prediction model to predict the performance data of the electromechanical equipment; Based on the predicted value output by the trained prediction model, a confidence interval of the performance of the electromechanical device under the probability 1-α is calculated, and the performance prediction result of the electromechanical device is obtained through the confidence interval; The process of predicting the reliability of electromechanical equipment includes: Obtain performance data of electromechanical equipment; The performance data is input into a trained prediction model to predict the reliability of the electromechanical equipment and obtain a performance failure threshold of the electromechanical equipment. The confidence interval probability boundary of the performance failure threshold is the reliability of the electromechanical equipment.

2. The electromechanical equipment performance prediction method based on a probabilistic deep neural network model according to claim 1 is characterized in that: The prediction model includes: an input module, a calculation module, an integrated decision module and an output module, wherein the input module, the calculation module, the integrated decision module and the output module are connected in series.

3. The electromechanical equipment performance prediction method based on a probabilistic deep neural network model according to claim 2 is characterized in that: The process of constructing the operation module includes: Several recurrent neural networks are connected in parallel to construct the operation module.

4. The electromechanical equipment performance prediction method based on a probabilistic deep neural network model according to claim 3 is characterized in that: The process of building the integrated decision module includes: The prediction results of the recurrent neural network are statistically analyzed based on the Monte Carlo algorithm, and the statistical results constitute the integrated decision module.

Images