Industrial equipment operation state evaluation method and system

Abstract

The invention relates to an industrial equipment operation state evaluation method and system, belongs to the technical field of industrial equipment, and solves the problems that in the existing industrial equipment operation state monitoring process, continuous quantitative description of operation state changes is difficult, and overall description of regional state degradation is difficult, and the like, the method comprises the following steps: S1, collecting multi-source operation signals of equipment; s2, performing feature extraction on the multi-source operation signal; s3, modeling the state vector to obtain a reference model; s4, comparing the state vector with a reference model, and calculating a regional health potential; s5, correcting and smoothing the regional health potential to obtain an evolution model; and S6, carrying out comprehensive evaluation on the overall change trend and the risk accumulation condition of the equipment operation state, and outputting an operation state evaluation result. According to the invention, comprehensive evaluation of the operation state of the equipment and advanced perception of the operation risk are realized, so that the engineering applicability and reliability of industrial equipment state monitoring are improved.

Description

Technical Field

[0001] This invention belongs to the field of industrial equipment technology, specifically relating to a method and system for assessing the operating status of industrial equipment. Background Technology

[0002] As industrial equipment develops towards larger scale, higher speed, and continuous operation, the stability and reliability of equipment operation have become crucial factors affecting production efficiency and safety levels. During long-term operation, various industrial equipment is affected by a variety of factors, including load changes, environmental conditions, material aging, and structural wear, resulting in complex and dynamic changes in its operating status. To ensure safe equipment operation and reduce the risk of sudden downtime, continuous monitoring and comprehensive evaluation of equipment operating status are typically required in industrial settings.

[0003] Existing industrial equipment condition monitoring technologies mostly rely on multi-source sensor signals such as vibration, current, temperature, and acoustics. By analyzing a single signal or a small number of features, they determine whether the equipment is in normal working condition. These methods were effective in early engineering practices, but their limitations have gradually become apparent as equipment structures become more complex and operating conditions more diverse.

[0004] First, industrial equipment operating conditions inherently exhibit significant multi-condition characteristics. Under different loads, speeds, or production cycles, the vibration amplitude, temperature distribution, and current characteristics of the equipment often undergo systematic changes. These changes do not necessarily indicate a fault in the equipment, but rather represent its normal response under different operating conditions. If fixed thresholds or simple statistical rules are used to evaluate operating signals, it is easy to misjudge changes in operating conditions as abnormal states, or to mask the true trend of condition degradation under high load conditions, thereby affecting the reliability of condition assessment.

[0005] Secondly, industrial equipment typically consists of multiple structural or functional areas, such as drive areas, transmission areas, support areas, and cooling areas. These areas are structurally coupled, and their operating states exhibit significant spatial correlation. When wear, loosening, or performance degradation occurs in one area, its effects often spread gradually to adjacent areas through structural transmission, resulting in regional state changes. However, most existing monitoring methods still rely primarily on single-point or single-sensor signal analysis, making it difficult to describe the equipment's operating state at the overall structural level and effectively characterize the collaborative changes between areas.

[0006] Furthermore, some research has begun to explore using machine learning or deep learning methods to model equipment operating states, identifying or classifying states in a data-driven manner. While these methods have shown certain advantages in specific application scenarios, they typically rely on a large number of labeled samples or explicit state category classifications. In real industrial environments, equipment is often in a normal or sub-healthy state for extended periods, the number of clearly defined fault samples is limited, and the state performance varies significantly across different equipment and operating conditions, thus limiting the model's generalization ability. Moreover, these methods often target discrete state determinations, making it difficult to reflect the continuous changes in equipment operating states and the trend of risk accumulation.

[0007] From an engineering application perspective, industrial sites are more concerned with whether equipment operating conditions are continuously deviating, whether these changes show a trend, and whether there is a gradual accumulation of potential risks, rather than simply determining the condition at a particular moment. Therefore, technical approaches based solely on classification or condition deviation assessment are insufficient to meet the practical needs of long-term equipment condition assessment and operational risk management.

[0008] Furthermore, equipment operation status monitoring also faces challenges such as signal noise interference, sensor accuracy differences, and environmental changes. Random noise, transient disturbances, or individual sensor anomalies can all cause short-term fluctuations in condition characteristics. Without comprehensive constraints on the overall equipment structure and regional correlations, the condition assessment results can easily become unstable, affecting the reliability of engineering decisions.

[0009] In summary, existing industrial equipment condition monitoring technologies still have shortcomings in the following aspects: First, it is difficult to continuously and quantitatively characterize the overall operating status of equipment; second, there is a lack of effective means to depict the evolution of operating status from the perspective of equipment regional structure; third, the adaptability to changes in operating conditions and long-term state deviations is limited; and fourth, it is difficult to maintain the stability of assessment results under noise and complex environmental conditions. Summary of the Invention

[0010] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method and system for evaluating the operating status of industrial equipment. This method and system are based on the evolution of regional health potential fields and solve the technical problems in the existing industrial equipment operating status monitoring process, such as the difficulty in continuously and quantitatively describing changes in operating status, the difficulty in comprehensively depicting regional state degradation, and the insufficient stability of state evaluation under complex working conditions.

[0011] To address the above problems, the technical solution of this invention is: a method for evaluating the operating status of industrial equipment, comprising the following steps: S1: Acquisition of multi-source operating signals from the acquisition device; S2: Extract features from multi-source operating signals to obtain state vectors that characterize the operating status of each operating area of ​​the equipment; S3: Model the state vector to obtain a reference model that reflects the distribution characteristics of normal operating states in each region; S4: Compare the state vector with the reference model and calculate the health potential of the device region based on the degree of its offset; S5: Combining the changes in the health potential of adjacent or related areas of the equipment, the regional health potential is corrected and smoothed to obtain the evolution model of the regional health potential field. S6: Combining statistical analysis methods, comprehensively assess the overall trend of equipment operating status changes and risk accumulation, and output the operating status assessment results.

[0012] Preferably, in step S2, the device is divided into N regions: R={R1,R2,...,R N}; Where R is the set of equipment operating areas, and N is the number of areas; Within each region, several sensors are deployed to collect multi-source operating signals; for the i-th region, a multi-dimensional state vector is extracted at time t: x i (t)=[x i1 (t),x i2 (t),...,x im (t)]; Where, x i (t) is the region state vector, x im (t) represents the m-th feature component, where m is the number of feature dimensions extracted from this region, and t represents the discrete time point during the device's operation. Using these feature dimensions, a multi-regional, multi-variable state space for the device is constructed, serving as input for subsequent models.

[0013] Preferably, in step S3, a reference model is constructed through clustering learning patterns, as follows: For region R i The historical normal sample set is: X i ={x i (t1),x i (t2),...,x i (t k )}; t k This represents the k-th historical moment, where k represents the historical sample index; Perform a clustering algorithm to divide the historical normal sample set into L i One normal pattern cluster: ; Among them, L i This represents the number of clusters in normal operating mode. The center of each pattern cluster is: ; μ ij As the center of the pattern cluster, C ij Representing region R i The j-th normal operation mode cluster.

[0014] Preferably, in step S4, the distance-based regional health potential is defined as: ; The distance metric used is: Euclidean distance: , Or Mahalanobis distance: ; Where d(a,b) is the distance function. It is the inverse covariance matrix.

[0015] Preferably, in step S3, when the amount of normal data is large, a reference model is constructed based on kernel density estimation: ; Where, p i (x) represents the region R i The probability density value of the state vector x in the normal operation distribution, K(·) is the kernel function used for kernel density estimation, h is the parameter that controls the smoothness of kernel density estimation, and K represents the number of historical samples used for density estimation. The lower the density value, the more "rare" the sample is, and the more likely it is to be abnormal.

[0016] Preferably, in step S4, the regional health potential based on rarity is: ; The lower the probability, the higher the potential energy; The potential field is constructed as follows: The overall equipment forms a regional health potential field: ; By analyzing the abnormal trends of the equipment from both spatial and temporal dimensions.

[0017] Preferably, in step S5, Region structures have natural adjacency relationships, which can be represented by a neighborhood structure diagram as follows: ; Represents region R i A set of indexes for regions that are structurally or functionally adjacent or related; To suppress isolated noise and develop a regional linkage state representation capability, neighborhood correction is applied to the regional health potential: ; in, H represents the regional health potential after neighborhood collaborative correction and smoothing. i (t) represents the region R i The degree of deviation of the operating state from the normal mode at time t, where α is the collaborative weight parameter. Agg(·) represents the mean, median, or weighted average.

[0018] Preferably, the mean neighborhood filter is: ; Median Neighborhood Filtering: .

[0019] Preferably, in step S6, within the sliding time window W, the mean and standard deviation of the corrected potential value are statistically analyzed: , ; in, For the health mean, Standard deviation of health potential; Dynamically generated thresholds: ; in Representing region R i The dynamic state reference threshold at time t, where k is the threshold sensitivity adjustment parameter; The final anomaly detection criteria are: like > An anomaly has occurred; The anomaly level is defined based on the extent of the out-of-bounds error: ; in, This represents the state offset magnitude.

[0020] Another object of the present invention is to provide an industrial equipment operating status assessment system, which is used in the above-mentioned industrial equipment operating status assessment method, and the system includes: Multi-source operating signal acquisition module, used to acquire multi-source operating signals of the equipment; The regional state modeling module is connected to the multi-source operation signal acquisition module. The regional state modeling module extracts features from the multi-source operation signals to obtain state vectors that characterize the operation status of each operating area of ​​the equipment. The normal operation mode characterization module is connected to the regional state modeling module. The normal operation mode characterization module is used to model the state vector to obtain a reference model that reflects the distribution characteristics of the normal operation state of each region. The health potential calculation module is connected to the normal operation mode characterization module. The health potential calculation module is used to compare the state vector with the reference model and calculate the health potential of the device area based on its offset. The neighborhood collaborative correction module is connected to the health potential calculation module. The neighborhood collaborative correction module is used to combine the changes in the health potential of adjacent or related areas of the device to correct and smooth the regional health potential, and obtain the evolution model of the regional health potential field. The Dynamic State Reference Boundary Construction and Operational Status Assessment Module is connected to the Neighborhood Collaborative Correction Module. This module combines statistical analysis methods to comprehensively assess the overall trend of equipment operation status changes and risk accumulation, and outputs the operational status assessment results.

[0021] Compared with the prior art, the beneficial effects of the present invention are as follows: Starting from the regional structure of equipment, this invention models the process of changes in operating status and constructs a quantitative representation method that can reflect the degree of state deviation and evolution trend, so as to achieve a comprehensive assessment of the operating status of equipment and early perception of operating risks, thereby improving the engineering applicability and reliability of industrial equipment condition monitoring.

[0022] This invention enables continuous quantitative assessment of the operating status of industrial equipment, and can characterize the change process of equipment operating status at the regional structure level. It does not rely on a large number of fault samples or complex learning strategies, and is suitable for industrial environments with multiple operating conditions and long-term operation. It has good engineering applicability and stability. Attached Figure Description

[0023] Figure 1 This is a block diagram of the overall structure of the present invention; Figure 2 This is a schematic diagram illustrating the change of multi-regional health potential over time according to the present invention; Figure 3 This is a schematic diagram illustrating the single-region health potential, dynamic threshold, and state assessment of the present invention. Figure 4 This is a schematic diagram comparing the regional health potential before and after the neighborhood collaborative correction of the present invention. Detailed Implementation

[0024] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0025] Example 1: As Figures 1-4 As shown, this embodiment provides a method for evaluating the operating status of industrial equipment, including the following steps: S1, Multi-source operating signals from the acquisition device.

[0026] S2. Extract features from multi-source operating signals to obtain state vectors that characterize the operating status of each operating area of ​​the equipment.

[0027] Specifically, this invention starts from the equipment structure and function level. Based on the structural characteristics of industrial equipment or the arrangement of sensors, the equipment is divided into several operating areas. The multi-source operating signals such as vibration, current, and temperature collected in each operating area are preprocessed and feature extracted to construct a state vector representing the operating state of each area. Through the regional state modeling method, the operating state of the equipment can be refined from the overall level to each local area, providing a basis for subsequent operating state evaluation.

[0028] S3. Model the state vector to obtain a reference model that reflects the distribution characteristics of normal operating states in each region.

[0029] Specifically, based on data from the historical stable operation phases of the equipment, this invention performs statistical analysis or cluster modeling on the state vectors of the operating states in each region to construct a reference model that reflects the distribution characteristics of the normal operating states in each region. This reference model is used to characterize the typical operating states of the equipment under different operating conditions, providing a benchmark for the quantification of the degree of deviation in the operating state, thereby avoiding misjudgment problems caused by changes in operating conditions.

[0030] S4: Compare the state vector with the reference model and calculate the health potential of the device region based on the degree of its offset.

[0031] Specifically, this invention compares the state vectors of the real-time operating status of each region with the reference model of the distribution characteristics of the corresponding normal operating status, and calculates the health potential value of each region based on the degree of deviation. The health potential value is used to quantify the deviation level of the operating status of a region relative to the normal state, and its value can reflect the degree of change of the operating status of the region, forming a continuous quantitative representation of the operating status of the equipment.

[0032] S5: Combining the changes in the health potential of adjacent or related areas of the equipment, the regional health potential is corrected and smoothed to obtain the evolution model of the regional health potential field.

[0033] Specifically, considering that there are usually structural or functional correlations between different operating areas of industrial equipment, this invention further introduces a regional collaborative constraint mechanism to jointly analyze the health potential of each region; by combining the changes in the health potential of adjacent or related regions, the regional health potential is corrected and smoothed, thereby constructing an evolution model of the regional health potential field; in order to suppress the influence of local random fluctuations on the evaluation results and enhance the expressive ability of regional state changes.

[0034] S6: Combining statistical analysis methods, comprehensively assess the overall trend of equipment operating status changes and risk accumulation, and output the operating status assessment results.

[0035] Specifically, this invention analyzes the evolution characteristics of the regional health potential field over time and combines statistical analysis methods to comprehensively assess the overall trend of equipment operating status and risk accumulation. When the regional health potential continuously and significantly shifts within a certain time range, it can serve as a quantitative indicator of increased equipment operating risk, providing support for equipment operation, maintenance, and risk management.

[0036] This invention does not aim to determine the discrete state of equipment operation. Instead, it constructs a regional health potential field model for the equipment to quantitatively characterize the deviation, trend, and risk accumulation process of the equipment's operating state. This enables a comprehensive assessment of the operating state of industrial equipment and early detection of operational risks. Through regional modeling, normal operation mode characterization, and neighborhood cooperative constraints, it achieves continuous quantitative characterization of the operating state of complex equipment. This invention can characterize the changing process of equipment operating state at the regional structure level, without relying on a large number of fault samples or complex learning strategies. It is suitable for industrial environments with multiple operating conditions and long-term operation, and has good engineering applicability and stability.

[0037] This invention comprises a complete technical chain consisting of five core components: regional state modeling, normal operation mode characterization, health potential calculation, neighborhood collaborative correction, and dynamic state reference boundary generation. To improve the interpretability and engineering feasibility of the model, this invention introduces mathematical descriptions and calculation formulas (as follows) in each component, enabling the entire method to be universally deployed on different types of industrial equipment and used for continuous quantitative evaluation of equipment operating status.

[0038] In the industrial equipment operating status assessment method of this embodiment, the mathematical description of step S2 is as follows: Industrial equipment typically consists of multiple functionally or structurally independent sub-regions. This invention divides the equipment into N regions: R={R1,R2,...,R N}; Where R is the set of equipment operating areas, and N is the number of areas; Within each region, several sensors are deployed to collect multi-source operating signals; for the i-th region, a multi-dimensional state vector is extracted at time t: x i (t)=[x i1 (t),x i2 (t),...,x im (t)]; Where, x i (t) is the region state vector, representing region R. i The multidimensional operating state feature vector at time t. im(t) represents the m-th feature component, indicating the region R. i The j-th operational feature at time t. m is the number of feature dimensions extracted from this region, representing the number of feature components contained in the state vector of each region. t represents the discrete time point during the operation of the device. Feature dimensions include, but are not limited to: vibration RMS, characteristic frequency band energy, current fluctuation amplitude, temperature rise gradient, envelope spectrum peak value, local variance, and local entropy. Through these multi-dimensional features, a multi-region, multi-variable state space of the device is constructed as the input for subsequent models.

[0039] In the industrial equipment operating status assessment method of this embodiment, the mathematical model for step S3 is as follows: To avoid relying on a large number of fault samples, this invention introduces an unsupervised learning mechanism to construct the normal operation mode of each region of the equipment through clustering learning patterns or density estimation methods. (1) Clustering learning models (K-means / FCM, etc.) are as follows: For region R i The historical normal sample set is: X i ={x i (t1),x i (t2),...,x i (t k )}; t k This represents the k-th historical moment, where k represents the historical sample index; Perform a clustering algorithm to divide the above historical normal sample set into L i One normal pattern cluster: ; Among them, L i This represents the number of clusters in normal operating mode. The center of each pattern cluster is: ; μ ij As the center of the pattern cluster, C ij Representing region R i The j-th normal operation mode cluster.

[0040] (2) The density estimation method (kernel density method KDE) is as follows: When the normal data volume is large, a distribution model can be constructed based on kernel density estimation: ; Where, p i (x) represents the region R iThe probability density value of the state vector x in the normal operation distribution, K(·) is the kernel function used for kernel density estimation, such as the Gaussian kernel function, h is a parameter that controls the smoothness of kernel density estimation, and K represents the number of historical samples used for density estimation. The lower the density value, the more "rare" the sample is, and the more likely it is to be abnormal.

[0041] In this embodiment of the industrial equipment operating status assessment method, the present invention proposes a regional health potential. This is used to quantify the deviation between the current state and the normal pattern, and is the core metric of the entire model. In step S4, the region health potential function is constructed as follows: (1) Distance-based regional health potential If clustering is used for learning, then the regional health potential is defined as: ; The distance metric used is: Euclidean distance: , Or Mahalanobis distance: ; Where d(a,b) is the distance function. It is the inverse covariance matrix.

[0042] (2) Regional health potential based on rarity (inverse density) If kernel density estimation is used, the regional health potential is defined as: ; The lower the probability, the higher the potential energy; (3) Potential field construction The overall equipment forms a regional health potential field: ; By analyzing the abnormal trends of the equipment from both spatial and temporal dimensions.

[0043] In the industrial equipment operating status assessment method of this embodiment, the mathematical mechanism of step S5 is as follows: Region structures have natural adjacency relationships, which are represented in this invention using a neighborhood structure diagram: ; Represents region R i A set of indexes for regions that are structurally or functionally adjacent or related; To suppress isolated noise and develop a regional linkage state representation capability, neighborhood correction is applied to the regional health potential: ; in, H represents the regional health potential after neighborhood collaborative correction and smoothing. i(t) represents the region R i The deviation of the operating state from the normal mode at time t, where α is the collaborative weighting parameter, representing the weighting coefficient between the current region's health potential and the neighborhood aggregation result. Agg(·) represents an operator that aggregates the health potential of a neighborhood region, and can be the mean, median, or weighted average.

[0044] Common choices are: Mean neighborhood filtering is: ; Median Neighborhood Filtering: .

[0045] This step enables anomalies to form a "regional linkage identification capability" in space.

[0046] In a method for evaluating the operating status of industrial equipment in this embodiment, in step S6, the mean and standard deviation of the corrected potential value are statistically analyzed within the sliding time window W: , ; in, For the health mean, Standard deviation of health potential; Dynamically generated thresholds: ; Representing region R i The dynamic state reference threshold at time t, k is the threshold sensitivity adjustment parameter (usually 2 to 4). The final anomaly detection criteria are: like > An anomaly has occurred; The anomaly level can be further defined based on the extent of the out-of-bounds error: ; in, This represents the state offset magnitude.

[0047] Figure 2 This figure illustrates the changes in regional health potential over time in multiple areas of an industrial device during operation, as described in an embodiment of the present invention. Each curve in the figure corresponds to a monitoring area of ​​the device, with the vertical axis representing the health potential value and the horizontal axis representing time. This figure reflects the differences and trends in the state of each area at different operating stages. Some areas show a significant increase in regional health potential during specific time periods, indicating that the state of these areas deviates from the normal operating mode. This figure is used to display the overall temporal distribution of the regional health potential field, providing a basis for subsequent location and trend analysis of state deviation areas.

[0048] Figure 3 The figure displays the health potential value, dynamic threshold, and status assessment results of device monitoring area 2 throughout the entire operating cycle. The solid line represents the area health potential curve after neighborhood collaborative correction, the dashed line represents the dynamic threshold curve generated based on sliding window statistics, and the circular markers indicate the moments deemed abnormal. This figure illustrates the status shift assessment process of the present invention: when the area health potential exceeds the dynamic threshold, that moment is identified as abnormal. By comparing the area health potential curve and the dynamic threshold curve, the abnormality triggering mechanism can be visually observed, and the adaptive capability of the dynamic threshold under different operating conditions can be verified.

[0049] Figure 4 The figure shows a comparison of the regional health potential changes in monitoring area 3 before and after applying neighborhood collaborative correction. The thinner curve represents the original regional health potential, while the thicker curve represents the corrected regional health potential after neighborhood collaboration. It can be seen that the corrected curve is significantly smoother in areas with large noise disturbances, while maintaining sensitivity during abnormal periods. This figure illustrates the effect of the neighborhood collaborative correction module, namely, by fusing spatial neighborhood information to suppress random fluctuations, enhance the significance of regional anomalies, and improve the stability and reliability of the regional health potential in complex environments.

[0050] Example 2: As Figures 1-4 As shown, this embodiment provides an industrial equipment operation status assessment system. The system is used in the industrial equipment operation status assessment method in Embodiment 1. The system includes a multi-source operation signal acquisition module, a regional state modeling module, a normal operation mode characterization module, a health potential calculation module, a neighborhood collaborative correction module, and a dynamic state reference boundary construction and operation status assessment module.

[0051] The multi-source operating signal acquisition module is used to acquire multi-source operating signals from the equipment; The regional state modeling module is connected to the multi-source operation signal acquisition module. The regional state modeling module is used to divide the equipment into several operating areas, preprocess and extract features from the multi-source operation signals acquired in each operating area, and construct a state vector representing the operating state of each area. The normal operation mode characterization module is connected to the regional state modeling module. The normal operation mode characterization module is used to perform statistical analysis or cluster modeling on the state vector of the operating state of each region based on the data of the equipment's historical stable operation phase, and to build a reference model that reflects the distribution characteristics of the normal operation state of each region. The health potential calculation module is connected to the normal operation mode characterization module. The health potential calculation module is used to compare the state vector of the real-time acquired operation status of each region with the reference model of the corresponding normal operation status distribution characteristics, and calculate the health potential value of each region according to its offset. The neighborhood collaborative correction module is connected to the health potential calculation module. The neighborhood collaborative correction module is used to introduce a regional collaborative constraint mechanism to jointly analyze the health potential of each region. By combining the changes in the health potential of adjacent or related regions of the device, the regional health potential is corrected and smoothed, thereby constructing an evolution model of the regional health potential field. The Dynamic State Reference Boundary Construction and Operational Status Assessment Module, connected to the Neighborhood Collaborative Correction Module, is used to comprehensively assess the overall trend of equipment operation status and risk accumulation by analyzing the evolution characteristics of the regional health potential field over time and combining statistical analysis methods, and output the operational status assessment results.

[0052] By adopting the above-mentioned technical solution, this invention realizes continuous quantitative evaluation of the operating status of industrial equipment. It can characterize the change process of equipment operating status at the regional structure level, without relying on a large number of fault samples or complex learning strategies. It is suitable for industrial environments with multiple operating conditions and long-term operation, and has good engineering applicability and stability.

Claims

1. A method of evaluating an operating state of an industrial plant, characterized by, The method comprises the following steps: S1: collecting multi-source operation signals of the equipment; S2: extracting features from the multi-source operation signals to obtain a state vector representing the operation state of each operation region of the equipment; S3: modeling the state vector to obtain a reference model reflecting the distribution characteristics of the normal operation state of each region; S4: comparing the state vector with the reference model, and calculating the regional health potential of the equipment according to the offset degree; S5: combining the change of the health potential of the adjacent or related regions of the equipment, correcting and smoothing the regional health potential, and obtaining an evolution model of the regional health potential field; S6: combining a statistical analysis method, comprehensively evaluating the overall change trend and risk accumulation of the operation state of the equipment, and outputting the operation state evaluation result.

2. The method of claim 1, wherein In step S2, The equipment is divided into N regions: R = {R1, R2,..., R N}; Wherein, R is a set of equipment operation regions, and N is the number of regions; In each region, a plurality of sensors are arranged to collect multi-source operation signals; for the ith region, a multi-dimensional state vector is extracted at time t: x i (t) = [x i1 (t), x i2 (t),..., x im (t)] ; wherein x i (t) is the region state vector, x im (t) is the mth feature component, m is the dimension number of the features extracted by the region, and t represents the discrete time instant in the process of device operation.

3. The method of claim 1, wherein In step S3, the reference model is constructed by clustering learning mode, and the steps are as follows: For the region R i The historical normal sample set for R is: X i = {x i (t1),x i (t2),...,x i (t k )}; t k denotes the k-th historical time instant, k denotes the historical sample index; performing a clustering algorithm to divide the historical normal sample set into L i normal pattern clusters: ； wherein L i is the number of normal operation mode clusters; Each mode cluster center is: ； μ ij C ij R i jthnormal operation mode cluster of region R 4. The method of claim 3, wherein In step S4, the distance-based regional health potential is defined as: ； Wherein, the distance measurement adopts: Euclidean distance: , or Mahalanobis distance: ; where d(a, b) is a distance function, is the inverse covariance matrix.

5. The method of claim 1, wherein In step S3, when the amount of normal data is large, the reference model is constructed based on kernel density estimation: ； where p i (x) denotes the probability density value of the state vector x of the region R i in the normal operation distribution, K(·) is a kernel function for kernel density estimation, h is a parameter for controlling the smoothness of kernel density estimation, and K denotes the number of historical samples for density estimation.

6. The method of claim 5, wherein In step S4, the regional health potential based on rarity is: ； The potential field is constructed as: Overall device formation area health potential field: ; The abnormal trend of the equipment is analyzed from the spatial and temporal dimensions.

7. The method of claim 1, wherein In step S5, The regional structure is represented as a neighborhood structure diagram: ； represents a region R i a set of region indices that are structurally or functionally adjacent or related; The neighborhood correction is performed on the regional health potential: ； wherein, H denotes the region health potential after neighborhood collaborative correction and smoothing processing, i (t) denotes the region R i the offset degree of the operating state at time t relative to the normal mode, and a is a collaborative weight parameter, α ∈ 0 1 , and Agg(·) is the mean, median, or weighted average.

8. The method of claim 7, wherein The mean neighborhood filtering is: ； Median neighborhood filtering: .

9. The method of claim 1, wherein In step S6, the mean value and standard deviation of the corrected potential value are calculated in a sliding time window W: ， ； wherein, is the healthy potential mean, is the healthy potential standard deviation; The threshold value is dynamically generated: ； representative region R i dynamic state reference threshold at time t, k is a threshold sensitivity adjustment parameter; The final abnormality judgment condition is: If An exception occurs;​ The abnormality level is defined according to the out-of-bound amplitude: ； wherein is the state offset magnitude.

10. An industrial equipment operating status assessment system, characterized in that, The system is used for the industrial equipment operation state evaluation method in any one of claims 1-9, and the system comprises: A multi-source operation signal acquisition module for acquiring multi-source operation signals of the equipment; A regional state modeling module connected with the multi-source operation signal acquisition module, the regional state modeling module extracts features from the multi-source operation signals to obtain a state vector representing the operation state of each operation region of the equipment; A normal operation mode description module connected with the regional state modeling module, the normal operation mode description module is used for modeling the state vector to obtain a reference model reflecting the distribution characteristics of the normal operation state of each region; A health potential calculation module connected with the normal operation mode description module, the health potential calculation module is used for comparing the state vector with the reference model, and calculating the regional health potential of the equipment according to the offset degree; A neighborhood collaborative correction module connected with the health potential calculation module, the neighborhood collaborative correction module is used for combining the change of the health potential of the adjacent or related regions of the equipment, correcting and smoothing the regional health potential, and obtaining an evolution model of the regional health potential field; The dynamic state reference boundary construction and running state evaluation module is connected with the neighborhood collaborative correction module, and is used for combining a statistical analysis method to comprehensively evaluate the overall change trend and risk accumulation situation of the equipment running state, and outputting a running state evaluation result.

Images