Subjective and objective combined equipment health evaluation method and system

Abstract

The invention discloses a subjective and objective combined equipment health evaluation method and system, and the method comprises the steps: synchronously obtaining objective sensor data and subjective expert evaluation data of equipment operation through a multi-source data collection module, carrying out the preprocessing, and carrying out the parallel calculation of an objective weight and a subjective weight through an improved entropy weight method and a group decision AHP algorithm, dynamically fusing the two types of weights by using a time-varying combination coefficient strategy, and introducing a momentum factor to smooth weight mutation; real-time data processing and complex calculation separation are realized through a cloud edge collaborative architecture, weight full-life-cycle traceability is realized in combination with block chain evidence storage, and evaluation accuracy is ensured by a built-in multi-dimensional verification mechanism. The method achieves the complementation of subjective and objective advantages, solves the problems of high subjectivity, poor adaptability and insufficient traceability of evaluation indexes in the prior art, improves the equipment health evaluation accuracy, reduces the delay control, and is suitable for the full-life-cycle health management of various industrial equipment.

Description

Technical Field

[0001] This invention relates to the field of industrial equipment fault diagnosis technology, and more specifically, to a method and system for evaluating equipment health that combines subjective and objective methods. Background Technology

[0002] Against the backdrop of the rapid development of intelligent diagnostic technology, the scientific validity and rationality of evaluation indicators directly determine the accuracy of industrial equipment fault diagnosis. In existing technologies, the determination of equipment health evaluation indicators often relies on a single method: either based solely on objective analysis of sensor data, ignoring domain knowledge from expert experience; or over-relying on subjective expert judgment, leading to significant differences in indicator allocation due to individual experience. This process is not only time-consuming, but the varying levels of knowledge and experience among experts increase the uncertainty of evaluation indicators, easily causing an imbalance in indicator proportions, ultimately affecting the accuracy of equipment health status assessment and failing to meet the demands of industrial sites for reliable and dynamically adaptable evaluation results. Therefore, there is an urgent need for an equipment health evaluation technology that can integrate subjective and objective advantages, dynamically adjust weight allocation, and possess traceability to address the shortcomings of existing technologies. Summary of the Invention

[0003] In view of the problems existing in the prior art, the purpose of this invention is to provide a method and system for equipment health evaluation that combines subjective and objective approaches. It aims to solve the problems of strong subjectivity, poor dynamic adaptability and insufficient traceability of evaluation indicators in the prior art, so as to achieve complementary advantages of subjective and objective approaches and improve the accuracy and engineering practicality of equipment health assessment.

[0004] The present invention adopts the following technical solution:

[0005] This invention discloses a method for evaluating equipment health that combines subjective and objective methods, comprising the following steps:

[0006] Step 1: Multi-source data acquisition

[0007] The system synchronously collects objective monitoring data and expert subjective evaluation data of the equipment operation. The objective monitoring data is acquired in real time through a sensor network, and the subjective evaluation data is collected and quantified through an expert knowledge acquisition platform.

[0008] Step 2: Data Preprocessing

[0009] Objective monitoring data undergoes outlier removal, missing value imputation, standardization, and noise reduction; subjective evaluation data undergoes digital conversion, consistency verification, and fuzzy quantization.

[0010] Step 3: Parallel weight calculation

[0011] An improved entropy weight method is used to process the preprocessed objective data to obtain objective weights; the group decision-making AHP algorithm is used to process the preprocessed subjective data to obtain subjective weights.

[0012] Step 4: Dynamic Weight Fusion

[0013] Based on the time-varying combination coefficient α, objective and subjective weights are dynamically fused. In the initial stage, α=0.6 emphasizes subjective weights. As data accumulates, it transitions to α=0.4 with an exponential decay law, emphasizing objective weights. The fusion process introduces a momentum factor β to smooth weight mutations, where β∈[0.1,0.3].

[0014] Step 5: Health Assessment and Feedback Optimization

[0015] The equipment health index is calculated based on the fusion weights, and the evaluation results and early warning information are output in real time. The model is self-updated through reverse verification and new failure mode monitoring.

[0016] This invention also discloses a device health evaluation system that combines subjective and objective methods, characterized in that it includes:

[0017] Multi-source data acquisition module: used to acquire sensor monitoring data and expert evaluation data, including sensor network, industrial bus interface, expert knowledge acquisition platform and data storage unit;

[0018] Data preprocessing module: Deployed on edge computing nodes, it realizes temporal alignment, outlier filtering, wavelet threshold denoising and standardization of objective data, and semantic parsing, fuzzy quantization and consistency verification of subjective data;

[0019] Core computing modules include an improved entropy weight calculation engine and a group decision AHP calculation engine, which run in parallel at the edge and in the cloud. The edge performs real-time entropy weight calculation (latency < 1s), while the cloud performs batch calculations of group decision AHP.

[0020] Dynamic fusion module: Automatically adjusts the combination coefficient α based on data quality indicators to achieve the fusion of subjective and objective weights and momentum smoothing.

[0021] Output and feedback module: Includes a health status visualization interface, multi-level early warning unit and model self-updating unit, seamlessly integrates with the work order system, and achieves elastic expansion through microservice architecture.

[0022] Beneficial effects

[0023] Complementary advantages of subjective and objective factors: The innovative integration of the improved entropy weight method and the group decision-making AHP algorithm utilizes both the objective regularity of sensor data and the domain knowledge of expert experience. By dynamically combining coefficients to automatically adjust the weight ratio, the influence of data noise and subjective bias is avoided, thus improving the accuracy of health assessment compared to a single method.

[0024] Strong dynamic adaptability: By adopting sliding window entropy weight calculation, time-varying combination strategy and momentum smoothing mechanism, the system can automatically adjust the weight allocation according to changes in equipment status and fluctuations in data quality, and adapt to the state evolution of industrial equipment throughout its entire life cycle.

[0025] Highly practical for engineering applications: It adopts a cloud-edge collaborative architecture, with real-time data processing completed at the edge (latency < 1s) and complex calculations performed in the cloud, meeting the real-time requirements of industrial sites; it provides a visual interface and multi-level early warning, and seamlessly integrates with the work order system, significantly improving equipment management efficiency.

[0026] The entire process is verifiable and traceable: Verification nodes are set up throughout the chain (consistency check, correlation analysis, AUC verification), and all weight versions are stored on the blockchain to form a traceable weight evolution curve, which meets the quality audit requirements of the industrial field. Attached Figure Description

[0027] Figure 1 This is a module architecture and data flow diagram of an embodiment of the equipment health evaluation system of the present invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0029] As shown in the figure, this invention discloses a method for evaluating equipment health that combines subjective and objective methods, comprising the following steps:

[0030] Step 1: Multi-source data acquisition

[0031] The system synchronously collects objective monitoring data and expert subjective evaluation data of the equipment operation. The objective monitoring data is acquired in real time through a sensor network, and the subjective evaluation data is collected and quantified through an expert knowledge acquisition platform.

[0032] Step 2: Data Preprocessing

[0033] Objective monitoring data undergoes outlier removal, missing value imputation, standardization, and noise reduction; subjective evaluation data undergoes digital conversion, consistency verification, and fuzzy quantization.

[0034] Step 3: Parallel weight calculation

[0035] An improved entropy weight method is used to process the preprocessed objective data to obtain objective weights; the group decision-making AHP algorithm is used to process the preprocessed subjective data to obtain subjective weights.

[0036] Step 4: Dynamic Weight Fusion

[0037] Based on the time-varying combination coefficient α, objective and subjective weights are dynamically fused. In the initial stage, α=0.6 emphasizes subjective weights, and as data accumulates, it transitions to α=0.4 with an exponential decay law, emphasizing objective weights. The fusion process introduces a momentum factor β to smooth weight abrupt changes. The momentum smoothing formula for the weight sequence is:

[0038] β∈[0.1,0.3]

[0039] in: Let be the final smoothed weight of the j-th evaluation index at time t;

[0040] Let t be the historical weight of the j-th indicator at time t-1 (i.e., the smoothed weight from the previous round).

[0041] The original weight of the j-th indicator newly calculated in the current round (the unsmoothed result obtained by fusing subjective and objective weights).

[0042] The processed data output consists of standardized health indicators with clear physical meaning, and a Sankey diagram visualization analysis report of the contribution of each indicator is generated. All processing steps have built-in data quality verification modules. When abnormal input is detected, an adaptive compensation calculation process is triggered to ensure that the system can maintain stable evaluation performance even when some data is missing.

[0043] Step 5: Health Assessment and Feedback Optimization

[0044] The equipment health index is calculated based on the fusion weights, and the evaluation results and early warning information are output in real time. The model is self-updated through reverse verification and new failure mode monitoring.

[0045] In one embodiment of the present invention, in step 1, the objective monitoring data includes vibration, current, displacement, noise, and rotational speed parameters. These parameters are collected through a sensor network composed of vibration acceleration sensors, dynamic current sensors, etc., based on the Modbus or Profinet protocol at a sampling frequency of not less than 1Hz. This data is also combined with semi-structured data such as historical maintenance records and spare parts replacement cycles. The subjective evaluation data is generated by 5-7 senior experts using the Delphi method in multiple rounds of evaluation. Qualitative indicators are assessed using the AHP questionnaire (Cronbach's α>0.8) for reliability and validity testing, and quantified using a 1-9 scale. Simultaneously, unstructured observation records such as equipment noise characteristics and oil color changes are collected through an expert collaborative annotation system. All subjective evaluation data are stored in a knowledge graph database after semantic analysis and fuzzy set transformation. A correlation mapping is established between the subjective evaluation data and the objective monitoring data using a unified timestamp and equipment ID, forming a digital twin dataset containing all dimensions of equipment status characteristics. This provides a fusion data foundation for subsequent weight calculations.

[0046] In one embodiment of the present invention, the subjective and objective data processing adopts a multi-stage fusion analysis process. First, the objective monitoring data collected by the sensors is processed by time-domain synchronization alignment and dimension normalization. The root mean square value, peak-to-peak value, waveform factor and other feature parameters of each indicator are extracted by sliding window statistics. An improved wavelet threshold denoising algorithm is used to eliminate signal distortion caused by electromagnetic interference in the industrial field. At the same time, the subjective data of expert evaluation is semantically parsed and fuzzy quantized. The triangular membership function is used to convert the language evaluation (such as "slight vibration" "severe abnormal noise") into a numerical score in the [0,1] interval. The multi-source subjective judgments are weighted and aggregated based on the expert authority weight. After the two types of data are uniformly processed by Box-Cox transformation to eliminate distribution skewness, the feature weights are calculated by the improved entropy weight method and the group decision AHP method, respectively.

[0047] The entropy weight method incorporates an adaptive sliding window mechanism to dynamically adjust the data time span. The improved entropy weight method introduces fuzzy entropy theory and an adaptive sliding window mechanism, and eliminates small sample bias through the Sigmoid entropy function. The entropy value is calculated using the following formula:

[0048]

[0049] When entropy value At that time, objective weight ;

[0050] When entropy value hour, ,in It is based on the historical average weight.

[0051] The equipment health evaluation system framework based on the improved entropy weighting method and the AHP (Advanced Hierarchical Organization for Health Management) adopts a modular design to achieve dynamic integration of subjective and objective weights. First, sensor monitoring data and expert evaluation questionnaires are acquired through a multi-source data acquisition module. Sensor data undergoes real-time preprocessing (including outlier removal, missing value imputation, and sliding window standardization) via edge computing nodes. Expert evaluation data undergoes digital conversion and consistency verification through a knowledge management system. The core computing layer runs the improved entropy weighting calculation engine and the AHP engine in parallel. Entropy weighting calculation uses fuzzified monitoring data, eliminating small sample bias by introducing the Sigmoid entropy function and a dynamic weight correction mechanism. The fuzzification formula before weighting calculation is as follows:

[0052]

[0053] in: For the first Unit equipment, the first The original characteristic values ​​of objective monitoring indicators (such as the root mean square value of vibration indicators, the fluctuation coefficient of current indicators, etc., are the original characteristics extracted from sensor data).

[0054] Yes Normalized feature values ​​after fuzzification (output results are compressed to the ([0,1]) interval);

[0055] a=10: Adjust the parameter (controls the steepness of the Sigmoid curve; the larger the value, the more drastic the curve changes around 0.5).

[0056] In step 3, the group decision-making AHP algorithm generates a group judgment matrix based on expert authority weighting. Through iterative optimization, the consistency ratio of the group judgment matrix is ​​controlled below 0.1, and the largest eigenvector is extracted as the subjective weight.

[0057] The formula for group aggregation of multi-expert evaluation data in group decision-making is:

[0058]

[0059] in: For the first The expert, regarding the first Each assessment object (such as a certain piece of equipment), the first The quantitative score given by each evaluation indicator (such as vibration, current, etc.) (e.g., subjective score after quantification using the 1-9 scale).

[0060] It is the first Weighting of expert authority;

[0061] It is the total number of experts who participated in the evaluation;

[0062] It is the final overall score of the group.

[0063] In one embodiment of the present invention, in step 4, the dynamic weight fusion adopts a multiplicative synthesis formula:

[0064]

[0065] in, Subjective weighting, For objective weighting.

[0066] In one embodiment of the present invention, in step 5, the reverse verification is performed using historical failure cases, requiring AUC≥0.85, and the consistency of weight ranking is analyzed by Spearman correlation coefficient, and the weights are merged and stored on the blockchain to form a traceable evolution curve.

[0067] This invention also discloses a device health evaluation system that combines subjective and objective methods, characterized in that it includes:

[0068] Multi-source data acquisition module: used to acquire sensor monitoring data and expert evaluation data, including sensor network, industrial bus interface, expert knowledge acquisition platform and data storage unit;

[0069] Data preprocessing module: Deployed on edge computing nodes, it realizes temporal alignment, outlier filtering, wavelet threshold denoising and standardization of objective data, and semantic parsing, fuzzy quantization and consistency verification of subjective data;

[0070] Core computing modules include an improved entropy weight calculation engine and a group decision AHP calculation engine, which run in parallel at the edge and in the cloud. The edge performs real-time entropy weight calculation, while the cloud performs batch calculations for group decision AHP.

[0071] Dynamic fusion module: Automatically adjusts the combination coefficient α based on data quality indicators to achieve the fusion of subjective and objective weights and momentum smoothing.

[0072] Output and feedback module: Includes a health status visualization interface, multi-level early warning unit and model self-updating unit, seamlessly integrates with the work order system, and achieves elastic expansion through microservice architecture.

[0073] The output and feedback module displays the equipment health index and the contribution of key indicators in real time. When the health index exceeds the threshold, it triggers multi-level warnings. At the same time, the system has a built-in feedback optimization mechanism. It achieves model self-update by periodically verifying the effectiveness of weights (using reverse verification AUC≥0.85) and monitoring new fault modes. The entire system is deployed on an industrial cloud platform. The microservice architecture ensures the independent scalability of each module. Edge-cloud collaborative computing ensures that the entropy weight calculation with high real-time requirements is completed locally (latency <1s), and complex group decision AHP calculation is executed in the cloud (daily batch updates). Finally, a closed-loop optimized equipment health management system is formed.

[0074] In one embodiment of the present invention, a cloud-edge collaborative architecture is adopted, where edge computing nodes complete real-time data preprocessing and entropy weight calculation, while the cloud performs group decision AHP calculation, weight fusion verification and model update, and performs full weight update daily. Special recalculation is triggered when a new fault mode or significant change in data distribution is detected.

[0075] Weight calculation deployment process:

[0076] The weight calculation process first standardizes the collected objective monitoring data through a data preprocessing module, extracts time-frequency domain features using a sliding window technique, and eliminates dimensional differences. Simultaneously, it performs fuzzy quantification and consistency verification on the subjective data from expert evaluations, constructing a complete indicator dataset. Then, it concurrently launches an improved entropy weighting method and a group decision-making AHP calculation engine. The entropy weighting method incorporates fuzzy entropy theory to handle data uncertainty, dynamically adjusts the calculation cycle through an adaptive sliding window, and sets an entropy threshold to trigger a weight correction mechanism. When the indicator data dispersion is detected to be too small, it automatically mixes historical average weights. The group decision-making AHP calculation first assigns individual weights based on an expert authority model, aggregates the judgment matrix using a weighted geometric average algorithm, and controls the consistency ratio of the group matrix below 0.1 through iterative optimization. Finally, it extracts the largest eigenvector as the subjective weight. The two types of weight results are input into a dynamic fusion module. This module automatically adjusts the subjective and objective combination coefficient α based on data quality indicators. When initial data is insufficient, it emphasizes expert weights (α=0.6). As monitoring data accumulates, it gradually transitions to emphasizing objective weights (α=0.4) according to an exponential decay law. It uses a multiplicative synthesis formula for normalization combination and adds a momentum factor to smooth weight abrupt changes. After each calculation, a verification program is executed. The consistency of weight ranking is analyzed through Spearman correlation coefficient and reverse verification is performed using historical failure cases to ensure that the AUC value is greater than 0.85. The final generated combined weights are updated to the health assessment model in real time and written to the blockchain for evidence storage, forming a traceable weight evolution curve. The entire process adopts a microservice architecture to realize the elastic expansion of the computing module. A full weight update is performed daily. When the system detects a new failure mode or a significant change in data distribution, a special recalculation is immediately triggered to ensure the timeliness and accuracy of weight allocation.

[0077] Please refer to the table below:

[0078]

[0079] The above description is merely a preferred embodiment of the present invention; however, the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and its improved concepts, should be covered within the scope of protection of the present invention.

Claims

1. A method for evaluating equipment health that combines subjective and objective methods, characterized in that: Includes the following steps: Step 1: Multi-source data acquisition The system synchronously collects objective monitoring data and expert subjective evaluation data of the equipment operation. The objective monitoring data is acquired in real time through a sensor network, and the subjective evaluation data is collected and quantified through an expert knowledge acquisition platform. Step 2: Data Preprocessing Objective monitoring data undergoes outlier removal, missing value imputation, standardization, and noise reduction; subjective evaluation data undergoes digital conversion, consistency verification, and fuzzy quantization. Step 3: Parallel weight calculation An improved entropy weighting method is used to process the preprocessed objective data to obtain objective weights. The preprocessed subjective data is processed using the group decision-making AHP algorithm to obtain subjective weights; Step 4: Dynamic Weight Fusion Based on the time-varying combination coefficient α, objective and subjective weights are dynamically fused. In the initial stage, α=0.6 emphasizes subjective weights. As data accumulates, it transitions to α=0.4 with an exponential decay law, emphasizing objective weights. The fusion process introduces a momentum factor β to smooth weight mutations, where β∈[0.1,0.3]. Step 5: Health Assessment and Feedback Optimization The equipment health index is calculated based on the fusion weights, and the evaluation results and early warning information are output in real time. The model is self-updated through reverse verification and new failure mode monitoring.

2. The equipment health evaluation method combining subjective and objective methods according to claim 1, characterized in that: In step 1, the objective monitoring data includes vibration, current, displacement, noise, and rotational speed parameters. These parameters are collected through a sensor network consisting of vibration acceleration sensors, dynamic current sensors, etc., based on the Modbus or Profinet protocol at a sampling frequency of not less than 1Hz. This data is also combined with semi-structured data such as historical maintenance records and spare parts replacement cycles. The subjective evaluation data is generated by 5-7 senior experts in the field using the Delphi method in multiple rounds of evaluation. The importance of qualitative indicators is determined through an AHP questionnaire with reliability and validity testing, and quantification is performed using the 1-9 scale.

3. The equipment health evaluation method combining subjective and objective methods according to claim 1, characterized in that: In step 3, the improved entropy weight method introduces fuzzy entropy theory and an adaptive sliding window mechanism, and eliminates small sample bias through the Sigmoid entropy function. The entropy value is calculated using the formula: When entropy value At that time, objective weight ; When entropy value hour, ,in It is based on the historical average weight.

4. The equipment health evaluation method combining subjective and objective methods according to claim 1, characterized in that: In step 3, the group decision-making AHP algorithm generates a group judgment matrix based on expert authority weighting. Through iterative optimization, the consistency ratio of the group judgment matrix is ​​controlled below 0.1, and the largest eigenvector is extracted as the subjective weight.

5. The equipment health evaluation method combining subjective and objective methods according to claim 1, characterized in that: In step 4, the dynamic weight fusion adopts a multiplicative synthesis formula: in, Subjective weighting, For objective weighting.

6. The equipment health evaluation method combining subjective and objective methods according to claim 5, characterized in that: In step 5, the reverse verification is performed using historical failure cases, requiring AUC ≥ 0.

85. At the same time, the consistency of weight ranking is analyzed by Spearman correlation coefficient, and the weights are merged and stored on the blockchain to form a traceable evolution curve.

7. A device health evaluation system that combines subjective and objective methods, characterized in that: include: Multi-source data acquisition module: used to acquire sensor monitoring data and expert evaluation data, including sensor network, industrial bus interface, expert knowledge acquisition platform and data storage unit; Data preprocessing module: Deployed on edge computing nodes, it realizes temporal alignment, outlier filtering, wavelet threshold denoising and standardization of objective data, and semantic parsing, fuzzy quantization and consistency verification of subjective data; Core computing modules include an improved entropy weight calculation engine and a group decision AHP calculation engine, which run in parallel at the edge and in the cloud. The edge performs real-time entropy weight calculation, while the cloud performs batch calculations for group decision AHP. Dynamic fusion module: Automatically adjusts the combination coefficient α based on data quality indicators to achieve the fusion of subjective and objective weights and momentum smoothing. Output and feedback module: Includes a health status visualization interface, multi-level early warning unit and model self-updating unit, seamlessly integrates with the work order system, and achieves elastic expansion through microservice architecture.

8. The equipment health evaluation system combining subjective and objective methods according to claim 7, characterized in that: Adopting a cloud-edge collaborative architecture, edge computing nodes complete real-time data preprocessing and entropy weight calculation, while the cloud performs group decision AHP calculation, weight fusion verification and model update. Full weight updates are performed daily on a scheduled basis, and special recalculation is triggered when new fault modes or significant changes in data distribution are detected.

Images