A method and system for predicting the wear life of a piston ring assembly

Abstract

The present invention relates to the technical field of piston ring group wear monitoring, specifically a method and system for predicting the wear life of a piston ring group; firstly, the working status data of the piston ring group is collected in real time, and the improved wavelet transform is used for denoising and the missing data is processed by the KNN interpolation method, then the data after standardization and normalization is processed, and the filter label generation and audit mechanism are used to ensure the ordered homogeneity of the data, and an ultra-limited learning machine model is constructed, and the hidden layer output matrix is ​​calculated by randomly generating hidden layer weights and bias vectors, and the model parameters are optimized by the white whale optimization algorithm to improve the prediction accuracy, and then the remaining life is predicted, and the final prediction result is used to calculate the fault risk coefficient, and compared with the preset threshold value to provide maintenance suggestions, and the real-time data and prediction results are displayed to assist users in making maintenance decisions. The present invention combines advanced signal processing technology and machine learning algorithms to improve the maintenance efficiency of the piston ring group and the reliability of equipment operation.

Description

Technical Field

[0001] The invention relates to the technical field of piston ring group wear monitoring, and in particular to a method and system for predicting the wear life of a piston ring group. Background Art

[0002] With the widespread application of industrial equipment, especially in the fields of automobiles, machinery manufacturing and aviation, the performance of piston ring groups as key power transmission components directly affects the operating efficiency and life of the engine. However, piston ring groups are prone to wear when working in a long-term high-load, high-temperature and high-pressure environment, resulting in performance degradation and even failure. Traditional maintenance methods mainly rely on regular inspections and manual judgment, which not only increases maintenance costs, but also makes it difficult to accurately predict the time of failure, posing a major safety hazard.

[0003] A Chinese invention patent application with publication number CN116757073A discloses a method for predicting the wear life of a piston ring group-cylinder liner, the method comprising: collecting real-time operating parameters, including operating time, engine speed, engine torque, intake particle size after air filtration, and intake particle concentration after air filtration; obtaining a pre-trained piston ring group-cylinder liner wear twin model, the training sample data is a multidimensional feature data set and the wear data in the multidimensional feature data set is obtained through a calibrated piston ring group-cylinder liner wear mechanism model; determining multiple real-time wear data corresponding to the piston ring group-cylinder liner through the piston ring group-cylinder liner wear twin model and real-time operating parameters, and determining the single-piece wear remaining life of each piston ring or cylinder liner according to the real-time wear data and wear threshold corresponding to each piston ring or cylinder liner; determining the wear remaining life of the piston ring group-cylinder liner based on the single-piece wear remaining life of each piston ring or cylinder liner.

[0004] With the rapid development of sensor technology and data analysis methods, real-time collection of equipment status data and life prediction combined with advanced prediction algorithms can significantly improve the maintenance accuracy and efficiency of equipment. Therefore, how to reduce the occurrence of failures and maintenance costs through accurate status monitoring and wear life prediction, quickly detect data loss or disorder during wireless transmission, and optimize the use and management of equipment has become an important issue that needs to be urgently addressed in the current field of engineering technology. Summary of the invention

[0005] The purpose of the present invention is to address the problems existing in the background technology and to propose a method and system for predicting the wear life of a piston ring set.

[0006] The technical solution of the present invention is a method for predicting the wear life of a piston ring assembly, comprising the following specific implementation steps:

[0007] S1. Obtaining the working status data of the piston ring assembly in real time;

[0008] S2. First, the noise in the working status data is eliminated by using the improved wavelet transform denoising method, and then the missing data is filled by using the KNN interpolation method. Then, the working status data is standardized and normalized, and a filter label is generated for the processed working status data.

[0009] S3. Based on the ELM model and the White Whale optimization algorithm, a piston ring group wear life prediction model is constructed and saved;

[0010] S4. Audit the ordered homogeneity of the working status data, convert the working status data into binary string form, and calculate the first-order filtering parameters and the second-order filtering parameters based on the bilinear mapping and the filter label. If the two match, the ordered homogeneity of the data is confirmed, and the trained piston ring group wear life prediction model is called, and the real-time working status data is used as input to output the remaining life prediction value. If the two do not match, the data is ignored and a safety alarm is triggered;

[0011] S5. Obtain the operating time and historical fault data of the current piston ring group, calculate the fault risk coefficient, and then compare the fault risk coefficient with the preset safety threshold and danger threshold to evaluate the safety status and generate maintenance suggestions;

[0012] S6. Display real-time working status data, remaining life prediction value, failure risk factor, assessed safety status and maintenance recommendations through a visual interface, and generate a maintenance report.

[0013] Preferably, the piston ring group working state data includes real-time vibration data, real-time working temperature data, real-time working pressure data and real-time friction data.

[0014] Preferably, the implementation process of eliminating noise in the working status data by using the improved wavelet transform denoising method is as follows:

[0015] S31. The signal is decomposed into approximate coefficients of different scales, i.e., the low-frequency part and detail coefficients, i.e., the high-frequency part, by discrete wavelet transform. The formula is:

[0016] ;

[0017] ;

[0018] In the formula, Represents original signal data, i.e., a vibration data signal, an operating temperature data signal, an operating pressure data signal or a friction data signal; represents the scale function, i.e., the low-frequency component; represents the wavelet function, i.e., the high-frequency component; represents the approximate coefficient of the jth layer, i.e. the low-frequency part; represents the detail coefficient of the jth layer, i.e., the high-frequency part; k represents the translation position of the wavelet function in the signal, which is used to determine the local characteristics of the signal in time; represents the approximation coefficient; represents the detail coefficient; t represents the time variable in the signal data;

[0019] S32. Remove high frequency noise by adaptive threshold method:

[0020] Adaptive Threshold :Automatically adjust the threshold according to the local characteristics of the signal, calculate the local noise standard deviation of each layer of detail coefficients, and select the adaptive threshold according to the local noise standard deviation of each layer, that is:

[0021] ;

[0022] ;

[0023] In the formula, represents the adaptive threshold of the jth layer; Represents the local noise standard deviation of the detail coefficient of the jth layer; Represents the number of samples of detail coefficients of the jth layer; represents the kth detail coefficient of the jth layer;

[0024] Hard threshold: directly set the coefficients less than the threshold λ to 0, that is:

[0025] ;

[0026] Soft threshold: Shrink the remaining coefficients based on the hard threshold, that is:

[0027] ;

[0028] In the formula, Represents the detail coefficient after threshold processing; Represents detail coefficient The symbol function of return The positive and negative signs of Represents detail coefficient The absolute value of indicates the strength of the coefficient;

[0029] S33, perform inverse wavelet transform (IDWT) on the coefficients after threshold processing to reconstruct the signal:

[0030] ;

[0031] In the formula, represents the signal after denoising;

[0032] S34, obtaining the denoised working status data.

[0033] Preferably, the generation process of the filter tag is as follows:

[0034] S41, real-time vibration data Vdata i , Real-time working temperature data Tdata i , Real-time working pressure data Pdata i And real-time friction data Fdata i Convert to binary string form;

[0035] S42, real-time vibration data Vdata i Generate unit filter label UL 1i :

[0036] ;

[0037] Among them, α1 represents the exclusive unit filter label generation code for all vibration data, α1∈ , based on which the exclusive filter review code Cf1 of the vibration data is calculated: ; p represents the order of the multiplicative cyclic groups G1 and G2; g1 represents the generator of the multiplicative cyclic group G1; Represents a random integer ring = select an element from {0,1,…,p-1}; h represents the hash function, h: {0,1} * →G2; accordingly, the parameters {G1, G2, g1, h, Cf1} are made public;

[0038] S43, real-time working temperature data Tdata i Generate unit filter label UL 2i :

[0039] ;

[0040] Among them, α2 represents the exclusive unit filter label generation code for all vibration data, α2∈ , based on which the exclusive filtering review code Cf2 of the working temperature data is calculated: And the exclusive filter review code Cf2 of the working temperature data is made public;

[0041] S44, real-time working pressure data Pdata i Generate unit filter label UL 3i :

[0042] ;

[0043] Among them, α3 represents the exclusive unit filter label generation code for all vibration data, α3∈ , based on which the exclusive filtering review code Cf3 of the work pressure data is calculated: and make the exclusive filter review code Cf3 of work pressure data public;

[0044] S45, real-time friction data Fdata i Generate unit filter label UL 4i :

[0045] ;

[0046] Among them, α4 represents the exclusive unit filter label generation code for all vibration data, α4∈ , based on which the exclusive filter review code Cf4 of the friction data is calculated: And the exclusive filter review code Cf4 of friction data is made public;

[0047] S46: Generate filter label Lf i : .

[0048] Preferably, the audit process for the ordered homogeneity of the audit work status data is as follows:

[0049] S51, real-time vibration data Vdata i , Real-time working temperature data Tdata i , Real-time working pressure data Pdata i And real-time friction data Fdata i Convert to binary string form;

[0050] S52, calculate the first-order filtering parameter Pf1=e(g1, Lf i );

[0051] Among them, e represents a bilinear mapping, e: G1×G2→G T , G T represents a q-order multiplicative cyclic group;

[0052] S53, calculate the second-order filtering parameter Pf2:

[0053] ;

[0054] S54, if Pf1 = Pf2, it means that the received working status data {Vdata i , Tdata i , Pdata i , Fdata i} has ordered homogeneity, otherwise, filtering ignores the working status data {Vdata i , Tdata i, Pdata i , Fdata i} and issue a safety warning.

[0055] Preferably, the construction process of the piston ring group wear life prediction model is as follows:

[0056] S61. Collect and generate feature vectors of piston ring wear data ;

[0057] in, represents the feature vector of the i-th sample, Including vibration data, working temperature data, working pressure data, friction data, ; d represents the feature dimension of each sample; n represents the number of samples;

[0058] S62, construct an ELM model, given the input data, calculate the hidden layer output matrix H, the calculation formula of the hidden layer output matrix H is:

[0059] ;

[0060] In the formula, H represents the hidden layer output matrix; W represents the weight matrix input to the hidden layer, ; M represents the number of hidden layers; b represents the hidden bias vector, which is randomly initialized. ; g() represents the activation function;

[0061] S63. Calculate the output layer weight:

[0062] ;

[0063] Where B represents the weight matrix of the output layer; y represents the actual wear life value;

[0064] S64. The parameters of the ELM model are optimized by using the White Whale optimization algorithm to obtain the best prediction model, which is used as the wear life prediction model.

[0065] Preferably, the optimization process of the White Whale optimization algorithm to optimize the parameters of the ELM model is as follows:

[0066] S71, randomly initialize the position and fitness value of the white whale in the search space;

[0067] S72. Determine whether the current stage is exploration or development based on the balance factor, that is, by adjusting the balance factor To gradually transition from the search stage to the development stage, the calculation formula is as follows:

[0068] ;

[0069] In the formula, represents the balance factor; represents a random number in the interval (0,1); t1 represents the current iteration number; T1 represents the total iteration number;

[0070] when When >0.5, the population is in the exploration stage; ≤0.5, the population is in the development stage;

[0071] S73. In the exploration phase, the beluga whales are randomly selected to simulate their swimming behavior, update their positions, and explore the global search space. The algorithm uses different update strategies based on the parity of the positions. The calculation formula is as follows:

[0072] ;

[0073] In the formula, Indicates the value of the position of the i-th beluga whale in the j-th dimension at the next iteration; Represents the value of the position of the i-th beluga whale on the random dimension p in the current iteration; represents the value of the position of the random white whale r on the random dimension p in the current iteration; r1 and r2 represent random numbers in the interval (0,1); n is an integer;

[0074] S74. In the development stage, the beluga whale’s predation behavior is used to cooperate in foraging and moving according to the location information of adjacent beluga whales to update the location of the beluga whale. The Levy flight strategy is introduced to improve convergence and control the local search in the design space, that is, collaborative predation is performed by sharing surrounding location information to find the optimal location. The calculation formula is as follows:

[0075] ;

[0076] ;

[0077] In the formula, Indicates the position of the i-th beluga whale in the next iteration; Indicates the best location for beluga populations; represents the current position of the i-th beluga whale; represents the random jump degree, which measures the intensity of Levi's flight; L F () Levy flight function; represents the current position of the random white whale r; t1 represents the current iteration number; T1 represents the total iteration number; r3 and r4 represent random numbers in the interval (0,1);

[0078] S75. In the whale fall stage, the threats to beluga whales during migration and foraging, as well as the phenomenon that a few beluga whales fail to survive and fall to the bottom of the sea are simulated. By selecting the probability of a whale falling from the population, the slight changes in the population are simulated, that is, the position of the beluga whale and the step length of the whale fall are used to update the position. The calculation formula is as follows:

[0079] ;

[0080] ;

[0081] ;

[0082] ;

[0083] In the formula, represents the step length of the whale fall; r5, r6 and r7 represent random numbers in the interval (0,1); represents the step-down factor; Indicates the whale fall step length; u b Indicates the maximum value that the whale can reach when moving or updating its position. This value sets the highest position or maximum speed that the whale can reach during the simulation. b Indicates the minimum value when the whale moves or updates its position. This value defines the lowest position or minimum speed that the whale can reach during the simulation.

[0084] S76, calculating the fitness value of the updated Beluga whale, and updating the optimal solution of the algorithm and the position of the optimal solution;

[0085] S77. Repeat steps S72 to S76 until the maximum number of iterations is reached or other set stop conditions are met, thereby completing the optimization of the model parameters.

[0086] Preferably, the implementation process for evaluating the safety status and generating maintenance recommendations is as follows:

[0087] S81. Get the running time of the current piston ring group and historical failure data;

[0088] S82. Calculate the failure risk factor :

[0089] ;

[0090] In the formula, Indicates the failure risk factor, which is used to measure the proportion of the piston ring group's operating time to its total life; Indicates the running time of the current piston ring group; Represents the predicted remaining life value;

[0091] S83, the failure risk factor With the set safety threshold and danger threshold Compare and output the current status of the piston ring group and maintenance suggestions:

[0092] like ≥ , the piston ring assembly is in a dangerous state and needs to be stopped immediately and emergency repairs should be carried out;

[0093] like ≤ < , the piston ring group is in a warning state and requires planned maintenance. It is recommended to replace the piston ring group within a certain time range;

[0094] like < , the piston ring assembly is in a safe state and no maintenance measures are required.

[0095] The technical solution of the present invention is a piston ring group wear life prediction system, which is used in the above-mentioned piston ring group wear life prediction method, comprising:

[0096] A data acquisition module, used to obtain real-time working status data related to the wear of the piston ring group through a number of sensors;

[0097] The data processing module is used to clean, denoise, standardize and process missing values ​​of the collected raw data to ensure that the data is suitable for subsequent analysis and modeling;

[0098] Model training and optimization module, which is responsible for building a wear life prediction model using machine learning algorithms, optimizing model parameters, and evaluating its prediction performance;

[0099] A wear life prediction module is used to predict the remaining life of the piston ring group using the trained wear life prediction model;

[0100] Fault warning and decision support module, which is used to trigger warnings based on the predicted remaining life and notify maintenance personnel in advance to intervene;

[0101] The user interaction module is used to provide users with an intuitive operation interface, display prediction results and generate reports.

[0102] Compared with the prior art, the above technical solution of the present invention has the following beneficial technical effects:

[0103] The present invention designs a piston ring wear life prediction method and system, which has significant beneficial technical effects in improving the wear life prediction accuracy, optimizing equipment maintenance decisions, enhancing data communication reliability and security, and improving the system intelligence and automation level. It can effectively improve equipment management efficiency, reduce equipment failure rate, extend equipment service life, and enhance user experience:

[0104] (1) Improving the accuracy of wear life prediction: The use of multi-sensor data acquisition can comprehensively and real-timely obtain the working status data of the piston ring group, thereby improving the accuracy of wear life prediction. The data is denoised through improved wavelet transform, and the KNN interpolation method is used to fill in missing data. Standardization and normalization processing methods can effectively reduce the impact of noise, outliers and data missing on the prediction results, making the input data more reliable. The ultra-limited learning machine (ELM) model is used, and the model parameters are optimized through the White Whale Optimization Algorithm (BWO) to further improve the performance of the prediction model;

[0105] (2) Improving the intelligence and automation level of the system: The introduction of the edge computing module (data processing module) enables data processing to be performed close to the data source, reducing the delay in data transmission and improving the real-time performance and response speed of the system. In addition, the data processing and wear life prediction processes are highly automated, reducing manual intervention and system complexity, enabling the system to monitor and analyze the working status of the piston ring group efficiently and intelligently.

[0106] (3) Optimizing equipment maintenance decisions: Providing real-time maintenance decision suggestions based on remaining life prediction and failure risk coefficient, making equipment maintenance more accurate and timely, avoiding excessive or insufficient maintenance, and being able to classify the operating status of the equipment (dangerous, warning, safe), and comprehensively assessing risks based on operating hours and historical failure data, providing equipment maintenance personnel with specific and actionable maintenance plans, thereby effectively reducing equipment failure rates and maintenance costs;

[0107] (4) Improve the reliability and security of data communication: During the data transmission process, filter tags are generated and verified to ensure the orderly homogeneity of data between modules, thereby improving the reliability and accuracy of system data transmission and avoiding possible loss or tampering of data during transmission;

[0108] (5) Improving user experience: Through the user interaction module, the system can display real-time monitoring data, wear life prediction, failure risk and maintenance suggestions in an intuitive visual interface, enhancing the user's operating experience and understanding of equipment status, facilitating timely maintenance decisions, and automatically generating maintenance reports, further improving the convenience and transparency of equipment management, and facilitating maintenance personnel to deal with potential failures in a timely manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0109] Figure 1 This is a system architecture diagram of a piston ring group wear life prediction system proposed by the present invention;

[0110] Figure 2 The present invention provides a flow chart of a method for predicting the wear life of a piston ring assembly. DETAILED DESCRIPTION

[0111] Embodiment 1, as Figure 1 As shown, a piston ring group wear life prediction system proposed by the present invention includes a data acquisition module, a data processing module, a model training and optimization module, a wear life prediction module, a fault warning and decision support module, and a user interaction module.

[0112] The data acquisition module is used to obtain working status data related to the wear of the piston ring group through a number of sensors, including but not limited to:

[0113] Vibration data collection: The vibration data of the piston ring during the operation of the engine is collected through the acceleration sensor to reflect the contact between the piston ring and the cylinder wall;

[0114] Temperature data collection: Real-time monitoring of the working temperature data of the piston ring group through the temperature sensor;

[0115] Pressure data collection: The pressure sensor is used to monitor the working pressure data in the cylinder;

[0116] Friction data collection: The friction data between the piston ring and the cylinder wall is measured by a friction sensor;

[0117] The data processing module is an edge computing module, which is used to clean, denoise, standardize and process missing values ​​of the collected raw data to ensure that the data is suitable for subsequent analysis and modeling;

[0118] The model training and optimization module is responsible for building a wear life prediction model using machine learning algorithms, optimizing model parameters, and evaluating its prediction performance;

[0119] The wear life prediction module is used to predict the remaining life (RUL) of the piston ring group using the trained wear life prediction model;

[0120] The fault warning and decision support module is used to trigger warnings based on the predicted remaining life and notify maintenance personnel in advance to intervene;

[0121] The user interaction module provides users with an intuitive operation interface to display prediction results and generate reports.

[0122] Embodiment 2, as Figure 2As shown, a piston ring group wear life prediction method proposed in the present invention is applied to a piston ring group wear life prediction system proposed in Example 1, which specifically includes the following specific implementation steps:

[0123] S1. The data acquisition module obtains the working status data signal of the piston ring group in real time. The working status data signal is obtained in the following ways:

[0124] The vibration data signal of the piston ring during the operation of the engine is collected through the acceleration sensor, the working temperature data signal of the piston ring group is monitored in real time through the temperature sensor, the working pressure data signal in the cylinder is monitored through the pressure sensor, and the friction data signal between the piston ring and the cylinder wall is measured through the friction sensor;

[0125] Afterwards, the collected working status data signals, namely {vibration data signal, working temperature data signal, working pressure data signal, friction data signal} are transmitted to the data processing module.

[0126] S2, the data processing module (i.e. the edge computing module) first eliminates the noise in the working status data signal through denoising and filtering, then fills in the missing data through the KNN interpolation method, and then standardizes and normalizes the data. The specific implementation process is as follows:

[0127] S21. Using the improved wavelet transform denoising method, the signal is decomposed into low-frequency and high-frequency parts to identify and remove noise. The specific implementation process is as follows:

[0128] S2101. Decompose the signal into approximate coefficients (i.e., low-frequency part) and detail coefficients (i.e., high-frequency part) of different scales through discrete wavelet transform. The formula is:

[0129] ;

[0130] ;

[0131] In the formula, Represents original signal data, i.e., a vibration data signal, an operating temperature data signal, an operating pressure data signal or a friction data signal; represents the scale function, i.e., the low-frequency component; represents the wavelet function, i.e., the high-frequency component; represents the approximate coefficient of the jth layer, i.e. the low-frequency part; represents the detail coefficient of the jth layer, i.e., the high-frequency part; k represents the translation position of the wavelet function in the signal, which is used to determine the local characteristics of the signal in time; represents the approximation coefficient; represents the detail coefficient; t represents the time variable in the signal data;

[0132] S2102, remove high frequency noise by improving the threshold method:

[0133] Adaptive Threshold :Automatically adjust the threshold according to the local characteristics of the signal, calculate the local noise standard deviation of each layer of detail coefficients, and select the adaptive threshold according to the local noise standard deviation of each layer, that is:

[0134] ;

[0135] ;

[0136] In the formula, represents the adaptive threshold of the jth layer; Represents the local noise standard deviation of the detail coefficient of the jth layer; Represents the number of samples of detail coefficients of the jth layer; represents the kth detail coefficient of the jth layer;

[0137] Hard threshold: directly set the coefficients less than the threshold λ to 0, that is:

[0138] ;

[0139] Soft threshold: Shrink the remaining coefficients based on the hard threshold, that is:

[0140] ;

[0141] In the formula, Represents the detail coefficient after threshold processing; Represents detail coefficient The symbol function of return The positive and negative signs of Represents detail coefficient The absolute value of indicates the strength of the coefficient;

[0142] S2103, perform inverse wavelet transform (IDWT) on the coefficients after threshold processing to reconstruct the signal:

[0143] ;

[0144] In the formula, represents the denoised signal;

[0145] S22. Use K-nearest neighbor interpolation (KNN) to calculate the distance between the missing point and the known data point, select the most similar K adjacent data, and then use the average value of the adjacent data to fill the missing data to obtain the supplemented data. , the basic formula of the KNN interpolation method is:

[0146] ;

[0147] In the formula, represents the missing value of the prediction; represents the i-th adjacent data among the K data adjacent to the missing value; K represents the number of data adjacent to the missing value;

[0148] S23. Supplemented data Perform standardization to obtain standardized data z:

[0149] The standardization formula is:

[0150] ;

[0151] In the formula, z represents the standardized data; x represents the original data, that is, the supplemented data ; μ represents the supplemented data The mean of ; σ represents the supplemented data The variance of

[0152] S24, normalize the standardized data z, and output the normalized data F(t), that is, output the processed working status data {Vdata i , Tdata i , Pdata i , Fdata i}, the normalized formula is:

[0153] ;

[0154] Where, F(t) represents the normalized data; max(z) and min(z) represent the maximum and minimum values ​​of the normalized data, respectively;

[0155] Among them, Vdata i Represents real-time vibration data; Tdata i Indicates real-time working temperature data; Pdata i Indicates real-time working pressure data; Fdata i Represents real-time friction data;

[0156] S25. Since the data processing module is an edge computing module, the communication with the wear life prediction module is realized through a wireless network. The data processing module sends the data and transmits it to the wear life prediction module through a wireless communication channel. The data processing module and the wear life prediction module are respectively at the two ends of the data communication. Therefore, ensuring the orderly homogeneity of the data between the data processing module and the wear life prediction module is part of the technical problem to be solved by the present invention. For this reason, the present invention provides a working state data {Vdatai , Tdata i , Pdata i , Fdata i}Generate a filter tag. The generation process is as follows:

[0157] S2501, real-time vibration data Vdata i , Real-time working temperature data Tdata i , Real-time working pressure data Pdata i And real-time friction data Fdata i Convert to binary string form;

[0158] S2502, real-time vibration data Vdata i Generate unit filter label UL 1i :

[0159] ;

[0160] Among them, α1 represents the exclusive unit filter label generation code for all vibration data, which is generated when the system is initialized, α1∈ , based on which the exclusive filter review code Cf1 of the vibration data is calculated: ; p is the order of the multiplicative cyclic group G1 and G2; g1 is the generator of the multiplicative cyclic group G1; Represents a random integer ring = select an element from {0,1,…,p-1}; h represents the hash function, h: {0,1} * →G2; accordingly, the parameters {G1, G2, g1, h, Cf1} are disclosed to the wear life prediction module;

[0161] S2503, real-time working temperature data Tdata i Generate unit filter label UL 2i :

[0162] ;

[0163] Among them, α2 represents the exclusive unit filter label generation code for all vibration data, which is generated when the system is initialized, α2∈ , based on which the exclusive filtering review code Cf2 of the working temperature data is calculated: And disclose the exclusive filter review code Cf2 of the working temperature data to the wear life prediction module;

[0164] S2504, real-time working pressure data Pdata i Generate unit filter label UL 3i :

[0165] ;

[0166] Among them, α3 represents the exclusive unit filter label generation code for all vibration data, which is generated when the system is initialized, α3∈ , based on which the exclusive filtering review code Cf3 of the work pressure data is calculated: And disclose the exclusive filter review code Cf3 of the working pressure data to the wear life prediction module;

[0167] S2505, real-time friction data Fdata i Generate unit filter label UL 4i :

[0168] ;

[0169] Among them, α4 represents the exclusive unit filter label generation code for all vibration data, which is generated when the system is initialized, α4∈ , based on which the exclusive filter review code Cf4 of the friction data is calculated: And disclose the exclusive filter review code Cf4 of the friction data to the wear life prediction module;

[0170] S2506: Generate filter label Lf i : ;

[0171] S26, {Lf i , Vdata i , Tdata i , Pdata i , Fdata i}Transferred to the wear life prediction module.

[0172] S3, model training and optimization module builds a piston ring wear life prediction model, and its construction and training process is as follows:

[0173] S31. Collect and generate feature vectors of piston ring group wear data ;

[0174] in, represents the feature vector of the i-th sample, ; d represents the feature dimension of each sample, and in this embodiment, the value is 4, that is, Contains vibration data, working temperature data, working pressure data, and friction data; n represents the number of samples;

[0175] S32, construct an ELM model, given the input data, calculate the hidden layer output matrix H, the calculation formula of the hidden layer output matrix H is:

[0176] ;

[0177] In the formula, H represents the hidden layer output matrix; W represents the weight matrix input to the hidden layer, ; M represents the number of hidden layers; b represents the hidden bias vector, which is randomly initialized. ; g() represents an activation function, and the Sigmoid activation function is used in this embodiment;

[0178] S33. Calculate the output layer weight:

[0179] ;

[0180] Where B represents the weight matrix of the output layer; y represents the actual wear life value;

[0181] It should be noted that ELM has the characteristics of fast training speed and strong generalization ability. It randomly generates the weights and biases of the hidden layer and then solves the weights of the output layer. Assuming there is a series of sample data (X i , t i ), when the network contains L hidden layer nodes, it can be expressed as follows: ;

[0182] Among them, X j represents the jth input vector; represents the weight vector connecting i hidden layer nodes and output layer nodes; g() represents the activation function of the hidden layer; W i represents the weight vector connecting the i-th input layer node and the hidden layer node, b i represents the threshold of the i-th hidden layer neuron, o j represents the output vector; N represents the number of input vectors;

[0183] S34. The Beluga Whale Optimization (BWO) algorithm is used to optimize the parameters of the ELM model to obtain the best prediction model, which is used as the wear life prediction model. The parameter optimization process of the BWO algorithm is as follows:

[0184] S3401, parameter initialization: randomly initialize the position and fitness value of the white whale in the search space as a candidate solution, and at the same time, set the parameters of the algorithm, including but not limited to the population size, the maximum number of iterations, and the upper and lower bounds of the search space;

[0185] S3402, determine whether the current stage is exploration or development based on the balance factor, that is, by adjusting the balance factor To gradually transition from the search stage to the development stage, the calculation formula is as follows:

[0186] ;

[0187] In the formula, represents the balance factor; represents a random number in the interval (0,1); t1 represents the current iteration number; T1 represents the total iteration number;

[0188] when When >0.5, the population is in the exploration stage; ≤0.5, the population is in the development stage;

[0189] S3403. In the exploration phase, beluga whales are randomly selected to simulate their swimming behavior, update their positions, and explore the global search space. Different update strategies are used according to the parity of the positions. The calculation formula is as follows:

[0190] ;

[0191] In the formula, Indicates the value of the position of the i-th beluga whale in the j-th dimension at the next iteration; Represents the value of the position of the i-th beluga whale on the random dimension p in the current iteration; represents the value of the position of the random white whale r on the random dimension p in the current iteration; r1 and r2 represent random numbers in the interval (0,1); n is an integer;

[0192] S3404. In the development stage, the beluga whale's predation behavior is used to cooperate in foraging and moving according to the location information of adjacent beluga whales to update the location of the beluga whale. The Levy flight strategy is introduced to improve convergence and control the local search in the design space, that is, collaborative predation is performed by sharing surrounding location information to find the optimal position. The calculation formula is as follows:

[0193] ;

[0194] ;

[0195] In the formula, Indicates the position of the i-th beluga whale in the next iteration; Indicates the best location for beluga populations; represents the current position of the i-th beluga whale; represents the random jump degree, which measures the intensity of Levi's flight; L F () Levy flight function; represents the current position of the random white whale r; t1 represents the current iteration number; T1 represents the total iteration number; r3 and r4 represent random numbers in the interval (0,1);

[0196] S3405. In the whale fall stage, the threats to beluga whales during migration and foraging, as well as the phenomenon that a few beluga whales fail to survive and fall to the bottom of the sea are simulated. By selecting the probability of a whale falling from the population, the slight changes in the population are simulated, that is, the position of the beluga whale and the step length of the whale fall are used to update the position. The calculation formula is as follows:

[0197] ;

[0198] ;

[0199] ;

[0200] ;

[0201] In the formula, represents the step length of the whale fall; r5, r6 and r7 represent random numbers in the interval (0,1); represents the step-down factor; Indicates the whale fall step length; u b Indicates the maximum value that the whale can reach when moving or updating its position. This value sets the highest position or maximum speed that the whale can reach during the simulation. b Indicates the minimum value when the whale moves or updates its position. This value defines the lowest position or minimum speed that the whale can reach during the simulation.

[0202] S3406, calculating the fitness value of the updated white whale, and updating the optimal solution and the position of the optimal solution;

[0203] S3407, repeat steps S3402 to S3406 until the maximum number of iterations is reached or other stopping conditions are met;

[0204] It should be noted that the Beluga Optimizer (BWO) is an emerging meta-heuristic optimization algorithm. Its design is inspired by the life habits of beluga whales, especially their swimming, hunting and whale falling behaviors. The algorithm achieves a balance between global exploration and local development in the search space by simulating these natural behaviors of beluga whales in the ocean. The BWO algorithm has the characteristics of global search and local search balance, adaptability, strong convergence and simple implementation. It performs well in solving single-peak and multi-peak optimization problems and shows the potential for wide application in complex real-world optimization problems.

[0205] S35, training to obtain a piston ring group wear life prediction model, and saving the piston ring group wear life prediction model to a model training and optimization module.

[0206] S4, the wear life prediction module receives {Lf i , Vdata i , Tdatai , Pdata i , Fdata i}, and perform the following operations:

[0207] S41. Verify the received filter tag Lf i , audit real-time vibration data Vdata i , Real-time working temperature data Tdata i , Real-time working pressure data Pdata i and real-time friction data Fdata i The audit process is as follows:

[0208] S4101, real-time vibration data Vdata i , Real-time working temperature data Tdata i , Real-time working pressure data Pdata i And real-time friction data Fdata i Convert to binary string form;

[0209] S4102, calculate the first-order filtering parameter Pf1=e(g1, Lf i );

[0210] Among them, e represents a bilinear mapping, e: G1×G2→G T , G T represents a q-order multiplicative cyclic group;

[0211] S4103, calculate the second-order filtering parameter Pf2:

[0212] ;

[0213] S4104, if Pf1 = Pf2, then the received working status data {Vdata i , Tdata i , Pdata i , Fdata i} has ordered homogeneity, then execute step S42; otherwise, filter and ignore the received working status data {Vdata i , Tdata i , Pdata i , Fdata i}, and issue a security warning to the user interaction module;

[0214] S42, call the piston ring wear life prediction model, input x new ={Vdata i , Tdata i , Pdatai , Fdata i}, output the predicted remaining life :

[0215] ;

[0216] ;

[0217] In the formula, represents the predicted remaining life; represents the trained output layer weights of the piston ring group wear life prediction model; represents the trained hidden layer output of the piston ring group wear life prediction model; represents the trained hidden bias vector of the piston ring group wear life prediction model; g() represents the activation function, and the Sigmoid activation function is used in this embodiment; The weight matrix representing the trained input to the hidden layer of the piston ring wear life prediction model;

[0218] S43. The remaining life of Transmitted to the fault warning and decision support module.

[0219] S5, Fault warning and decision support module uses the predicted remaining wear life of the piston ring group , to warn of potential failures and provide decision support solutions for equipment maintenance. The specific implementation process is as follows:

[0220] S51. Get the running time of the current piston ring group and historical failure data;

[0221] S52. Calculate the failure risk factor :

[0222] ;

[0223] In the formula, Indicates the failure risk factor, which is used to measure the proportion of the piston ring group's operating time to its total life; Indicates the running time of the current piston ring group;

[0224] S53, the failure risk factor With the set safety threshold and danger threshold Compare and output the current status of the piston ring group and maintenance suggestions:

[0225] like ≥ , the piston ring assembly is in a dangerous state and needs to be stopped immediately and emergency repairs should be carried out;

[0226] like ≤ < , the piston ring group is in a warning state and requires planned maintenance. It is recommended to replace the piston ring group within a certain time range;

[0227] like < , the piston ring assembly is in a safe state and no maintenance measures are required;

[0228] S54, the failure risk factor , the current status of the piston ring group and maintenance recommendations are transmitted to the user interaction module.

[0229] S6, the user interaction module displays real-time data through a visual interface {Vdata i , Tdata i , Pdata i , Fdata i}、Predicted remaining life , Failure risk factor , current status of piston ring pack and maintenance recommendations, and generate maintenance reports.

[0230] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited thereto, and various changes can be made within the knowledge scope of technicians in the relevant technical field without departing from the purpose of the present invention.

Claims

1. A method for predicting the wear life of a piston ring assembly, characterized in that: The specific implementation steps include the following: S1. Obtaining the working status data of the piston ring assembly in real time; S2. First, the noise in the working status data is eliminated by using the improved wavelet transform denoising method, and then the missing data is filled by using the KNN interpolation method. Then, the working status data is standardized and normalized, and a filter label is generated for the processed working status data. The generation process of filter tags is as follows: S41, real-time vibration data Vdata i , Real-time working temperature data Tdata i , Real-time working pressure data Pdata i And real-time friction data Fdata i Convert to binary string form; S42, real-time vibration data Vdata i Generate unit filter label UL 1i : ； Among them, α1 represents the exclusive unit filter label generation code for all vibration data, α1∈ , based on which the exclusive filter review code Cf1 of the vibration data is calculated: ; p represents the order of the multiplicative cyclic groups G1 and G2; g1 represents the generator of the multiplicative cyclic group G1; Represents a random number from the integer ring = select an element from {0,1,…,p-1}; h represents the hash function, h: {0,1} * →G2; accordingly, the parameters {G1, G2, g1, h, Cf1} are made public; S43, real-time working temperature data Tdata i Generate unit filter label UL 2i : ； Among them, α2 represents the exclusive unit filter label generation code for all vibration data, α2∈ , based on which the exclusive filtering review code Cf2 of the working temperature data is calculated: And the exclusive filter review code Cf2 of the working temperature data is made public; S44, real-time working pressure data Pdata i Generate unit filter label UL 3i : ； Among them, α3 represents the exclusive unit filter label generation code for all vibration data, α3∈ , based on which the exclusive filtering review code Cf3 of the work pressure data is calculated: and make the exclusive filter review code Cf3 of work pressure data public; S45, real-time friction data Fdata i Generate unit filter label UL 4i : ； Among them, α4 represents the exclusive unit filter label generation code for all vibration data, α4∈ , based on which the exclusive filter review code Cf4 of the friction data is calculated: And the exclusive filter review code Cf4 of friction data is made public; S46: Generate filter label Lf i : ; S3. Based on the ELM model and the White Whale optimization algorithm, a piston ring group wear life prediction model is constructed and saved; S4. Audit the ordered homogeneity of the working status data, convert the working status data into binary string form, and calculate the first-order filtering parameters and the second-order filtering parameters based on the bilinear mapping and the filter label. If the two match, the ordered homogeneity of the data is confirmed, and the trained piston ring group wear life prediction model is called, and the real-time working status data is used as input to output the remaining life prediction value. If the two do not match, the data is ignored and a safety alarm is triggered; S5. Obtain the operating time and historical fault data of the current piston ring group, calculate the fault risk coefficient, and then compare the fault risk coefficient with the preset safety threshold and danger threshold to evaluate the safety status and generate maintenance suggestions; S6. Display real-time working status data, remaining life prediction value, failure risk factor, assessed safety status and maintenance recommendations through a visual interface, and generate a maintenance report.

2. The method for predicting the wear life of a piston ring set according to claim 1, characterized in that: The working status data of the piston ring group includes real-time vibration data, real-time working temperature data, real-time working pressure data and real-time friction data.

3. The method for predicting the wear life of a piston ring set according to claim 1, characterized in that: The implementation process of eliminating noise in working status data by improved wavelet transform denoising method is as follows: S31. The signal is decomposed into approximate coefficients of different scales, i.e., the low-frequency part and detail coefficients, i.e., the high-frequency part, by discrete wavelet transform. The formula is: ； ； In the formula, Represents original signal data, i.e., a vibration data signal, an operating temperature data signal, an operating pressure data signal or a friction data signal; represents the scale function, i.e., the low-frequency component; represents the wavelet function, i.e., the high-frequency component; represents the approximate coefficient of the jth layer, i.e. the low-frequency part; represents the detail coefficient of the jth layer, i.e., the high-frequency part; k represents the translation position of the wavelet function in the signal, which is used to determine the local characteristics of the signal in time; represents the approximation coefficient; represents the detail coefficient; t represents the time variable in the signal data; S32. Remove high frequency noise by adaptive threshold method: Adaptive Threshold :Automatically adjust the threshold according to the local characteristics of the signal, calculate the local noise standard deviation of each layer of detail coefficients, and select the adaptive threshold according to the local noise standard deviation of each layer, that is: ； ； In the formula, represents the adaptive threshold of the jth layer; Represents the local noise standard deviation of the j-th layer detail coefficient; Represents the number of samples of detail coefficients of the jth layer; represents the kth detail coefficient of the jth layer; Hard threshold: directly set the value less than the threshold λ j The coefficient of is set to 0, that is: ； Soft threshold: Shrink the remaining coefficients based on the hard threshold, that is: ； In the formula, Represents the detail coefficient after threshold processing; Represents detail coefficient The symbol function of return The positive and negative signs of Represents detail coefficient The absolute value of indicates the strength of the coefficient; S33, perform inverse wavelet transform IDWT on the coefficients after threshold processing to reconstruct the signal: ； In the formula, represents the denoised signal; S34, obtaining the denoised working status data.

4. The method for predicting the wear life of a piston ring set according to claim 1, characterized in that: The audit process for the ordered homogeneity of audit work status data is as follows: S51, real-time vibration data Vdata i , Real-time working temperature data Tdata i , Real-time working pressure data Pdata i And real-time friction data Fdata i Convert to binary string form; S52, calculate the first-order filtering parameter Pf1=e(g1, Lf i ); Among them, e represents a bilinear mapping, e: G1×G2→G T , G T represents a q-order multiplicative cyclic group; g1 represents the generator of the multiplicative cyclic group G1; G2 represents the multiplicative cyclic group; Lf i Indicates the filter tag; S53, calculate the second-order filtering parameter Pf2: ； Among them, Cf1 represents the exclusive filter review code of vibration data; Cf2 represents the exclusive filter review code of working temperature data; Cf3 represents the exclusive filter review code of working pressure data; Cf4 represents the exclusive filter review code of friction data; S54, if Pf1 = Pf2, it means that the received working status data {Vdata i , Tdata i , Pdata i , Fdata i } has ordered homogeneity, otherwise, filtering ignores the working status data {Vdata i , Tdata i , Pdata i , Fdata i } and issue a safety warning.

5. The method for predicting the wear life of a piston ring set according to claim 1, characterized in that: The construction process of the piston ring group wear life prediction model is as follows: S61. Collect and generate feature vectors of piston ring wear data ; in, represents the feature vector of the i-th sample, Including vibration data, working temperature data, working pressure data, friction data, ; d represents the feature dimension of each sample; n represents the number of samples; S62, construct an ELM model, given the input data, calculate the hidden layer output matrix H, the calculation formula of the hidden layer output matrix H is: ； In the formula, H represents the hidden layer output matrix; W represents the weight matrix input to the hidden layer, ; M represents the number of hidden layers; b represents the hidden bias vector, which is randomly initialized. ; g() represents the activation function; S63. Calculate the output layer weight: ； Where B represents the weight matrix of the output layer; y represents the actual wear life value; S64. The parameters of the ELM model are optimized by using the White Whale optimization algorithm to obtain the best prediction model, which is used as the wear life prediction model.

6. A method for predicting the wear life of a piston ring set according to claim 5, characterized in that: The optimization process of the White Whale optimization algorithm to optimize the parameters of the ELM model is as follows: S71, randomly initialize the position and fitness value of the white whale in the search space; S72. Determine whether the current stage is exploration or development based on the balance factor, that is, by adjusting the balance factor To gradually transition from the search stage to the development stage, the calculation formula is as follows: ； In the formula, represents the balance factor; Represents a random number in the interval (0,1); t1 represents the current iteration number; T1 represents the total number of iterations; when When >0.5, the population is in the exploration stage; ≤0.5, the population is in the development stage; S73. In the exploration phase, the beluga whales are randomly selected to simulate their swimming behavior, update their positions, and explore the global search space. The algorithm uses different update strategies based on the parity of the positions. The calculation formula is as follows: ； In the formula, Indicates the value of the position of the i-th beluga whale in the j-th dimension at the next iteration; Represents the value of the position of the i-th beluga whale on the random dimension p in the current iteration; represents the value of the position of the random white whale r on the random dimension p in the current iteration; r1 and r2 represent random numbers in the interval (0,1); n is an integer; S74. In the development stage, the beluga whale’s predation behavior is used to cooperate in foraging and moving according to the location information of adjacent beluga whales to update the location of the beluga whale. The Levy flight strategy is introduced to improve convergence and control the local search in the design space, that is, collaborative predation is performed by sharing surrounding location information to find the optimal location. The calculation formula is as follows: ； ； In the formula, Indicates the position of the i-th beluga whale in the next iteration; Indicates the best location for beluga populations; represents the current position of the i-th beluga whale; represents the random jump degree, which measures the intensity of Levi's flight; L F () Levy flight function; Indicates the current position of the random white whale r; t1 indicates the current iteration number; T1 represents the total number of iterations; r3 and r4 represent random numbers in the interval (0,1); S75. In the whale fall stage, the threats to beluga whales during migration and foraging, as well as the phenomenon that a few beluga whales fail to survive and fall to the bottom of the sea are simulated. By selecting the probability of a whale falling from the population, the slight changes in the population are simulated, that is, the position of the beluga whale and the step length of the whale fall are used to update the position. The calculation formula is as follows: ； ； ； ； In the formula, represents the step length of the whale fall; r5, r6 and r7 represent random numbers in the interval (0,1); represents the step-down factor; Indicates the whale fall step length; u b Indicates the maximum value that the whale can reach when moving or updating its position. This value sets the highest position or maximum speed that the whale can reach during the simulation. b Indicates the minimum value when the whale moves or updates its position. This value defines the lowest position or minimum speed that the whale can reach during the simulation. S76, calculating the fitness value of the updated Beluga whale, and updating the optimal solution of the algorithm and the position of the optimal solution; S77. Repeat steps S72 to S76 until the maximum number of iterations is reached or other set stop conditions are met, thereby completing the optimization of the model parameters.

7. The method for predicting the wear life of a piston ring set according to claim 1, characterized in that: The implementation process for evaluating the safety status and generating maintenance recommendations is as follows: S81. Get the running time of the current piston ring group and historical failure data; S82. Calculate the failure risk factor : ； In the formula, Indicates the failure risk factor, which is used to measure the proportion of the piston ring group's operating time to its total life; Indicates the running time of the current piston ring group; represents the predicted remaining life; S83, the failure risk factor With the set safety threshold and danger threshold Compare and output the current status of the piston ring group and maintenance suggestions: like ≥ , the piston ring assembly is in a dangerous state and needs to be stopped immediately and emergency repairs should be carried out; like ≤ < , the piston ring group is in a warning state and requires planned maintenance. It is recommended to replace the piston ring group within a certain time range; like < , the piston ring assembly is in a safe state and no maintenance measures are required.

8. A piston ring wear life prediction system, used to execute a piston ring wear life prediction method according to any one of claims 1 to 7, characterized in that: include: A data acquisition module, used to obtain real-time working status data related to the wear of the piston ring group through a number of sensors; The data processing module is used to clean, denoise, standardize and process missing values ​​of the collected raw data to ensure that the data is suitable for subsequent analysis and modeling; Model training and optimization module, which is responsible for building a wear life prediction model using machine learning algorithms, optimizing model parameters, and evaluating its prediction performance; A wear life prediction module is used to predict the remaining life of the piston ring group using the trained wear life prediction model; Fault warning and decision support module, which is used to trigger warnings based on the predicted remaining life and notify maintenance personnel in advance to intervene; The user interaction module is used to provide users with an intuitive operation interface, display prediction results and generate reports.

Images