Method, device and equipment for predicting service life of rotary mechanical part, medium and product

Abstract

The invention discloses a life prediction method, device, equipment, medium and product of a rotating mechanical component, and relates to the technical field of state monitoring and fault prediction of mechanical equipment, and the method comprises the steps: obtaining an operation monitoring signal of a target rotating mechanical component in a current time period, and extracting a preset key feature; analyzing the characteristic stability through multi-scale wavelet transform to obtain a current steady-state index, and determining the health state of the rotary mechanical part according to the current steady-state index; and inputting the characteristics into the corresponding steady-state or non-steady-state working condition life prediction model according to the state, and outputting a residual life prediction value. According to the method, calculation redundancy is reduced by screening key features, working condition self-adaptive division is realized by using multi-scale wavelet transform, and differentiated prediction models are constructed for different health states, so that the precision and robustness of service life prediction of the rotary mechanical part are remarkably improved under complex working conditions, and the service life prediction efficiency of the rotary mechanical part is improved. And light-weight and high-reliability intelligent maintenance decision support is realized.

Description

Technical Field

[0001] This application relates to the field of mechanical equipment condition monitoring and fault prediction technology, and in particular to a method, device, equipment, medium and product for predicting the life of rotating mechanical components. Background Technology

[0002] Rotating mechanical components are key parts of machinery and widely used in industry, and their performance directly affects equipment reliability and operating efficiency. Accurately predicting the remaining service life of rotating mechanical components is a core technology for achieving predictive maintenance and avoiding unplanned downtime.

[0003] In related technologies, life prediction methods for rotating mechanical components mainly rely on the analysis of signals such as vibration and temperature generated during the operation of rotating mechanical components. However, under complex operating conditions (such as load fluctuations, speed changes, or noise interference), prediction methods based on statistical models of steady-state signal characteristics or using a single fuzzy system (such as a Type-1 Fuzzy System, T1 FS) have significant shortcomings: on the one hand, the high-dimensional features extracted from the original signal have a large amount of redundancy, increasing the computational burden of the model and making it difficult to meet real-time requirements; on the other hand, signal noise and uncertainty under complex operating conditions can significantly interfere with the prediction accuracy of a single model, and the model itself is difficult to adaptively handle the differences in data characteristics under different operating conditions.

[0004] Therefore, there is an urgent need for a life prediction method for rotating mechanical components that can effectively handle feature redundancy, adapt to changes in operating conditions, and properly address data uncertainty, so as to improve prediction accuracy and practicality and provide reliable support for intelligent maintenance decisions. Summary of the Invention

[0005] The purpose of this application is to provide a method, device, equipment, medium, and product for predicting the life of rotating mechanical components. This method can effectively eliminate high-dimensional redundant features, automatically identify steady-state and non-steady-state operating conditions of rotating mechanical components, and model the uncertainty of operating conditions. As a result, it can still output high-precision and lightweight remaining life prediction results under complex operating conditions, providing reliable support for intelligent maintenance decisions.

[0006] To achieve the above objectives, this application provides the following solution: In a first aspect, this application provides a method for predicting the lifespan of rotating mechanical components, comprising: Acquire the operational monitoring signals of the target rotating mechanical component within the current time period; the operational monitoring signals include vibration signals, temperature signals, and rotational speed signals; Based on the operational monitoring signals within the current time period, extract preset key features; Based on preset key features, the current steady-state index of the target rotating mechanical component is obtained by performing a multi-scale wavelet transform method to analyze the stability of the working condition. Based on the current steady-state index, the health status of the target rotating mechanical component is determined for the current time period; the health status includes normal status, minor fault status, and severe fault status. If the health status is normal, the preset key features are input into the trained steady-state working condition life prediction model to obtain the remaining life prediction value of the target rotating mechanical component in the current time period. The steady-state working condition life prediction model is obtained by iteratively training the preset first machine learning model using the first training dataset. The first training dataset is constructed based on the preset key features and corresponding real values ​​of remaining life in the historical time period under normal working conditions. If the health status is a minor fault or a severe fault, the unsteady-state operating condition life prediction model trained with preset key features is used to obtain the remaining life prediction value of the target rotating mechanical component in the current time period. The unsteady-state operating condition life prediction model is obtained by iteratively training a preset second machine learning model using a second training dataset. The second training dataset is constructed based on preset key features and corresponding real values ​​of remaining life in historical time periods under minor and severe fault states.

[0007] Secondly, this application provides a life prediction device for rotating mechanical components, comprising: The data acquisition module is used to acquire the operation monitoring signals of the target rotating mechanical component within the current time period; the operation monitoring signals include vibration signals, temperature signals, and rotational speed signals; The feature extraction module is used to extract preset key features based on the operation monitoring signals within the current time period; The working condition analysis module is used to perform working condition stability analysis based on preset key features and through multi-scale wavelet transform method to obtain the current steady-state index of the target rotating mechanical component. The status determination module is used to determine the health status of the target rotating mechanical component in the current time period based on the current steady-state index; the health status includes normal status, minor fault status and severe fault status; The steady-state prediction module is used to input preset key features into the trained steady-state working condition life prediction model if the health state is normal, so as to obtain the remaining life prediction value of the target rotating mechanical component in the current time period. The steady-state working condition life prediction model is obtained by iteratively training a preset first machine learning model using a first training dataset. The first training dataset is constructed based on preset key features and corresponding real values ​​of remaining life in historical time periods under normal working conditions. The non-steady-state prediction module is used to obtain the remaining life prediction value of the target rotating mechanical component in the current time period by using a non-steady-state working condition life prediction model trained with preset key features if the health state is a minor fault state or a severe fault state. The non-steady-state working condition life prediction model is obtained by iteratively training a preset second machine learning model using a second training dataset. The second training dataset is constructed based on preset key features and corresponding real values ​​of remaining life in historical time periods under minor and severe fault states.

[0008] Thirdly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the life prediction method for rotating mechanical parts as described above.

[0009] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the life prediction method for rotating mechanical parts described above.

[0010] Fifthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the life prediction method for rotating mechanical components described above.

[0011] According to the specific embodiments provided in this application, this application has the following technical effects: This application provides a method, apparatus, equipment, medium, and product for predicting the lifespan of rotating mechanical components. It acquires operational monitoring signals of the target rotating mechanical component within a current time period; these signals include vibration, temperature, and rotational speed signals. Based on these signals, preset key features are extracted, solving the problems of high feature redundancy, high model computational complexity, and difficulty in ensuring real-time performance when directly modeling from high-dimensional raw signals. This fundamentally reduces data dimensionality and computational load, providing high-quality input for building a lightweight prediction model. By performing operational stability analysis using a multi-scale wavelet transform method based on the preset key features, the current steady-state index of the target rotating mechanical component is obtained. Based on this index, the health status of the target rotating mechanical component within the current time period is determined. This solves the problems of traditional methods struggling to accurately perceive and classify the real-time operating state of rotating mechanical components under complex dynamic conditions such as load fluctuations and rotational speed changes, and the difficulty in universally applying single prediction models. This achieves a comprehensive prediction model based on... Adaptive operating condition perception and refined state classification of signal time-frequency domain characteristics provide a precise basis for subsequent differentiated modeling. By inputting preset key features into a trained steady-state operating condition life prediction model if the health state is normal, the remaining life prediction value of the target rotating mechanical component for the current time period is obtained. Conversely, if the health state is a minor or severe fault state, preset key features are input into a trained non-steady-state operating condition life prediction model to obtain the remaining life prediction value of the target rotating mechanical component for the current time period. This solves the problems that a single model cannot simultaneously consider efficiency under steady-state conditions and robustness under non-steady-state conditions, and the insufficient uncertainty handling capability of traditional fuzzy systems. It achieves prediction models with different complexities and fault tolerance capabilities for different health states, and effectively characterizes and handles data uncertainties caused by noise and changes in operating conditions. Ultimately, it achieves high-precision, high-reliability, and lightweight prediction of the remaining life of rotating mechanical components in complex industrial environments, providing effective technical support for intelligent equipment maintenance and life management. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is an application environment diagram of a life prediction method for a rotating mechanical component according to an embodiment of this application; Figure 2 A flowchart illustrating a life prediction method for a rotating mechanical component provided in an embodiment of this application; Figure 3 A structural block diagram of a type of fuzzy system model provided in an embodiment of this application; Figure 4 This is a structural block diagram of an interval type II fuzzy system model provided in an embodiment of this application; Figure 5 This is a schematic diagram of an interval type II fuzzy set provided in an embodiment of this application; Figure 6 A flowchart illustrating a life prediction method for a rotating mechanical component, provided as another embodiment of this application; Figure 7 This is a schematic diagram of fuzzy set initialization for fuzzy clustering provided in an embodiment of this application; Figure 8 A schematic flowchart of a genetic algorithm provided in an embodiment of this application; Figure 9 for Figure 8 A schematic diagram of chromosome encoding in a genetic algorithm; Figure 10 A schematic diagram of the functional modules of a life prediction device for a rotating mechanical component provided in an embodiment of this application; Figure 11 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0014] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0015] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0016] The life prediction method for rotating mechanical components provided in this application can be applied to, for example... Figure 1In the application environment shown, terminal 101 communicates with server 102 via a network. A data storage system can store the data that server 102 needs to process. The data storage system can be set up independently, integrated into server 102, or placed in the cloud or on another server. Terminal 101 can send the operating monitoring signal of the target rotating mechanical component within the current time period to server 102. After receiving the operating monitoring signal of the target rotating mechanical component within the current time period, server 102 extracts preset key features based on the operating monitoring signal within the current time period; based on the preset key features, it performs a working condition stability analysis using a multi-scale wavelet transform method to obtain the current steady-state index of the target rotating mechanical component; based on the current steady-state index, it determines the health status of the target rotating mechanical component within the current time period; the health status includes normal state, minor fault state, and severe fault state; if the health status is normal, the preset key features are input into a trained steady-state working condition life prediction model to obtain the health status of the target rotating mechanical component within the current time period. The remaining life prediction value is obtained by iteratively training a preset first machine learning model using a first training dataset. The first training dataset is constructed based on preset key features and corresponding actual remaining life values ​​within a historical time period under normal operating conditions. If the health state is a minor or severe fault state, the remaining life prediction value of the target rotating mechanical component is obtained by iteratively training a preset second machine learning model using a second training dataset. The second training dataset is constructed based on preset key features and corresponding actual remaining life values ​​within a historical time period under minor and severe fault states. The server 102 can feed back the obtained remaining life prediction value of the target rotating mechanical component to the terminal 101. In addition, in some embodiments, the life prediction method for rotating mechanical components can also be implemented by the server 102 or the terminal 101 separately. For example, the terminal 101 can directly perform life prediction processing on the operation monitoring signal of the target rotating mechanical component in the current time period, or the server 102 can obtain the operation monitoring signal of the target rotating mechanical component in the current time period from the data storage system and perform life prediction processing on the operation monitoring signal of the target rotating mechanical component in the current time period.

[0017] The terminal 101 can be, but is not limited to, various desktop computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. The server 102 can be implemented using a standalone server or a server cluster consisting of multiple servers, or it can be a cloud server.

[0018] In one exemplary embodiment, such as Figure 2 As shown, a method for predicting the lifespan of rotating mechanical components is provided. This method is executed by a computer device, specifically a terminal or server, or both. In this embodiment, the method is applied to... Figure 1 Taking server 102 as an example, the explanation includes the following steps 201 to 206. Wherein: Step 201: Obtain the operational monitoring signals of the target rotating mechanical component within the current time period; the operational monitoring signals include vibration signals, temperature signals, and rotational speed signals. This application is applicable to rotating mechanical components with similar operational characteristics and fault evolution patterns, such as gears, shafts, and motor rotors. In these application scenarios, life prediction can be achieved simply by obtaining the operational monitoring signals of the corresponding rotating component and processing them according to the steps described in this application, maintaining the same technical principles and processing flow.

[0019] Step 202: Extract preset key features based on the operation monitoring signals within the current time period.

[0020] Step 203: Based on preset key features, perform operational stability analysis using multi-scale wavelet transform to obtain the current steady-state index of the target rotating mechanical component.

[0021] Step 204: Based on the current steady-state index, determine the health status of the target rotating mechanical component in the current time period; the health status includes normal status, minor fault status and severe fault status.

[0022] Step 205: If the health status is normal, then input the preset key features into the trained steady-state working condition life prediction model to obtain the remaining life prediction value of the target rotating mechanical component in the current time period; the steady-state working condition life prediction model is obtained by iteratively training the preset first machine learning model using the first training dataset, and the first training dataset is constructed based on the preset key features and corresponding real values ​​of remaining life in the historical time period under normal working conditions.

[0023] Step 206: If the health status is a minor fault state or a severe fault state, then the unsteady-state operating condition life prediction model trained with preset key features is used to obtain the remaining life prediction value of the target rotating mechanical component in the current time period. The unsteady-state operating condition life prediction model is obtained by iteratively training a preset second machine learning model using a second training dataset. The second training dataset is constructed based on preset key features and corresponding real values ​​of remaining life in historical time periods under minor and severe fault states.

[0024] By implementing steps 201 to 206 above, this application can effectively reduce the dimensionality and redundancy of input data by screening preset key features, significantly improving the representativeness and computational efficiency of features; it overcomes the shortcomings of traditional methods in adapting to dynamic working conditions by adaptively identifying and refining the operating state under complex working conditions through multi-scale wavelet transform; and based on this, it uses steady-state and non-steady-state working condition life predictions that match the state for differentiated inference, significantly enhancing the model's ability to handle data uncertainty and the robustness of overall prediction while ensuring lightweight design. Therefore, this application achieves more accurate, reliable, and efficient prediction of the remaining service life of rotating mechanical components under complex and variable operating conditions in industrial scenarios, providing strong technical support for predictive maintenance decisions.

[0025] In another exemplary embodiment of this application, the method for determining the preset key features specifically includes: Acquire the operational monitoring signals and corresponding actual remaining life values ​​of multiple rotating mechanical components within a historical time period.

[0026] Based on the operation monitoring signals of multiple rotating mechanical components within a historical time period, a candidate feature set is obtained by performing time-domain and frequency-domain feature calculations through feature extraction methods. The candidate feature set includes the maximum value, mean, standard deviation, root mean square, root mean square shape factor, skewness, kurtosis, crest factor, latitude factor, impulse factor, temperature value, and rotational speed value of the vibration signal in the vertical and horizontal directions, respectively.

[0027] The mutual information algorithm is used to calculate the first mutual information value between each candidate feature in the candidate feature set and its corresponding true remaining lifetime value, as well as the second mutual information value between any two candidate features in the candidate feature set.

[0028] A greedy search strategy is adopted to maximize the mRMR value as the optimization objective. Features are selected from the candidate feature set, and preset key features are obtained when preset stopping conditions are met. The mRMR value is calculated based on the first mutual information value and the second mutual information value. The preset stopping conditions include reaching a first preset maximum number of iterations or the change in mRMR value within a consecutive first preset number of iterations being less than a preset change threshold. The preset key features include the standard deviation, root mean square, and maximum value of the vibration signal in the vertical direction, the standard deviation, root mean square, and crest factor of the vibration signal in the horizontal direction, the temperature value, and the rotational speed value.

[0029] In another exemplary embodiment of this application, based on preset key features, a multi-scale wavelet transform method is used to perform operational stability analysis to obtain the current steady-state index of the target rotating mechanical component, specifically including: Based on preset key features, multi-scale wavelet transform is performed using the following formula: .

[0030] .

[0031] in, This represents the first-order wavelet transform; Represents the preset key features at time t; This represents the result of performing a first-order wavelet transform on the preset key features at time t. Describe the wavelet basis function at time t and scale s; express and Perform convolution operations; Indicates the second-order wavelet transform; This represents the result of performing a second-order wavelet transform on the preset key features at time t; Let represent the square of the wavelet basis function at time t and scale s; express and Perform convolution operations.

[0032] like If , then the steady-state index of the target rotating mechanical component at time t is 0; where, This represents the first preset discrimination threshold.

[0033] like ,and Then the steady-state index of the target rotating mechanical component at time t is 1; where, This indicates the second preset discrimination threshold.

[0034] If neither of the above two conditions is met, the steady-state index of the target rotating mechanical component at time t is calculated using the following formula: .

[0035] in, The steady-state exponent is represented at time t; The smoothing function representing the scale s at time t; Represents the weighting coefficients of the first-order wavelet transform; This represents the weighting coefficients of the second-order wavelet transform.

[0036] The current steady-state index is obtained by summing and averaging the steady-state indices at all points within the current time period.

[0037] In another exemplary embodiment of this application, the health status of the target rotating mechanical component in the current time period is determined based on the current steady-state index, specifically including: If the current steady-state index is greater than or equal to the first preset steady-state threshold, then the health state is normal.

[0038] If the current steady-state index is less than the first preset steady-state threshold and greater than or equal to the second preset steady-state threshold, then the health state is a minor fault state.

[0039] If the current steady-state index is less than the second preset steady-state threshold, the health status is a severe fault state.

[0040] In another exemplary embodiment of this application, the training process of the steady-state operating condition life prediction model specifically includes: Based on the first training dataset, each preset key feature is clustered using a fuzzy clustering algorithm to obtain the cluster center value of each preset key feature.

[0041] Based on the cluster center value of each preset key feature, a membership function of a preset type is constructed for each preset key feature, and the center parameter of the membership function of each preset key feature is initialized to the corresponding cluster center value; the preset type includes Gaussian type or triangular type.

[0042] Based on the first training dataset and the membership function initialized for each preset key feature, the linear parameters of the preset first machine learning model are identified by the least squares method to obtain the initial linear parameters of the steady-state operating condition life prediction model; the first machine learning model is a type of fuzzy system model.

[0043] With the goal of minimizing the root mean square error between the predicted remaining lifetime output by the first machine learning model and the corresponding actual remaining lifetime, the central parameters of the membership function of each preset key feature and the linear parameters of the steady-state operating condition life prediction model are iteratively optimized until the second preset maximum number of iterations is reached or the change in root mean square error within the second preset number of iterations is less than the preset root mean square error change threshold, thus obtaining the trained steady-state operating condition life prediction model.

[0044] In another exemplary embodiment of this application, the training process of the non-steady-state operating condition life prediction model specifically includes: Based on the second training dataset, each preset key feature is clustered using a fuzzy clustering algorithm to obtain the cluster center value of each preset key feature.

[0045] Based on the cluster center value of each preset key feature, an interval type II membership function is constructed for each preset key feature, and the center parameter of the interval type II membership function of each preset key feature is initialized to the corresponding cluster center value; the interval type II membership function includes an uncertainty interval formed by the lower membership function and the upper membership function.

[0046] Based on the second training dataset and the interval type II membership function initialized for each preset key feature, a genetic algorithm is used to globally optimize the parameters of the second machine learning model until a third preset maximum number of iterations is reached or the change in fitness function value within a consecutive third preset number of iterations is less than a preset threshold for the change in fitness function value, thus obtaining a trained unsteady-state operating condition life prediction model; the fitness function value is the negative of the root mean square error between the predicted remaining life value output by the second machine learning model and the corresponding true remaining life value; the second machine learning model is an interval type II fuzzy system model.

[0047] The following example illustrates this application using a specific life prediction process for rotating mechanical components.

[0048] To address the problems of feature redundancy, insufficient uncertainty handling, and poor adaptability to operating conditions in existing life prediction methods for rotating mechanical components, this application proposes a lightweight electric motor rotating mechanical component life prediction method based on fuzzy systems. This method reduces data dimensionality through mutual information feature selection, achieves operating condition classification through multi-scale wavelet transform, handles uncertainties by combining IT2 FS (Interval Type-2 Fuzzy System), and optimizes parameters through a genetic algorithm, thereby achieving efficient and accurate RUL (Remaining Useful Life) prediction.

[0049] The technical solution adopted by this application to solve its technical problem is: Step 1: Collect operating signals from rotating mechanical components of the motor, including vibration signals (sampling frequency 25.6kHz), temperature signals (sampling frequency 10Hz), and speed signals, and convert them into digital signals through a signal conditioning circuit.

[0050] This embodiment uses rolling and rotating mechanical parts as the prediction object, and the signal acquisition equipment is a high-precision accelerometer and a thermocouple.

[0051] Step 2: Extract the operating features of rotating mechanical components from the acquired signals, including the maximum value, mean, standard deviation, root mean square, root mean square shape factor, skewness, kurtosis, crest factor, latitude factor, impulse factor, and temperature in the vertical and horizontal directions of the vibration signal. Standardize the features and construct a dataset containing RUL (Remaining Usable Life) labels.

[0052] As an optional implementation, step 2 is performed as follows: Step 2.1: Feature Extraction. Throughout the entire lifecycle of the rotating mechanical component, a set of data is collected every 10 seconds, and 20 features are extracted. Each set of 20 extracted features includes: the maximum value in the vertical direction (unit: m / s). 2); Vertical mean (unit: m / s) 2 ); Vertical standard deviation (unit: m / s) 2 ); Root mean square in the vertical direction (unitless); Root mean square shape factor in the vertical direction (unitless); Skewness in the vertical direction (unitless); Kurtosis in the vertical direction (unitless); Crest factor in the vertical direction (unitless); Latitude factor in the vertical direction (unitless); Impulse factor in the vertical direction (unitless); Maximum value in the horizontal direction (unit: m / s) 2 ); mean value in the horizontal direction (unit: m / s) 2 ); Standard deviation in the horizontal direction (unit: m / s) 2 ); Horizontal root mean square (unitless); Horizontal root mean square shape factor (unitless); Horizontal skewness (unitless); Horizontal kurtosis (unitless); Horizontal crest factor (unitless); Horizontal latitude factor (unitless); Horizontal impulse factor (unitless); Corresponding characteristics and temperature of the vibration signal in the horizontal direction (unit: °C).

[0053] Step 2.2: Apply min-max normalization to all operational features, using the following formula: .

[0054] in, Represents the normalized eigenvalues; Represents the original operating characteristic values; This represents the minimum value of the feature in the current dataset; This represents the maximum value of this feature in the current dataset.

[0055] Step 2.3: Construct the dataset with RUL labels indicating a lifecycle countdown (e.g., from 2000 cycles to 1), containing 100 samples, 70 training samples, and 30 validation samples.

[0056] Step 3: Design a feature selection model based on mutual information (quantifying the degree of interdependence between two random variables. It is a non-parametric measure of statistical correlation between variables, capable of capturing not only linear relationships but also non-linear ones). The Minimum Redundancy Maximum Relevance (mRMR) evaluation index is used to calculate the RUL and mutual information between features, screening key features (such as vibration standard deviation, root mean square, and temperature) and eliminating redundant features. In this embodiment, mutual information calculation is based on a joint probability distribution, and the mRMR optimization objective is to achieve maximum correlation and minimum redundancy.

[0057] The process of step 3 is as follows: Step 3.1: Calculate mutual information: .

[0058] in, This represents the first mutual information value between the i-th candidate key feature and the corresponding true value of remaining lifetime; Let represent the joint probability distribution function between the i-th candidate key feature and the corresponding true value of remaining lifetime; The marginal probability distribution function represents the feature of the i-th candidate key. The marginal probability distribution function represents the true value of remaining lifespan y.

[0059] Step 3.2: Calculate the mRMR value: .

[0060] in, The second mutual information value between the i-th candidate key feature and the j-th candidate key feature is represented; S represents the selected feature set, and m is the number of features contained in the current feature set.

[0061] Step 3.3: Use a greedy algorithm to select 8 key features and save the model.

[0062] The process of using a greedy algorithm to select key features is as follows: (1) Initialize the candidate feature set F, and calculate the first mutual information value between each candidate feature and the output variable (remaining life of rotating mechanical parts). .

[0063] (2) Select the feature with the largest mutual information value and add it to the feature set S.

[0064] (3) Features not selected The average redundancy between the selected feature and other features in the current feature set is calculated, and a comprehensive evaluation index is constructed. The calculation process is as follows: Candidate features The average redundancy is defined as: .

[0065] Based on this, a comprehensive evaluation index for candidate features is constructed: .

[0066] in, Representing candidate features Compared with the actual remaining lifespan The first mutual information value between them; Representing candidate features With selected features The second mutual information value between them; S represents the currently selected feature set; This represents the number of features in set S.

[0067] (4) Choose The largest feature is added to set S.

[0068] (5) Repeat the above steps until the evaluation index is stable or the number of features meets the optimal dimension.

[0069] Finally, after a greedy search combining mutual information and the mRMR index, the following eight key features were selected: standard deviation_vertical direction (reflecting vibration intensity changes), root mean square_vertical direction (characterizing energy level), crest factor_horizontal direction (reflecting impact characteristics), maximum value_vertical direction (peak load response), standard deviation_horizontal direction (lateral stability index), root mean square_horizontal direction (lateral energy change), temperature (thermal effect and lubrication state), and rotational speed (influence of dynamic load changes).

[0070] Step 4: Feature Subset Training and Validation Evaluation. Feature selection is performed using the labeled dataset. Mutual information values ​​are calculated and sorted. A greedy algorithm is used for optimization, with mRMR loss as the loss function. After training the model, a selected feature set is obtained. The correlation is evaluated using the validation set, and the correlation coefficient is calculated. The iteration stops after a preset number of iterations.

[0071] After obtaining the candidate feature subset using the mRMR greedy algorithm in step 3, the feature subset is trained and validated based on the labeled dataset to determine the final feature set used for lifetime prediction. The specific process of step 4 is as follows: Step 4.1: Data Loading. Use data processing tools to load the labeled dataset, where each sample contains a feature vector and its corresponding remaining lifetime label.

[0072] Step 4.2: Model Training and Feature Subset Update. Based on the candidate feature subsets obtained in Step 3, the prediction model is trained, and the feature subsets are updated progressively during the training process to evaluate the impact of different feature combinations on prediction performance.

[0073] Step 4.3: Correlation Evaluation and Stopping Criterion. Calculate the correlation coefficient between the predicted results and the actual remaining lifetime on the validation set. This coefficient is used to evaluate the effectiveness of the current feature subset. The calculation formula is as follows: .

[0074] Where CC represents the correlation coefficient, which measures the degree of linear correlation between the predicted result and the actual value; N represents the sample size. Indicates the first The true remaining lifetime value of each sample; Indicates the first Predicted remaining lifetime values ​​for each sample; This represents the mean of the actual remaining lifespan sample; This represents the mean of the predicted remaining lifespan sample; The standard deviation of the true remaining lifetime sample; This represents the standard deviation of the predicted remaining lifespan sample.

[0075] When the correlation coefficient no longer significantly improves with feature subset updates, or when the preset number of iterations is reached, the feature update process stops, and the final feature subset is output.

[0076] Step 5: Export the filtered feature set as a CSV (Comma-Separated Values) file.

[0077] Step 6: Construct a working condition classification model: Calculate the steady-state index γ using multi-scale wavelet transform, classify the data into steady-state and unsteady-state working conditions, and store the corresponding data. In this embodiment, the steady-state index γ is calculated based on wavelet transform, and the formula is: .

[0078] .

[0079] in, This represents the first-order wavelet transform; Represents the preset key features at time t; This represents the result of performing a first-order wavelet transform on the preset key features at time t. Describe the wavelet basis function at time t and scale s; express and Perform convolution operations; Indicates the second-order wavelet transform; This represents the result of performing a second-order wavelet transform on the preset key features at time t; Let represent the square of the wavelet basis function at time t and scale s; express and Perform convolution operations.

[0080] like If , then the steady-state index of the target rotating mechanical component at time t is 0; where, This represents the first preset discrimination threshold.

[0081] like ,and Then the steady-state index of the target rotating mechanical component at time t is 1; where, This indicates the second preset discrimination threshold.

[0082] If neither of the above two conditions is met, the steady-state index of the target rotating mechanical component at time t is calculated using the following formula: .

[0083] in, The steady-state exponent is represented at time t; The smoothing function representing the scale s at time t; The weighting coefficients of the first-order wavelet transform are used to describe the degree of gradation in the signal. This represents the weighting coefficients of the second-order wavelet transform.

[0084] The current steady-state index is obtained by summing and averaging the steady-state indices at all times within the current time period. To reflect the overall operating state of the target rotating mechanical component within the current time period, the steady-state indices at all times within this time period are averaged over time, and the resulting average value is used as the current steady-state index for subsequent operating condition determination. Using an average value can comprehensively reflect the steady-state level over a period of time, avoiding excessive influence of instantaneous impacts on operating condition judgment, and is more in line with engineering application scenarios.

[0085] Based on the comprehensive time-domain γ value (current steady-state index), the operating condition is determined according to the threshold range: γ≥0.8: steady-state operating condition (normal state of rotating mechanical parts); 0.3<γ<0.8: weakly unsteady operating condition (minor fault state); γ≤0.3: strong unsteady operating condition (severe fault state).

[0086] like Figure 6 The diagram illustrates a flowchart of a life prediction method for rotating mechanical components. By classifying the state of the rotating mechanical component into normal, minor fault, and severe fault states, and establishing corresponding fault prediction models for each, the method can more accurately predict faults under various operating conditions. The flowchart of the life prediction model for rotating mechanical components categorizes the state of the rotating mechanical component into normal, minor fault, and severe fault states. Among these, steady-state operating conditions correspond to the normal state, characterized by stable signals and operation; weakly unsteady operating conditions correspond to the minor fault state, exhibiting slight fluctuations and wear; and strongly unsteady operating conditions correspond to the severe fault state, characterized by significant non-stationarity and fault deterioration. Establishing steady-state and unsteady-state fault prediction models for different operating conditions can more accurately reflect the life evolution of rotating mechanical components under various operating states.

[0087] Step 7: Combine the filtered feature set with an interval type II fuzzy system (IT2 FS) and a genetic algorithm to write a prediction program and deploy it to a cloud platform.

[0088] In this example, step 7 is as follows: Step 7.1: Build the IT2 FS network using Python (e.g., Figure 4 (As shown).

[0089] Step 7.2: Develop the Flask API to receive edge data.

[0090] Step 7.3: Process: Feature input → Working condition division → IT2 inference → GA optimization → Output prediction.

[0091] Step 8: Perform multi-scale wavelet transform on the input samples to divide the data into different scenarios. Construct T1 FS prediction models for steady-state data and IT2 FS prediction models for non-steady-state data. The IT2 FS prediction models optimize fuzzy parameters using a genetic algorithm, selecting the model with the smallest root mean square error (RMSE) as the final prediction result. Figure 3 As shown, a structural block diagram of a fuzzy system model is provided, illustrating the basic components of the fuzzy inference system in this application, including fuzzification of the input signal, a fuzzy rule base, an inference engine, and a defuzzification module, which are used to realize the mapping process from precise input to precise output.

[0092] like Figure 4 As shown, a structural block diagram of an interval type-II fuzzy system model is provided, illustrating the fuzzy logic unit, inference engine, and type reduction (output). And a defuzzer, to handle uncertainty mechanisms. This represents the final precise output value obtained after type reduction and defuzzification processing of the interval type-II fuzzy system; This represents the left endpoint value of the output interval obtained during the type reduction process; This represents the right endpoint value of the output interval obtained during the type reduction process.

[0093] In this example, step 8 is as follows: Step 8.1: Divide the working conditions and calculate γ. If γ is close to 1, it is a steady state; (1) Steady-state data modeling: Steady-state data corresponds to the normal operating state of rotating mechanical parts, with small signal fluctuations and stable statistical characteristics. For this type of data, a type fuzzy system (T1-FS) model is adopted. The membership function of the input features is determined by the fuzzy clustering algorithm (Fuzzy C-Means, FCM), and the linear parameters are identified by the least squares method to establish the mapping relationship between the input features and the remaining life (RUL) of rotating mechanical parts. This model has low computational complexity and is suitable for life prediction under steady-state working conditions.

[0094] Step 8.2: Modeling Unsteady-State Data: Unsteady-state data corresponds to minor or severe fault states of rotating mechanical components, exhibiting significant signal fluctuations and substantial uncertainty. For this type of data, an improved Interval Type II Fuzzy System (IT2-FS) model is employed. This model introduces upper and lower boundaries into the fuzzy membership degrees, forming an uncertainty envelope interval to characterize the fuzziness caused by changes in operating conditions. A Genetic Algorithm (GA) is used to globally optimize the fuzzy rule parameters and membership function parameters to improve the model's prediction accuracy and robustness.

[0095] GA is used to optimize the following parameters: the center parameter value m of the input membership function, the left and right widths δ1 and δ2; and the output parameters of each fuzzy rule in the interval type-II fuzzy system. ; The upper and lower boundary parameters of the activation interval.

[0096] like Figure 8 As shown, the genetic algorithm first randomly generates a population of chromosomes. Each chromosome encodes a candidate solution to the optimization problem. Then, the fitness of all individuals relative to the optimization task is evaluated using a fitness function. According to Darwin's principle, highly healthy individuals are more likely to be selected to reproduce. Genetic operators, such as crossover and mutation, are applied to the parents to generate a new generation of candidate solutions. The optimization steps are as follows: (1) Chromosome Encoding: The above fuzzy parameters are encoded into chromosomes to form the initial population. A schematic diagram of chromosome encoding is shown below. Figure 9 As shown, A 1，1 A represents the antecedent fuzzy parameter corresponding to the first input variable in the first fuzzy rule. 1，C In the first fuzzy rule, A represents the antecedent fuzzy parameter corresponding to the C-th input variable; C represents the total number of input variables; n，1 A represents the set of antecedent fuzzy parameters for n fuzzy rules on the first input variable, where n represents the total number of fuzzy rules; n，C This represents the set of antecedent fuzzy parameters for n fuzzy rules on the Cth input variable; This represents the consequent parameter corresponding to the Cth fuzzy rule, which is used to determine the output value of the fuzzy rule. This represents the consequent parameter corresponding to the Cth fuzzy rule, used to determine the output value of the fuzzy rule.

[0097] (2) Fitness calculation: The prediction error RMSE is used as the fitness function to evaluate each individual.

[0098] (3) Selection operation: Select individuals with high fitness from the current population to enter the next generation based on fitness.

[0099] (4) Cross operation: Cross the selected individuals to generate new candidate parameter combinations.

[0100] (5) Mutation operation: Perturb some genes to increase population diversity.

[0101] (6) Termination judgment: If the set number of iterations is reached or RMSE convergence is achieved, the iteration is stopped and the optimal fuzzy parameters are output.

[0102] The formation process of activation intervals and type-reduced output intervals, and Figure 5 The structure of the interval type II fuzzy system shown is consistent with the rule activation and type reduction process.

[0103] The fuzzy set is initialized using FCM, and the activation range of each fuzzy rule is calculated: .

[0104] in, Indicates the first The activation range corresponding to the fuzzy rule; This represents the lower bound of the activation interval (Lower MF). This represents the upper bound of the activation interval (Upper MF).

[0105] Based on this, type reduction is performed to obtain the output range. : .

[0106] Output: .

[0107] GA optimizes membership function parameters and fitness: .

[0108] like Figure 5 As shown, a schematic diagram of an interval type II fuzzy set is provided, where the horizontal axis represents the range of values ​​of the input variable, and the vertical axis represents the membership degree of the membership function. The membership degree is a dimensionless quantity with a value range of [0, 1]. The membership degree of the IT2 fuzzy set is an interval (footprint of uncertainty, FOU) which is enclosed by the upper and lower membership functions.

[0109] A schematic diagram of initializing a fuzzy set for fuzzy clustering, as shown below. Figure 7 As shown, the fuzzy C-means (FCM) algorithm is used to cluster the input feature space, thereby initializing the fuzzy membership function of each input variable. This includes clustering the input feature space for each input variable X. i Generate its corresponding fuzzy membership function A i,1 A i,2 A i,3The cluster centers C1, C2, and C3 required for the antecedents of the fuzzy rules are obtained based on the clustering results. The initialized fuzzy sets serve as the basic parameter inputs for the inference process of the IT2 fuzzy system.

[0110] Step 9: Start the edge acquisition device, calculate the running features in real time and upload them to the cloud platform to realize RUL prediction of rotating mechanical parts.

[0111] The technical concept of this application is as follows: to reduce the computational burden by jointly optimizing feature selection through mutual information and expert prior knowledge; to dynamically divide the working conditions using multi-scale wavelet transform; to use T1 FS to process steady-state data and IT2 FS to process data uncertainty, and to combine genetic algorithm to optimize parameters, thereby achieving prediction that is both lightweight and highly accurate.

[0112] The beneficial effects of this application are: significantly improving the accuracy and robustness of RUL prediction, reducing computational complexity, adapting to complex industrial conditions, and supporting intelligent maintenance decision-making. This is mainly reflected in: through multi-stage collaboration, leveraging information theory's ability to filter features, utilizing wavelet transform for dynamic division of operating conditions, and employing fuzzy systems to handle uncertainties, ultimately achieving high-precision life prediction for lightweight rotating mechanical components.

[0113] Based on the same inventive concept, this application also provides a rotating mechanical component life prediction device for implementing the above-mentioned rotating mechanical component life prediction method. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations in one or more embodiments of the rotating mechanical component life prediction device provided below can be found in the limitations of the rotating mechanical component life prediction method described above, and will not be repeated here.

[0114] In one exemplary embodiment, such as Figure 10 As shown, a life prediction device for rotating mechanical parts is provided, comprising: The data acquisition module 301 is used to acquire the operation monitoring signals of the target rotating mechanical component within the current time period; the operation monitoring signals include vibration signals, temperature signals and rotation speed signals.

[0115] The feature extraction module 302 is used to extract preset key features based on the operation monitoring signals within the current time period.

[0116] The working condition analysis module 303 is used to perform working condition stability analysis based on preset key features and through multi-scale wavelet transform method to obtain the current steady-state index of the target rotating mechanical component.

[0117] The status determination module 304 is used to determine the health status of the target rotating mechanical component in the current time period based on the current steady-state index; the health status includes normal status, minor fault status and severe fault status.

[0118] The steady-state prediction module 305 is used to input preset key features into the trained steady-state working condition life prediction model if the health state is normal, so as to obtain the remaining life prediction value of the target rotating mechanical component in the current time period. The steady-state working condition life prediction model is obtained by iteratively training a preset first machine learning model using a first training dataset. The first training dataset is constructed based on preset key features and corresponding real values ​​of remaining life in historical time periods under normal working conditions.

[0119] The non-steady-state prediction module 306 is used to obtain the remaining life prediction value of the target rotating mechanical component in the current time period by training a non-steady-state working condition life prediction model with preset key features if the health state is a minor fault state or a severe fault state. The non-steady-state working condition life prediction model is obtained by iteratively training a preset second machine learning model using a second training dataset. The second training dataset is constructed based on preset key features and corresponding real values ​​of remaining life in historical time periods under minor and severe fault states.

[0120] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 11 As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs in the non-volatile storage media. The database stores data for predicting the lifespan of rotating mechanical components. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network. When the computer program is executed by the processor, it implements a method for predicting the lifespan of rotating mechanical components.

[0121] Those skilled in the art will understand that Figure 11The structures shown are merely block diagrams of some structures related to the present application and do not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than shown in the figures, or combine certain components, or have different component arrangements. In an exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0122] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0123] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0124] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0125] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0126] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0127] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0128] The technical solution of this application uses a rolling and rotating mechanical component as a specific embodiment for illustration, aiming to clearly and completely explain the technical principles and implementation process of the invention. However, the life prediction method proposed in this application is based on feature extraction of the operating monitoring signal of the rotating component, multi-scale wavelet steady-state analysis, and fuzzy prediction model construction. Its core idea does not depend on the specific object of the rotating mechanical component.

[0129] Therefore, the method of this application is also applicable to rotating mechanical components with similar operating characteristics and fault evolution patterns, such as gears, shafts, and motor rotors. In these application scenarios, life prediction can be achieved simply by acquiring the operating monitoring signals of the corresponding rotating components and processing them according to the steps described in this application, and the technical principles and processing procedures remain consistent.

[0130] It should be noted that the use of rotating mechanical components as the object of protection in the claims of this application is based on the selection of specific embodiments and does not constitute a substantial limitation on the scope of application of the technical solution of this invention.

[0131] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for predicting the lifespan of rotating mechanical components, characterized in that, The life prediction method for rotating mechanical components includes: Acquire the operational monitoring signals of the target rotating mechanical component within the current time period; the operational monitoring signals include vibration signals, temperature signals, and rotational speed signals; Based on the operational monitoring signals within the current time period, extract preset key features; Based on preset key features, the current steady-state index of the target rotating mechanical component is obtained by performing a multi-scale wavelet transform method to analyze the stability of the working condition. Based on the current steady-state index, the health status of the target rotating mechanical component is determined for the current time period; the health status includes normal status, minor fault status, and severe fault status. If the health status is normal, the preset key features are input into the trained steady-state working condition life prediction model to obtain the remaining life prediction value of the target rotating mechanical component in the current time period. The steady-state working condition life prediction model is obtained by iteratively training the preset first machine learning model using the first training dataset. The first training dataset is constructed based on the preset key features and corresponding real values ​​of remaining life in the historical time period under normal working conditions. If the health status is a minor fault state or a severe fault state, the unsteady-state operating condition life prediction model trained with preset key features is used to obtain the remaining life prediction value of the target rotating mechanical component in the current time period. The unsteady-state operating condition life prediction model is obtained by iteratively training a preset second machine learning model using a second training dataset. The second training dataset is constructed based on preset key features and corresponding real values ​​of remaining life in historical time periods under minor and severe fault states.

2. The life prediction method for rotating mechanical components according to claim 1, characterized in that, The method for determining the preset key features specifically includes: Acquire the operational monitoring signals and corresponding actual remaining life values ​​of multiple rotating mechanical components within a historical time period; Based on the operation monitoring signals of multiple rotating mechanical components within a historical time period, a candidate feature set is obtained by performing time-domain and frequency-domain feature calculations through feature extraction methods. The candidate feature set includes the maximum value, mean, standard deviation, root mean square, root mean square shape factor, skewness, kurtosis, crest factor, latitude factor, impulse factor, temperature value, and rotational speed value of the vibration signal in the vertical and horizontal directions, respectively. The first mutual information value between each candidate feature in the candidate feature set and its corresponding true remaining lifetime value, and the second mutual information value between any two candidate features in the candidate feature set are calculated using the mutual information algorithm. A greedy search strategy is adopted to maximize the mRMR value as the optimization objective. Features are selected from the candidate feature set, and preset key features are obtained when preset stopping conditions are met. The mRMR value is calculated based on the first mutual information value and the second mutual information value. The preset stopping conditions include reaching a first preset maximum number of iterations or the change in mRMR value within a consecutive first preset number of iterations being less than a preset change threshold. The preset key features include the standard deviation, root mean square, and maximum value of the vibration signal in the vertical direction, the standard deviation, root mean square, and crest factor of the vibration signal in the horizontal direction, the temperature value, and the rotational speed value.

3. The life prediction method for rotating mechanical components according to claim 1, characterized in that, Based on preset key features, a multi-scale wavelet transform method is used to perform operational stability analysis, obtaining the current steady-state index of the target rotating mechanical component, specifically including: Based on preset key features, multi-scale wavelet transform is performed using the following formula: ； ； in, This represents the first-order wavelet transform; Represents the preset key features at time t; This represents the result of performing a first-order wavelet transform on the preset key features at time t. Describe the wavelet basis function at time t and scale s; express and Perform convolution operations; Indicates the second-order wavelet transform; This represents the result of performing a second-order wavelet transform on the preset key features at time t; Let represent the square of the wavelet basis function at time t and scale s; express and Perform convolution operations; like If , then the steady-state index of the target rotating mechanical component at time t is 0; where, This represents the first preset discrimination threshold; like ,and Then the steady-state index of the target rotating mechanical component at time t is 1; where, This indicates the second preset discrimination threshold; If neither of the above two conditions is met, the steady-state index of the target rotating mechanical component at time t is calculated using the following formula: ； in, The steady-state exponent is represented at time t; The smoothing function representing the scale s at time t; Represents the weighting coefficients of the first-order wavelet transform; Represents the weighting coefficients of the second-order wavelet transform; The current steady-state index is obtained by summing and averaging the steady-state indices at all times within the current time period.

4. The life prediction method for rotating mechanical components according to claim 3, characterized in that, Based on the current steady-state index, the health status of the target rotating mechanical component in the current time period is determined, specifically including: If the current steady-state index is greater than or equal to the first preset steady-state threshold, then the health state is normal. If the current steady-state index is less than the first preset steady-state threshold and greater than or equal to the second preset steady-state threshold, then the health state is a mild fault state. If the current steady-state index is less than the second preset steady-state threshold, the health status is a severe fault state.

5. The life prediction method for rotating mechanical components according to claim 1, characterized in that, The training process of the steady-state operating condition life prediction model specifically includes: Based on the first training dataset, each preset key feature is clustered using a fuzzy clustering algorithm to obtain the cluster center value of each preset key feature; Based on the cluster center value of each preset key feature, a membership function of a preset type is constructed for each preset key feature, and the center parameter of the membership function of each preset key feature is initialized to the corresponding cluster center value; the preset type includes Gaussian type or triangular type. Based on the first training dataset and the membership function initialized for each preset key feature, the linear parameters of the preset first machine learning model are identified by the least squares method to obtain the initial linear parameters of the steady-state operating condition life prediction model; the first machine learning model is a type of fuzzy system model. With the goal of minimizing the root mean square error between the predicted remaining lifetime output by the first machine learning model and the corresponding actual remaining lifetime, the central parameters of the membership function of each preset key feature and the linear parameters of the steady-state operating condition life prediction model are iteratively optimized until the second preset maximum number of iterations is reached or the change in root mean square error within the second preset number of iterations is less than the preset root mean square error change threshold, thus obtaining the trained steady-state operating condition life prediction model.

6. The life prediction method for rotating mechanical components according to claim 1, characterized in that, The training process of the unsteady-state life prediction model specifically includes: Based on the second training dataset, each preset key feature is clustered using a fuzzy clustering algorithm to obtain the cluster center value of each preset key feature; Based on the cluster center value of each preset key feature, an interval type II membership function is constructed for each preset key feature, and the center parameter of the interval type II membership function of each preset key feature is initialized to the corresponding cluster center value; the interval type II membership function includes an uncertainty interval formed by the lower membership function and the upper membership function. Based on the second training dataset and the interval type II membership function initialized for each preset key feature, a genetic algorithm is used to globally optimize the parameters of the second machine learning model until a third preset maximum number of iterations is reached or the change in fitness function value within a consecutive third preset number of iterations is less than a preset threshold for the change in fitness function value, thus obtaining a trained unsteady-state operating condition life prediction model; the fitness function value is the negative of the root mean square error between the predicted remaining life value output by the second machine learning model and the corresponding true remaining life value; the second machine learning model is an interval type II fuzzy system model.

7. A life prediction device for rotating mechanical parts, characterized in that, The life prediction of the rotating mechanical component uses the life prediction method for rotating mechanical components according to any one of claims 1-6, wherein the life prediction device for the rotating mechanical component includes: The data acquisition module is used to acquire the operation monitoring signals of the target rotating mechanical component within the current time period; the operation monitoring signals include vibration signals, temperature signals, and rotational speed signals; The feature extraction module is used to extract preset key features based on the operation monitoring signals within the current time period; The working condition analysis module is used to perform working condition stability analysis based on preset key features and through multi-scale wavelet transform method to obtain the current steady-state index of the target rotating mechanical component. The status determination module is used to determine the health status of the target rotating mechanical component in the current time period based on the current steady-state index; the health status includes normal status, minor fault status and severe fault status; The steady-state prediction module is used to input preset key features into the trained steady-state working condition life prediction model if the health state is normal, so as to obtain the remaining life prediction value of the target rotating mechanical component in the current time period. The steady-state working condition life prediction model is obtained by iteratively training a preset first machine learning model using a first training dataset. The first training dataset is constructed based on preset key features and corresponding real values ​​of remaining life in historical time periods under normal working conditions. The non-steady-state prediction module is used to obtain the remaining life prediction value of the target rotating mechanical component in the current time period by using a non-steady-state working condition life prediction model trained with preset key features if the health state is a minor fault state or a severe fault state. The non-steady-state working condition life prediction model is obtained by iteratively training a preset second machine learning model using a second training dataset. The second training dataset is constructed based on preset key features and corresponding real values ​​of remaining life in historical time periods under minor and severe fault states.

8. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the life prediction method for a rotating mechanical component according to any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the life prediction method for rotating mechanical components as described in any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the life prediction method for rotating mechanical components as described in any one of claims 1-6.

Images