Equipment state management-oriented data optimization method, intelligent diagnosis method and system

Abstract

The invention discloses an equipment state management-oriented data optimization method, an intelligent diagnosis method and an intelligent diagnosis system, and belongs to the field of equipment state monitoring. The existing equipment state sampling technology has the problems of data redundancy, anomaly detection lag and lack of real data. The method comprises the following steps: constructing a state adaptive monitoring sampling framework, and dynamically adjusting a sample interval; based on state sequence conversion and binarization technologies, performing three-dimensional compression by fusing short-time Fourier transform, a hidden Markov model and a compression method of a failure mechanism; a mixed compression strategy of a cross-Nyquist space and a sub-Nyquist space is constructed, and data properties are maintained through a local permutation orthogonal matrix. Amplitude difference between simulation data and real data is eliminated based on state sequence conversion of spectrum clustering, and a migration diagnosis model is constructed in combination with a spiking neural network and a domain adaptation technology; and forming an anomaly detection method fused with security domain state evaluation. According to the invention, the data acquisition and processing process is optimized, and the accuracy and timeliness of fault diagnosis are improved.

Description

Technical Field

[0001] This invention relates to a data optimization method, diagnostic method, and system for equipment operating status, and particularly to a data optimization method, intelligent diagnostic method, and system for equipment status management. Background Technology

[0002] During the operation of industrial equipment, issues such as redundancy, missing data, and inconsistent distribution of condition monitoring data (e.g., vibration data) severely impact the effectiveness of Predictive Fault and Health Management (PHM). Traditional methods have shortcomings in sampling strategies, data compression, anomaly detection, and fault diagnosis. For example, fixed-time sampling leads to sample redundancy or missing data; single-dimensional data compression struggles to balance compression ratio and feature retention; the "black box" nature of health phase condition assessment results in highly unpredictable anomaly detection; and the scarcity of real fault data leads to insufficient generalization ability of diagnostic models. Therefore, a systematic approach covering the entire lifecycle is urgently needed to optimize data acquisition, processing, and diagnostic processes, thereby improving the reliability and efficiency of equipment health management. Summary of the Invention

[0003] The purpose of this invention is to address the problems of data redundancy, delayed anomaly detection, and lack of real data in existing equipment condition sampling technologies, and to propose a data optimization method, intelligent diagnostic method, and system for equipment condition management.

[0004] The above objectives are achieved through the following technical solutions: A data optimization method for equipment status management, the method comprising the following: The adaptive sample interval dynamic adjustment method for slowly degrading processes refers to constructing a state-adaptive monitoring and sampling framework and dynamically adjusting the sample interval using the degradation state prediction results. The vibration data compression method in the Nyquist sampling space refers to a compression method based on state sequence transformation and binarization technology, which integrates short-time Fourier transform, hidden Markov model and time-frequency domain fusion compression to compress the fault mechanism and perform three-dimensional data compression of the sample. Cross-sampling space compression refers to a hybrid compression strategy that combines clustering binarization, discrete wavelet transform filtering, and compressed sensing to construct a cross-Nyquist and sub-Nyquist space, and preserves data properties by locally permuting orthogonal matrices.

[0005] Furthermore, the method for constructing a state-adaptive monitoring sampling framework and dynamically adjusting the sample interval using the degradation state prediction results includes the following: The steps include: adding new sampling target hyperparameters; applying piecewise cubic Hermite interpolation; using the differential autoregressive moving average model; variable interchange; constructing a state-adaptive monitoring sampling framework, converting the time series into a degenerate series through variable interchange, and combining the prediction model of the differential autoregressive moving average model to achieve quantitative adjustment of the sample interval and eliminate sampling bias.

[0006] Furthermore, in the vibration data compression method in the Nyquist sampling space, the three-dimensional compression of the sample data refers to the compression of the sample in the length dimension based on time-frequency transformation, the compression in the height dimension based on the time-frequency distribution state, and the compression in the width dimension based on wavelet compression downsampling.

[0007] An intelligent diagnostic method for equipment condition management includes the following: State sequence transformation based on spectral clustering eliminates the amplitude difference between simulation and real data, and a transfer diagnostic model is constructed by combining spiking neural networks and domain adaptation techniques. Common feature extraction is enhanced by leakage integral discharge model, forming an anomaly detection method that integrates safety domain state assessment.

[0008] Furthermore, the anomaly detection method for the integrated security domain state assessment specifically includes: Steps for introducing auxiliary health status information; The dual-task learning steps based on the univariate support vector data description model unify anomaly detection and health status assessment within the security domain into the support vector data description framework; and monotonicity and discriminative indices are introduced by optimizing the hypersphere center and boundary.

[0009] Furthermore, the construction of the transfer diagnostic model includes: The steps to construct a dataset consisting of virtual and real data; Steps for state sequence transformation based on spectrum clustering; The steps for optimizing feature learning of compressed data include introducing an integral discharge model for leakage. Steps for constructing a transfer diagnostic model based on a spiking neural network.

[0010] A system for a data optimization method and an intelligent diagnostic method for equipment condition management is provided. The system comprises: The data acquisition module is used to collect equipment status monitoring data through sensors and to optimize the sampling time and density using a dynamic sampling strategy. The data optimization module uses data optimization methods oriented towards equipment health management to process the collected data; The status monitoring module uses optimized data to monitor the health status of equipment in real time. The fault diagnosis module uses intelligent diagnostic methods for equipment health management to diagnose abnormal conditions. The decision support module provides maintenance decision support based on monitoring and diagnostic results.

[0011] The beneficial effects of this invention are as follows: This invention provides a comprehensive, efficient, and intelligent solution for equipment health management through innovative data optimization strategies and intelligent diagnostic methods. By optimizing data acquisition and processing, and improving the accuracy and timeliness of fault diagnosis, this invention helps advance industrial equipment operation and maintenance technology, meeting the modern industrial demands for high reliability, high safety, and high efficiency in equipment. Detailed Implementation

[0012] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Specific implementation method one:

[0014] This embodiment provides a data optimization method for equipment status management, the method comprising the following: The adaptive sample interval dynamic adjustment method for slow degradation processes refers to constructing a state-adaptive monitoring and sampling framework, dynamically adjusting the sample interval using degradation state prediction results, optimizing the sampling process, reducing sample redundancy and missing data, and achieving adaptive monitoring of slow degradation processes. The vibration data compression method in the Nyquist sampling space refers to a compression method based on state sequence transformation and binarization techniques, which integrates short-time Fourier transform (STFT), hidden Markov model (HMM) and time-frequency domain fusion compression to compress the fault mechanism. It performs three-dimensional compression of the sample data, which significantly improves the compression ratio while preserving fault characteristics, and reduces data storage and transmission costs. Cross-sampling space compression methods refer to the construction of a hybrid compression strategy across Nyquist and sub-Nyquist spaces by combining clustering binarization, discrete wavelet transform filtering, and compressed sensing (CS). By preserving data properties through local permutation orthogonal matrix (PPO), the compression ratio is further improved and the data is encrypted. Specific Implementation Method Two:

[0016] This embodiment of the data optimization method for equipment condition management differs from Specific Embodiment 1 in that the method for constructing a condition adaptive monitoring sampling framework and dynamically adjusting the sample interval using degradation condition prediction results includes the following: To address the issue of missing sampling adjustment targets, a new sampling target hyperparameter is added; piecewise cubic Hermite interpolation is applied to solve the problem of predicting irregular time series; the steps of the differential autoregressive moving average model are adopted to address the issue of sampling adjustment not considering the trend of state changes; variable interchange is used to address the problem that sample intervals cannot be directly predicted; a state-adaptive monitoring sampling framework is constructed, which converts time series into degenerate series through variable interchange, and combines the prediction model of the differential autoregressive moving average model to achieve quantitative adjustment of sample intervals, eliminate sampling bias, and improve the targeting and effectiveness of data collection. Specific implementation method three:

[0018] This embodiment of the data optimization method for equipment status management differs from specific embodiments one or two in that the vibration data compression method in the Nyquist sampling space involves three-dimensional compression of the sample data, which means compressing the sample in the length dimension (i.e., number of points) based on time-frequency conversion, in the height dimension (i.e., frequency components) based on the time-frequency distribution state, and in the width dimension (i.e., sample point size) based on wavelet compression downsampling. Specific implementation method four:

[0020] This embodiment of an intelligent diagnostic method for equipment status management includes the following: To address the scarcity of real-world data, a state sequence transformation based on spectral clustering eliminates the amplitude difference between simulation and real data. A transfer diagnostic model is constructed by combining spiking neural networks (SNN) and domain adaptation techniques to achieve a virtual-to-real-world transfer fault diagnosis method, thus improving fault diagnosis performance. Furthermore, a common feature extraction is enhanced through a leakage integral discharge model (LIF), forming an anomaly detection method that integrates safety domain state assessment. The safety domain solution process is reconstructed, unifying anomaly detection and health status assessment within the same framework. This is achieved using a dual-task learning method, reducing the distribution difference between the source and target domains and enabling the transfer of virtual data knowledge to real-world scenarios. Specific implementation method five:

[0022] This embodiment of the intelligent diagnostic method for equipment status management differs from Specific Embodiment Four in that the anomaly detection method integrating security domain status assessment specifically includes: The steps of introducing auxiliary health status information optimize the health status assessment within the security domain; The dual-task learning steps based on the univariate support vector data description model unify anomaly detection and health status assessment within the safety domain (HSESR) within the support vector data description (SVDD) framework; by optimizing the hypersphere center and boundary, monotonicity and discriminative indices are introduced to achieve quantitative assessment of health stage status and early detection of anomalies. Specific implementation method six:

[0024] This embodiment of an intelligent diagnostic method for equipment status management differs from specific embodiments four or five in that the construction of a migration diagnostic model to realize a virtual-physical migration fault diagnosis method includes: The steps to construct a dataset consisting of virtual and real data; The state sequence transformation based on spectrum clustering is used to reduce the difference in feature distribution between simulation and real data; The steps for optimizing feature learning of compressed data include introducing an integral discharge model for leakage. Steps for constructing a transfer diagnostic model based on a spiking neural network. Specific implementation method seven:

[0026] This embodiment provides a system for a data optimization method and an intelligent diagnostic method for equipment status management. The system is used in both the data optimization method and the intelligent diagnostic method for equipment status management. The system includes: The data acquisition module is used to collect equipment status monitoring data (such as equipment vibration data) through sensors, and to optimize the sampling time and density using a dynamic sampling strategy; The data optimization module uses data optimization methods oriented towards equipment health management to process the collected data; The status monitoring module uses optimized data to monitor the health status of equipment in real time. The fault diagnosis module uses intelligent diagnostic methods for equipment health management to diagnose abnormal conditions. The decision support module provides maintenance decision support based on monitoring and diagnostic results.

[0027] The embodiments disclosed herein are preferred embodiments, but are not limited thereto. Those skilled in the art can readily grasp the spirit of the present invention based on the above embodiments and make different extensions and variations, but as long as they do not depart from the spirit of the present invention, they are all within the protection scope of the present invention.

Claims

1. A data optimization method for equipment state management, characterized in that: The method comprises the following contents: The adaptive sample interval dynamic adjustment method for the slowly changing degradation process is to construct a state adaptive monitoring sampling framework, and dynamically adjust the sample interval by using the degradation state prediction result; The vibration data compression method in the Nyquist sampling space is a compression method that fuses the short-time Fourier transform, hidden Markov model and time-frequency domain fusion compression of the fault mechanism based on state sequence conversion and binarization technology, and performs three-dimensional compression of the sample data; The cross-sampling space compression method is to combine clustering binarization, discrete wavelet transform filtering and compressed sensing to construct a hybrid compression strategy across the Nyquist and sub-Nyquist spaces, and maintain the data properties through local permutation orthogonal matrices.

2. The data optimization method for equipment state management according to claim 1, characterized in that, The method of constructing a state adaptive monitoring sampling framework and dynamically adjusting the sample interval by using the degradation state prediction result comprises the following contents: The new sampling target hyperparameter is added; the piecewise cubic Hermite interpolation is applied; the steps of adopting the difference autoregressive moving average model; the variable exchange is counted; the state adaptive monitoring sampling framework is constructed, the time series is converted into a degradation sequence through variable exchange, the quantitative adjustment of the sample interval is realized by combining the prediction model of the difference autoregressive moving average model, and the sampling deviation is eliminated.

3. The data optimization method for equipment state management according to claim 1 or 2, characterized in that, In the vibration data compression method in the Nyquist sampling space, the three-dimensional compression of the sample data is to perform length dimension compression based on time-frequency conversion, height dimension compression based on time-frequency distribution state, and width dimension compression based on wavelet compression downsampling.

4. An intelligent diagnostic method oriented to equipment state management, characterized in that, It comprises the following contents: The state sequence conversion based on spectral clustering eliminates the amplitude difference between simulation and real data, a transfer diagnosis model is constructed by combining the pulse neural network and domain adaptation technology, common feature extraction is enhanced through the leakage integral discharge model, and an abnormality detection method that fuses safety domain state evaluation is formed.

5. The intelligent diagnostic method for equipment state management according to claim 4, characterized in that, The abnormality detection method that fuses safety domain state evaluation specifically comprises: The step of introducing health state auxiliary information; The step of double-task learning based on the single-variable support vector data description model, which unifies abnormality detection and health state evaluation in the safety domain in the support vector data description framework; and the step of introducing monotonicity and discriminability indexes by optimizing the center and boundary of the hypersphere.

6. The intelligent diagnostic method for equipment state management according to claim 4 or 5, characterized in that, The construction of the transfer diagnosis model comprises: The step of constructing a data set composed of virtual data and real data; The step of state sequence conversion based on spectral clustering; The step of introducing the leakage integral discharge model to optimize feature learning of compressed data; The step of constructing a transfer diagnosis model based on the pulse neural network.

7. A system for equipment state management-oriented data optimization method and intelligent diagnosis method, characterized in that, The system is used for the data optimization method for equipment state management in any one of claims 1-3, and the intelligent diagnosis method for equipment state management in any one of claims 4-6, and the system comprises: A data acquisition module for acquiring state monitoring data of equipment through a sensor and optimizing sample acquisition time and density by using a dynamic sampling strategy; A data optimization module for processing the acquired data by using the data optimization method for equipment health management; A state monitoring module for monitoring the real-time health state of equipment by using the optimized data. A fault diagnosis module uses intelligent diagnosis methods for equipment health management to diagnose abnormal conditions. A decision support module provides maintenance decision support based on monitoring and diagnosis results.