Method, system and equipment for predicting residual life of equipment component and medium

Abstract

The invention discloses a method, a system, equipment and a medium for predicting the residual life of an equipment component, and particularly relates to the field of component life prediction.The technical key points are as follows: performing two-stage maximum likelihood estimation on historical degradation indexes and historical failure data, and fusing the two-stage maximum likelihood estimation to obtain a full log likelihood function; performing maximum likelihood estimation on the full-log likelihood function to obtain a model parameter estimation value, and updating model parameters of the residual life distribution model by using the model parameter estimation value to obtain an updated residual life distribution model; and updating the multi-stage degradation model of the component based on the predicted value of the degradation state of the component by using Kalman filtering in combination with an EM algorithm to obtain an updated multi-stage degradation model, inputting the current multi-source sensing data into the updated multi-stage degradation model to obtain an estimated value of the degradation state, and calculating the degradation state of the component according to the estimated value of the degradation state. And inputting the degradation state estimation value into the updated residual life distribution model to obtain a residual life prediction result.

Description

Technical Field

[0001] The present invention relates to the field of component life prediction, and in particular to a method, system, device and medium for predicting the remaining life of equipment components. Background Art

[0002] Equipment condition monitoring and remaining life prediction are of great significance to ensure the safe and efficient operation of equipment. At present, there are two problems in the prediction of the remaining life of equipment under variable working conditions: on the one hand, the frequent changes in the working conditions of the equipment during its life cycle will affect the amplitude of the degradation data, making the remaining life prediction of equipment under time-varying working conditions more difficult; on the other hand, it is difficult to fully reflect the degradation process of the equipment by using a single degradation indicator for modeling.

[0003] A Chinese patent with the publication number CN118294138A discloses a rolling bearing life prediction method and system, including a vibration sensor to obtain the original vibration signal of the rolling bearing, the collected original vibration signal is summarized and the root mean square value is calculated, the bearing working information is monitored by the root mean square value of the vibration signal, and the root mean square value is stored in real time; the time domain signal of the rolling bearing is collected when it vibrates, and the obtained time domain signal is sampled by sliding window based on the frequency domain analysis method to obtain the time domain signal of multiple time steps. The present invention collects the vibration signal measured by the sensor and pre-processes it to obtain the effective value, combines the frequency domain analysis and filtering algorithm, and realizes comprehensive and accurate monitoring and life prediction of the rolling bearing. However, in this technical solution, it only relies on the root mean square value of the vibration signal and the frequency domain coherence analysis, and does not fuse multi-source data, resulting in relatively limited degradation feature characterization.

[0004] Another Chinese patent with the publication number CN11967640A discloses a method for predicting the remaining life of a tool taking into account tool wear and surface roughness, including the following steps: Step 1, establishing a single degradation index tool remaining life prediction model; Step 2, measuring the flank wear and workpiece surface roughness at the same time, and performing a first-stage prior estimation of the two-stage maximum likelihood estimation for the relevant parameters used to describe the differences between individual tools and the commonalities of similar tools in the tool flank wear degradation model or the workpiece surface roughness degradation model in step 1; Step 3, establishing a multi-degradation index tool remaining life prediction model based on the parameter values ​​obtained by the prior estimation in step 2; Step 4, importing the parameters estimated in steps 2 and 3 into the multi-degradation index tool remaining life prediction model to obtain the tool remaining life prediction result. This solves the problem of a single indicator for evaluating the remaining life of a tool in the prior art, resulting in inaccurate results. Although this technical solution solves the problem of a single indicator, its degradation model cannot be dynamically updated according to the indicator parameters.

[0005] Therefore, the present invention aims to provide a method, system, device and medium for predicting the remaining life of equipment components to solve the above-mentioned related problems. Summary of the invention

[0006] The technical problem to be solved by the present invention is that the evaluation index of the existing remaining life prediction method is single, and the degradation model cannot be dynamically updated according to the index parameters, which easily leads to related problems of inaccurate prediction results. The purpose is to provide a method, system, device and medium for predicting the remaining life of equipment parts. By fusing multi-source sensor data signals such as vibration, current, cutting force, etc., and extracting features in the time domain, frequency domain, and time-frequency domain, a multi-source mapping relationship function is constructed to comprehensively capture nonlinear degradation characteristics, thereby solving the problem of single prediction evaluation index; by dynamically adjusting the multi-source mapping relationship function, the predicted value of the multi-source mapping relationship function is closer to the actual degradation state. , which improves the accuracy of remaining life prediction and can further reduce the prediction error, while ensuring the stability and adaptability of the multi-source mapping relationship function in long-term operation; through the historical degradation index and the ischange function, the stage boundary value can be adaptively divided, avoiding the subjectivity of manually setting the threshold value to improve the accuracy of the degradation process description; at the same time, by combining the Kalman filter and the EM algorithm, the drift coefficient and diffusion coefficient of each stage can be dynamically updated, which solves the problems of prediction deviation caused by fixed parameters and the inability of the offline model to adapt to real-time degradation, and also ensures the continuity of the multi-stage degradation model when switching between multiple stages.

[0007] The present invention is achieved through the following technical solutions: A method for predicting the remaining life of equipment components, the method comprising: Obtain the time domain features, frequency domain features and time-frequency domain features in the historical multi-source sensor data, construct the historical degradation index, and establish a multi-source mapping relationship function that characterizes the historical multi-source sensor data and the degradation state of the component based on the time domain features, frequency domain features and time-frequency domain features of the historical multi-source sensor data; The component degradation cutoff value is calculated by using the ischange function combined with the historical degradation index, and multiple component degradation stages are obtained. Then, based on the component degradation cutoff value and the single-stage degradation function, a component multi-stage degradation model is constructed. Based on the chi-square Markov property of the component Wiener degradation process, a remaining life distribution model of the component degradation stage is constructed. Perform two-stage maximum likelihood estimation on historical degradation indicators and historical failure data, and integrate the two-stage maximum likelihood estimation to obtain the full log-likelihood function, then perform maximum likelihood estimation on the full log-likelihood function to obtain model parameter estimates, and use the model parameter estimates to update the model parameters of the remaining life distribution model to obtain the updated remaining life distribution model; The component degradation state prediction value corresponding to the current multi-source sensor data is obtained through the multi-source mapping relationship function, and the component multi-stage degradation model is updated based on the component degradation state prediction value by using Kalman filtering combined with the EM algorithm to obtain an updated multi-stage degradation model, and the current multi-source sensor data is input into the updated multi-stage degradation model to obtain the degradation state estimation value, and the degradation state estimation value is input into the updated remaining life distribution model to obtain the remaining life prediction result.

[0008] Furthermore, the time domain features, frequency domain features and time-frequency domain features in the historical multi-source sensor data are obtained to construct the historical degradation index, and according to the time domain features, frequency domain features and time-frequency domain features of the historical multi-source sensor data, a multi-source mapping relationship function is established to characterize the historical multi-source sensor data and the degradation state of the component, wherein the multi-source sensor data includes vibration data, current data and cutting force data, specifically: The time domain characteristics, frequency domain characteristics and time-frequency domain characteristics of the historical vibration data, the historical current data and the historical cutting force data are respectively obtained; Calculate the Pearson correlation coefficients between multiple time domain features, frequency domain features, time-frequency domain features and component degradation states, and select features with high correlation as historical degradation indicators; The historical degradation indicators are input into the feature fusion network of the attention mechanism to predict the component degradation status, and a multi-source mapping relationship function is constructed to characterize the historical multi-source sensor data and the component degradation status.

[0009] Furthermore, the method also includes: based on the observation function, constructing a state relationship function between the degradation state mapping result of the multi-source mapping relationship function and the actual degradation state, and using the state relationship function to optimize and update the function parameters of the multi-source mapping relationship function to obtain the multi-source mapping relationship function after parameter optimization and update.

[0010] Furthermore, based on the Chi-square Markov property of the component Wiener degradation process, a remaining life distribution model of the component degradation stage is constructed, specifically: Define the time from the initial degradation moment to the first time the degradation state exceeds the The time of the degradation stage threshold is , according to the chi-square Markov property of the component Wiener degradation process, It is expressed as: ,in, , ,in, represents the p-th stage life, represents the degradation value of the pth stage, It represents the degradation threshold value of the component in the pth degradation stage; The inverse Gaussian distribution function is used to construct the remaining life distribution model of the component degradation stage, specifically: ,in, represents the degradation model parameters of the pth stage, represents the drift function, represents the integral operator, represents the p-th stage life, Indicates p Diffusion parameters for the stage degradation model.

[0011] Furthermore, the component degradation cutoff value is calculated by combining the ischange function with the historical degradation index, and multiple component degradation stages are obtained. Then, a component multi-stage degradation model is constructed based on the component degradation cutoff value and the single-stage degradation function, specifically: The ischange function is used in combination with historical degradation indicators to find multiple component degradation cutoff values ​​and obtain multiple component degradation stages; Based on the single-stage degradation function, a multi-stage degradation model is constructed using multiple component degradation cutoff values, specifically: ,in, Indicates a degraded state. represents the first stage degradation drift function, represents the integral operator, represents the first stage degradation model parameters, represents the diffusion parameter of the first stage degradation model, represents the standard Brownian motion, Represents a representative function, represents the degradation threshold of the component in the first degradation stage, represents the second stage degradation drift function, represents the second stage degradation model parameters, represents the diffusion parameter of the second stage degradation model, represents the standard Brownian motion, Represents a representative function, represents the degradation threshold of the component in the second degradation stage, represents the third stage degradation drift function, represents the third stage degradation model parameters, represents the diffusion parameter of the third stage degradation model, represents the standard Brownian motion, Represents a representative function.

[0012] Furthermore, a two-stage maximum likelihood estimation is performed on the historical degradation index and the historical failure data, and the two-stage maximum likelihood estimation is integrated to obtain the full logarithmic likelihood function, and then the full logarithmic likelihood function is subjected to maximum likelihood estimation to obtain the model parameter estimates. The model parameter estimates are used to update the model parameters of the remaining life distribution model to obtain the updated remaining life distribution model, which is specifically: Based on The first log-likelihood function of the first model parameter and the second model parameter of the remaining life distribution model is constructed based on the historical degradation index of the component degradation stage, and the maximum likelihood estimation method is used to perform maximum likelihood estimation to obtain the estimated value of the first model parameter; Based on The historical failure data of the components in the degradation stage are used to construct the second log-likelihood function of the third model parameter and the fourth model parameter of the remaining life distribution model, and the first log-likelihood function and the second log-likelihood function are fused to obtain the full log-likelihood function, and the full log-likelihood function is maximum-likelihood estimated to obtain the estimated value of the second model parameter; The model parameters of the remaining life distribution model are updated using the first model parameter estimate value and the second model parameter estimate value to obtain an updated remaining life distribution model.

[0013] The present invention also provides a system for predicting the remaining life of equipment components, which is used in any one of the above-mentioned methods for predicting the remaining life of equipment components, and the system comprises: The first module is used to obtain the time domain features, frequency domain features and time-frequency domain features in the historical multi-source sensor data, construct the historical degradation index, and establish a multi-source mapping relationship function that characterizes the historical multi-source sensor data and the degradation state of the component based on the time domain features, frequency domain features and time-frequency domain features of the historical multi-source sensor data; The second module is used to calculate the component degradation cutoff value by using the ischange function combined with the historical degradation index, and obtain multiple component degradation stages. Then, based on the component degradation cutoff value and the single-stage degradation function, a component multi-stage degradation model is constructed, and based on the chi-square Markov property of the component Wiener degradation process, a remaining life distribution model of the component degradation stage is constructed; The third module is used to perform two-stage maximum likelihood estimation on historical degradation indicators and historical failure data, and fuse the two-stage maximum likelihood estimation to obtain the full logarithmic likelihood function, and then perform maximum likelihood estimation on the full logarithmic likelihood function to obtain model parameter estimates, and use the model parameter estimates to update the model parameters of the remaining life distribution model to obtain the updated remaining life distribution model; The fourth module is used to obtain the component degradation state prediction value corresponding to the current multi-source sensor data through the multi-source mapping relationship function, and use Kalman filtering combined with the EM algorithm to update the component multi-stage degradation model based on the component degradation state prediction value to obtain an updated multi-stage degradation model, input the current multi-source sensor data into the updated multi-stage degradation model to obtain the degradation state estimation value, and input the degradation state estimation value into the updated remaining life distribution model to obtain the remaining life prediction result.

[0014] The present invention also provides a computer device, comprising a system memory and a processor, wherein the system memory stores a computer program, and the processor implements the steps of any one of the above-mentioned methods when executing the computer program.

[0015] The present invention also provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of any one of the methods described above are implemented.

[0016] The present invention also provides a computer program product comprising instructions, and when the instructions are executed by a computer device cluster, the computer device cluster executes any of the above methods.

[0017] Compared with the prior art, the present invention has the following advantages and beneficial effects: In the present invention, by fusing multi-source sensor data signals such as vibration, current, cutting force, etc., and extracting features in the time domain, frequency domain, and time-frequency domain, a multi-source mapping relationship function is constructed to comprehensively capture nonlinear degradation characteristics, thereby solving the problem of a single prediction evaluation index; by dynamically adjusting the multi-source mapping relationship function, the predicted value of the multi-source mapping relationship function is made closer to the actual degradation state, thereby improving the accuracy of the remaining life prediction, further reducing the prediction error, and ensuring the stability and adaptability of the multi-source mapping relationship function in long-term operation; by using historical degradation indicators and the ischange function, the stage boundary values ​​can be adaptively divided, thereby avoiding the subjectivity of manually setting thresholds, thereby improving the accuracy of the description of the degradation process; by combining the Kalman filter with the EM algorithm, the drift coefficient and diffusion coefficient of each stage can be dynamically updated, thereby solving the problems of prediction deviation caused by fixed parameters and the inability of the offline model to adapt to real-time degradation, and ensuring the continuity of the multi-stage degradation model during multi-stage switching. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, other relevant drawings can be obtained based on these drawings without creative work. In the drawings: Figure 1 This is a method flow chart of a method for predicting the remaining life of equipment components in this embodiment; Figure 2a This is a schematic diagram of the wear state estimation result of Cutter4 using the technical solution in this embodiment; Figure 2b This is a schematic diagram of the wear state estimation result of Cutter7 using the technical solution in this embodiment; Figure 3a This is a schematic comparison diagram of Cutter5 wear state estimation results of four existing prediction methods in this embodiment and the prediction method of this application; Figure 3b This is a schematic comparison diagram of Cutter9 wear state estimation results of four existing prediction methods in this embodiment and the prediction method of this application; Figure 4 This is a schematic diagram of system modules of a system for predicting the remaining life of equipment components in this embodiment; Figure 5 It is a structural schematic diagram of a computer device in this embodiment. DETAILED DESCRIPTION

[0019] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope of the present disclosure. Similarly, for the sake of clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0020] In the present disclosure, unless otherwise specified, the use of the terms "first", "second", etc. to describe various elements is not intended to limit the positional relationship, timing relationship, or importance relationship of these elements, and such terms are only used to distinguish one element from another element. In some examples, the first element and the second element may refer to the same instance of the element, and in some cases, based on the description of the context, they may also refer to different instances.

[0021] The terms used in the description of various examples in this disclosure are only for the purpose of describing specific examples and are not intended to be limiting. Unless the context clearly indicates otherwise, if the number of elements is not specifically limited, the element can be one or more. In addition, the term "and / or" used in this disclosure covers any one of the listed items and all possible combinations.

[0022] Example 1

[0023] See also Figure 1 , shows a method flow chart of a method for predicting the remaining life of equipment components, the method comprising: S1: Obtain the time domain features, frequency domain features and time-frequency domain features in the historical multi-source sensor data, construct the historical degradation index, and establish a multi-source mapping relationship function that characterizes the historical multi-source sensor data and the degradation state of the component based on the time domain features, frequency domain features and time-frequency domain features of the historical multi-source sensor data; Specifically, in this embodiment, the multi-source sensor data includes vibration data, current data and cutting force data, and a multi-source mapping relationship function characterizing the historical multi-source sensor data and the degradation state of the component is established, specifically: the time domain characteristics, frequency domain characteristics and time-frequency domain characteristics of the historical vibration data, the historical current data and the historical cutting force data are respectively obtained; It should be noted that, in this embodiment, the time domain features, frequency domain features and time-frequency domain features of vibration data, current data and cutting force data are extracted from the multi-source sensors respectively. The specific number depends on the actual situation and is not restricted here.

[0024] Specifically, in this embodiment, by fusing multiple source signals such as vibration, current, cutting force, etc., and extracting features in the time domain, frequency domain, and time-frequency domain, the nonlinear degradation characteristics are fully captured.

[0025] Calculate the Pearson correlation coefficients between multiple time domain features, frequency domain features, time-frequency domain features and component degradation states, and select features with high correlation as historical degradation indicators; It should be noted that, in this embodiment, the Pearson correlation coefficient between each feature and the degradation state of the component is calculated, and the calculation function is: ,in, Representation characteristics and component degradation history The Pearson correlation coefficient between represents the covariance function, Representation characteristics The standard deviation of Indicates historical degradation state The standard deviation of ; features with high Pearson correlation coefficients are selected as indicators of historical degradation; The historical degradation indicators are input into the feature fusion network of the attention mechanism to predict the degradation status of the components, and a multi-source mapping relationship function is constructed to characterize the historical multi-source sensor data and the degradation status of the components. It should be noted that, in this embodiment, the feature fusion network using the attention mechanism predicts the degradation state of the component according to the degradation index, which is a conventional technical means in the art and will not be described in detail here; at the same time, a multi-source mapping relationship function representing the historical multi-source sensor data and the degradation state of the component is constructed, specifically: ,in, represents the multi-source sensor signal collected at the current time k, D represents the number of channels of the sensor signal, and D is a positive integer. Represents the mapping relationship from the features of multi-source sensor data to the degradation status of components.

[0026] Specifically, in this embodiment, by adopting a feature fusion network with an attention mechanism, the impact of operating condition changes on degradation indicators (such as amplitude fluctuations) is adaptively eliminated, thereby solving the problem of poor monitoring stability under variable operating conditions.

[0027] S2: The component degradation cutoff value is calculated by using the ischange function combined with the historical degradation index to obtain multiple component degradation stages. Then, based on the component degradation cutoff value and the single-stage degradation function, a component multi-stage degradation model is constructed. Based on the chi-square Markov property of the component Wiener degradation process, a remaining life distribution model of the component degradation stage is constructed. It should be noted that, in this embodiment, based on the component degradation cutoff value and the single-stage degradation function, a component multi-stage degradation model is constructed, specifically: firstly, a component single-stage degradation function is established based on the Wiener process: ,in, express The degradation state of the component at the time, represents the initial degradation state, set to 0, represents the drift function, represents the model parameters, represents the diffusion coefficient, It is a standard Brownian motion; this technical means is a conventional technical means in this field and will not be described in detail here; Then, the ischange function is used in combination with the historical degradation index to calculate the component degradation cutoff value and construct a multi-stage component degradation model, specifically: ,in, Indicates a degraded state. represents the first stage degradation drift function, represents the integral operator, represents the first stage degradation model parameters, represents the diffusion parameter of the first stage degradation model, represents the standard Brownian motion, Represents a representative function, represents the degradation threshold of the component in the first degradation stage, represents the second stage degradation drift function, represents the second stage degradation model parameters, represents the diffusion parameter of the second stage degradation model, represents the standard Brownian motion, Represents a representative function, represents the degradation threshold of the component in the second degradation stage, represents the third stage degradation drift function, represents the third stage degradation model parameters, represents the diffusion parameter of the third stage degradation model, represents the standard Brownian motion, Represents a representative function.

[0028] Specifically, in this embodiment, based on the historical degradation index and the ischange function, the stage boundary value (such as the boundary value , , avoiding the subjectivity of manually setting thresholds to improve the accuracy of the description of the degradation process.

[0029] Based on the Chi-square Markov property of the component Wiener degradation process, the remaining life distribution model of the component degradation stage is constructed. Specifically, it is defined as: from the initial degradation moment to the first time the degradation state exceeds the first The time of the degradation stage threshold is , expressed as: According to the Chi-square Markov property of the component Wiener degradation process, Rewritten as: ,Then It can be expressed as: ,in, , ,in, represents the p-th stage life, represents the degradation value of the pth stage, It represents the degradation threshold value of the component in the pth degradation stage; The inverse Gaussian distribution function is used to construct the remaining life distribution model of the component degradation stage, specifically: ,in, represents the degradation model parameters of the pth stage, represents the drift function, represents the integral operator, represents the p-th stage life, Indicates p Diffusion parameters for the stage degradation model.

[0030] S3: Perform two-stage maximum likelihood estimation on historical degradation indicators and historical failure data, and fuse the two-stage maximum likelihood estimation to obtain the full log-likelihood function, then perform maximum likelihood estimation on the full log-likelihood function to obtain model parameter estimates, and use the model parameter estimates to update the model parameters of the remaining life distribution model to obtain the updated remaining life distribution model; Specifically, in this embodiment, a two-stage maximum likelihood estimation is performed on the historical degradation index and the historical failure data, and the two-stage maximum likelihood estimation is integrated to obtain the full logarithmic likelihood function, and then the full logarithmic likelihood function is subjected to maximum likelihood estimation to obtain the model parameter estimation value, specifically: based on the first The historical degradation index of the degradation stage of each component is used to construct the first model parameter and the first log-likelihood function of the second model parameter of the remaining life distribution model, which is specifically: ,in, represents the first model parameter, represents the second model parameter, represents the degradation state value, Indicates the number of historical devices. Indicates The historical equipment is in The number of degradation indicators in each degradation stage, Indicates the time interval, represents the degradation difference, Indicates time, represents the drift function; and uses the maximum likelihood estimation method to perform maximum likelihood estimation to obtain the estimated value of the first model parameter; Based on The historical failure data of the components in the degradation stage are used to construct the second logarithmic likelihood function of the third model parameter and the fourth model parameter of the remaining life distribution model, which is specifically: ,in, represents the third model parameter, represents the fourth model parameter, Indicates the time in stage 3, Indicates the number of historical data. represents the number of data with degraded data, represents the i-th historical failure time, represents the failure threshold, represents the drift function, represents a nonlinear function that characterizes the degradation rate. The first log-likelihood function and the second log-likelihood function are fused to obtain the full log-likelihood function, which is: , and perform maximum likelihood estimation on the full log-likelihood function to obtain the second model parameter estimate; using the first model parameter estimate and the second model parameter estimate, the model parameters of the remaining life distribution model are updated to obtain the updated remaining life distribution model.

[0031] Specifically, in this embodiment, by processing historical degradation indicators and historical failure data in stages and constructing log-likelihood functions respectively, the risk of overfitting of the degradation process can be avoided; by fusing degradation data reflecting the dynamics of the process with failure data reflecting the statistical characteristics of the end of life, the robustness of parameter estimation can be improved.

[0032] S4: Obtain the component degradation state prediction value corresponding to the current multi-source sensor data through the multi-source mapping relationship function, and use Kalman filtering combined with the EM algorithm to update the component multi-stage degradation model based on the component degradation state prediction value to obtain an updated multi-stage degradation model, input the current multi-source sensor data into the updated multi-stage degradation model to obtain the degradation state estimation value, and input the degradation state estimation value into the updated remaining life distribution model to obtain the remaining life prediction result.

[0033] It should be noted that, in this embodiment, the multi-stage degradation model of the component is updated by combining the Kalman filter with the EM algorithm to obtain an updated multi-stage degradation model, specifically: the component degradation state prediction value corresponding to the current multi-source sensor data is obtained by establishing a multi-source mapping relationship function; the component degradation state prediction value is introduced into the Kalman filter to obtain the estimated value of the component degradation state ,in, represents the mean, Represents and variance; dynamically updates the multi-stage degradation model by combining the estimated value with the EM algorithm to obtain an updated multi-stage degradation model; Specifically, in this embodiment, by combining Kalman filtering with the EM algorithm, the drift coefficient and diffusion coefficient of each stage can be dynamically updated, which solves the related problems of prediction deviation caused by fixed parameters and the inability of the offline model to adapt to real-time degradation, while also ensuring the continuity of the multi-stage degradation model when switching between multiple stages.

[0034] The current multi-source sensing data is input into the updated multi-stage degradation model to obtain the degradation state estimation value, and the degradation state estimation value is input into the updated remaining life distribution model to obtain the remaining life prediction result, which is specifically: ,in, , , represents the remaining life probability density function, Remaining life, represents the translation and scaling parameters, Represents the multi-source mapping result, represents the state estimation variance, represents the diffusion coefficient, represents the nonlinear function that characterizes the degradation rate, Indicates time, represents the state estimated mean, represents the drift function; time, satisfy When the remaining life of the core components can be expressed as: ,in, express tk The state of constant degradation, express The state of constant degradation, express The state of constant degradation, Remaining life, represents the lifetime of the p+1th stage, Indicates the third stage of life.

[0035] In another embodiment, the method further includes: constructing a state relationship function between the degradation state mapping result of the multi-source mapping relationship function and the actual degradation state based on the observation function, and optimizing and updating the function parameters of the multi-source mapping relationship function using the state relationship function to obtain the multi-source mapping relationship function after parameter optimization and update.

[0036] It should be noted that, in this embodiment, the observation function is described by a linear function, which can be expressed as: ,in, Represents the multi-source mapping result, and represents the translation and scaling parameters, Represents mapping noise; After obtaining the degradation state prediction value through the aforementioned multi-source mapping relationship function, by minimizing the distance between the prediction value and the true degradation state To estimate and The estimated value of . It is expressed as: , Represents the multi-source mapping result, and represents the shift-scaling parameter estimate, Indicates a degraded state. Indicates the number of data.

[0037] Specifically, in this embodiment, by dynamically adjusting the multi-source mapping relationship function, the predicted value of the multi-source mapping relationship function is closer to the actual degradation state, which improves the accuracy of the remaining life prediction and can further reduce the prediction error. At the same time, it can ensure the stability and adaptability of the multi-source mapping relationship function in long-term operation.

[0038] To verify the prediction effect of this method, 9 full-life tool milling data were used for verification. The cutting process starting from one side of the workpiece and completing on the other side of the workpiece was set as a cycle. In each cycle, five types of signals were collected through the triaxial accelerometer PCB 356A15, the current sensor CSA201-P030T01, the KISTLER 9257B triaxial force sensor and the free-field microphone Bruel&Kjaer 4966. After completing 8 cutting cycles, the tool was disassembled and the flank wear VB was measured using a visual microscope. In the offline estimation stage, first, the features of each channel data were extracted from the time domain, frequency domain and time-frequency domain. Among them, the time domain includes five dimensional features, namely, mean, variance, root mean square value, average absolute value and peak value, and six dimensionless features, namely, waveform index, peak index, pulse index, margin, skewness and kurtosis. The frequency domain signal consists of full-band signal and sub-band signal. Eleven features are extracted, including spectrum sum, spectrum mean, spectrum variance, spectrum entropy, spectrum root mean square, spectrum kurtosis, spectrum skewness, spectrum shape mean, spectrum shape variance, spectrum shape third-order moment and spectrum shape fourth-order moment. Eight time-frequency domain features are generated by performing Haar wavelet packet transform and three-layer decomposition. Then, by calculating the Pearson correlation coefficient between each feature and the tool wear value, highly representative features are selected to train the feature fusion network based on the attention mechanism.

[0039] Finally, the tool degradation stage change decomposition value is obtained through the collected wear data, and the multi-stage degradation model parameters are estimated. In the online prediction stage, first, the online tool multi-source signal is input into the feature fusion network to obtain the degradation state prediction value. Then the tool degradation state is obtained through Kalman filtering and the degradation model parameters are updated. Finally, the tool state and parameters are brought into the remaining life prediction distribution to realize the state monitoring and remaining life prediction of core components. The wear state estimation results of Cutter4 and Cutter7 can be found in Figure 2a-2b As shown, it can be seen that the prediction method of the present application is effective in estimating tool wear.

[0040] The prediction results using different methods are as follows Figure 3a-Figure 3b As shown in the figure, M1 is based on a nonlinear Wiener process for remaining life prediction. M2 takes into account the measurement variability based on M1. However, they do not consider the multi-stage characteristics of tool wear degradation. M3 is a multi-stage degradation model based on a linear Wiener process. M4 is the most commonly used deep learning tool prediction method (CNN+LSTM network). It can be seen from the figure that compared with the other four methods, this method fully considers the stage and nonlinear characteristics of tool degradation, and achieves RUL prediction under different working conditions with higher accuracy.

[0041] Example 2 See also Figure 4As shown, the present invention also provides a system for predicting the remaining life of equipment components, which is used in any one of the above-mentioned methods for predicting the remaining life of equipment components, and the system includes: The first module 100 is used to obtain time domain features, frequency domain features and time-frequency domain features in historical multi-source sensor data, construct historical degradation indicators, and establish a multi-source mapping relationship function that characterizes the historical multi-source sensor data and the degradation state of the component based on the time domain features, frequency domain features and time-frequency domain features of the historical multi-source sensor data; The second module 200 is used to calculate the component degradation cutoff value by using the ischange function in combination with the historical degradation index, obtain multiple component degradation stages, and then construct a component multi-stage degradation model based on the component degradation cutoff value and the single-stage degradation function, and construct a remaining life distribution model of the component degradation stage based on the chi-square Markov property of the component Wiener degradation process; The third module 300 is used to perform two-stage maximum likelihood estimation on the historical degradation index and the historical failure data, and fuse the two-stage maximum likelihood estimation to obtain a full logarithmic likelihood function, and then perform maximum likelihood estimation on the full logarithmic likelihood function to obtain model parameter estimates, and use the model parameter estimates to update the model parameters of the remaining life distribution model to obtain an updated remaining life distribution model; The fourth module 400 is used to obtain the component degradation state prediction value corresponding to the current multi-source sensor data through the multi-source mapping relationship function, and use Kalman filtering combined with the EM algorithm to update the component multi-stage degradation model based on the component degradation state prediction value to obtain an updated multi-stage degradation model, input the current multi-source sensor data into the updated multi-stage degradation model to obtain the degradation state estimation value, and input the degradation state estimation value into the updated remaining life distribution model to obtain the remaining life prediction result.

[0042] It should be noted that the modules in the system of Example 2 correspond to the steps in the method of Example 1. The steps in the method of Example 1 have been described in detail in Example 1. In this Example 2, the contents of the modules in the system will not be described in detail.

[0043] Example 3 See also Figure 5 As shown, this embodiment further provides a computer device, including a system memory 1005 and a processor 1001, wherein the system memory 1005 stores a computer program, and the processor 1001 implements the steps of any of the above methods when executing the computer program.

[0044] It should be noted that the processor 1001 is used to execute the steps in the above method embodiments according to the instructions in the program code. Alternatively, the processor 1001 implements the functions of each module / unit in the above system / device embodiments when executing the computer program.

[0045] Specifically, in this embodiment, the computer program may be divided into one or more modules / units, one or more modules / units are stored in the system memory 1005, and are executed by the processor 1001 to complete the present application. One or more modules / units may be a series of computer program instruction segments that can complete specific functions, and the instruction segments are used to describe the execution process of the computer program in the terminal device.

[0046] The terminal device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The terminal device may include, but is not limited to, a processor 1001 and a system memory 1005. Those skilled in the art will appreciate that this does not constitute a limitation on the terminal device, and may include more or less components than shown in the figure, or combine certain components, or different components. For example, the terminal device may also include an input / output device 1003, a network access device 1002, a bus 1006, etc.

[0047] The processor 1001 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0048] The system memory 1005 may be an internal storage unit of the terminal device, such as a hard disk or memory of the terminal device. The system memory 1005 may also be a storage device 1004 of the terminal device, such as a plug-in hard disk, a smart memory card (SmartMedia Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), etc. equipped on the terminal device. Further, the system memory 1005 may also include both the internal storage unit of the terminal device and the storage device 1004. The system memory 1005 is used to store computer programs and other programs and data required by the terminal device. The system memory 1005 may also be used to temporarily store data that has been output or is to be output.

[0049] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0050] Example 4 This embodiment provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of any one of the above methods are implemented.

[0051] Among them, the computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared or semiconductor system, system or device, or any combination of the above. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection with one or more wires, a portable computer disk, a hard disk. Random Access Memory (RAM), Read-Only Memory (ROM), Erasable Programmable Read Only Memory (EPROM), registers, hard disks, optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above, or any other form of computer-readable storage medium known in the art.

[0052] An exemplary storage medium is coupled to a processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can be located in an application-specific integrated circuit (ASIC). In an embodiment of the present invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system, or device.

[0053] Example 5 This embodiment also provides a computer program product including instructions. When the instructions are executed by a computer device cluster, the computer device cluster executes the method described in Embodiment 1.

[0054] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for predicting the remaining life of equipment components, characterized in that the method include: Obtain the time domain features, frequency domain features and time-frequency domain features in the historical multi-source sensor data, construct the historical degradation index, and establish a multi-source mapping relationship function that characterizes the historical multi-source sensor data and the degradation state of the component based on the time domain features, frequency domain features and time-frequency domain features of the historical multi-source sensor data; Based on the observation function, a state relationship function between the degradation state mapping result of the multi-source mapping relationship function and the actual degradation state is constructed, and the function parameters of the multi-source mapping relationship function are optimized and updated using the state relationship function to obtain the multi-source mapping relationship function after parameter optimization and update; The component degradation cutoff value is calculated by using the ischange function combined with the historical degradation index, and multiple component degradation stages are obtained. Then, based on the component degradation cutoff value and the single-stage degradation function, a component multi-stage degradation model is constructed. Based on the chi-square Markov property of the component Wiener degradation process, a remaining life distribution model of the component degradation stage is constructed. Perform two-stage maximum likelihood estimation on historical degradation indicators and historical failure data, and integrate the two-stage maximum likelihood estimation to obtain the full log-likelihood function, then perform maximum likelihood estimation on the full log-likelihood function to obtain model parameter estimates, and use the model parameter estimates to update the model parameters of the remaining life distribution model to obtain the updated remaining life distribution model; The component degradation state prediction value corresponding to the current multi-source sensor data is obtained through the multi-source mapping relationship function, and the component multi-stage degradation model is updated based on the component degradation state prediction value by using Kalman filtering combined with the EM algorithm to obtain an updated multi-stage degradation model, and the current multi-source sensor data is input into the updated multi-stage degradation model to obtain the degradation state estimation value, and the degradation state estimation value is input into the updated remaining life distribution model to obtain the remaining life prediction result.

2. A method for predicting the remaining life of equipment components according to claim 1, characterized in that: The time domain features, frequency domain features and time-frequency domain features in the historical multi-source sensor data are obtained, and the historical degradation index is constructed. According to the time domain features, frequency domain features and time-frequency domain features of the historical multi-source sensor data, a multi-source mapping relationship function is established to characterize the historical multi-source sensor data and the degradation state of the component. The multi-source sensor data includes vibration data, current data and cutting force data, which are specifically: The time domain characteristics, frequency domain characteristics and time-frequency domain characteristics of the historical vibration data, the historical current data and the historical cutting force data are respectively obtained; Calculate the Pearson correlation coefficients between multiple time domain features, frequency domain features, time-frequency domain features and component degradation states, and select features with high correlation as historical degradation indicators; The historical degradation indicators are input into the feature fusion network of the attention mechanism to predict the degradation status of the component, and a multi-source mapping relationship function is constructed to characterize the historical multi-source sensor data and the degradation status of the component.

3. The method for predicting the remaining life of equipment components according to claim 1, characterized in that: Based on the Chi-square Markov property of the component Wiener degradation process, the remaining life distribution model of the component degradation stage is constructed, which is as follows: Define the time from the initial degradation moment to the first time the degradation state exceeds the The time of the degradation stage threshold is , according to the chi-square Markov property of the component Wiener degradation process, It is expressed as: ,in, , ,in, represents the p-th stage life, represents the degradation value of the pth stage, It represents the degradation threshold value of the component in the pth degradation stage; The inverse Gaussian distribution function is used to construct the remaining life distribution model of the component degradation stage, specifically: ,in, represents the degradation model parameters of the pth stage, represents the drift function, represents the integral operator, represents the p-th stage life, Indicates p Diffusion parameters for the stage degradation model.

4. The method for predicting the remaining life of equipment components according to claim 1, characterized in that: The component degradation cutoff value is calculated by combining the ischange function with the historical degradation index, and multiple component degradation stages are obtained. Then, a component multi-stage degradation model is constructed based on the component degradation cutoff value and the single-stage degradation function, specifically: The ischange function is used in combination with historical degradation indicators to find multiple component degradation cutoff values ​​and obtain multiple component degradation stages; Based on the single-stage degradation function, a multi-stage degradation model is constructed using multiple component degradation cutoff values, specifically: ,in, Indicates a degraded state. represents the first stage degradation drift function, represents the integral operator, represents the first stage degradation model parameters, represents the diffusion parameter of the first stage degradation model, represents the standard Brownian motion, Represents a representative function, represents the degradation threshold of the component in the first degradation stage, represents the second stage degradation drift function, represents the second stage degradation model parameters, represents the diffusion parameter of the second stage degradation model, represents the standard Brownian motion, Represents a representative function, represents the degradation threshold of the component in the second degradation stage, represents the third stage degradation drift function, represents the third stage degradation model parameters, represents the diffusion parameter of the third stage degradation model, represents the standard Brownian motion, Represents a representative function.

5. The method for predicting the remaining life of equipment components according to claim 1, characterized in that: The historical degradation index and historical failure data are estimated by two-stage maximum likelihood estimation, and the full logarithmic likelihood function is obtained by integrating the two-stage maximum likelihood estimation. The full logarithmic likelihood function is then estimated by maximum likelihood estimation to obtain the model parameter estimates. The model parameter estimates are used to update the model parameters of the remaining life distribution model to obtain the updated remaining life distribution model, which is specifically: Based on The first log-likelihood function of the first model parameter and the second model parameter of the remaining life distribution model is constructed based on the historical degradation index of the component degradation stage, and the maximum likelihood estimation method is used to perform maximum likelihood estimation to obtain the estimated value of the first model parameter; Based on The historical failure data of the components in the degradation stage are used to construct the second log-likelihood function of the third model parameter and the fourth model parameter of the remaining life distribution model, and the first log-likelihood function and the second log-likelihood function are fused to obtain the full log-likelihood function, and the full log-likelihood function is maximum-likelihood estimated to obtain the estimated value of the second model parameter; The model parameters of the remaining life distribution model are updated using the first model parameter estimate value and the second model parameter estimate value to obtain an updated remaining life distribution model.

6. A system for predicting the remaining life of equipment components, characterized in that: The system is used in a method for predicting the remaining life of equipment components according to any one of claims 1 to 5, and the system comprises: The first module is used to obtain the time domain features, frequency domain features and time-frequency domain features in the historical multi-source sensor data, construct the historical degradation index, and establish a multi-source mapping relationship function that characterizes the historical multi-source sensor data and the degradation state of the component based on the time domain features, frequency domain features and time-frequency domain features of the historical multi-source sensor data; The second module is used to calculate the component degradation cutoff value by using the ischange function combined with the historical degradation index, and obtain multiple component degradation stages. Then, based on the component degradation cutoff value and the single-stage degradation function, a component multi-stage degradation model is constructed, and based on the chi-square Markov property of the component Wiener degradation process, a remaining life distribution model of the component degradation stage is constructed; The third module is used to perform two-stage maximum likelihood estimation on historical degradation indicators and historical failure data, and fuse the two-stage maximum likelihood estimation to obtain the full logarithmic likelihood function, and then perform maximum likelihood estimation on the full logarithmic likelihood function to obtain model parameter estimates, and use the model parameter estimates to update the model parameters of the remaining life distribution model to obtain the updated remaining life distribution model; The fourth module is used to obtain the component degradation state prediction value corresponding to the current multi-source sensor data through the multi-source mapping relationship function, and use Kalman filtering combined with the EM algorithm to update the component multi-stage degradation model based on the component degradation state prediction value to obtain an updated multi-stage degradation model, input the current multi-source sensor data into the updated multi-stage degradation model to obtain the degradation state estimation value, and input the degradation state estimation value into the updated remaining life distribution model to obtain the remaining life prediction result.

7. A computer device comprising a system memory and a processor, wherein the system memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 5 are implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

9. A computer program product comprising instructions, characterized in that When the instructions are executed by a computer device cluster, the computer device cluster executes the method according to any one of claims 1 to 5.

Images