A method, device and medium for adaptive health assessment of a multi-condition device

Abstract

This application discloses an adaptive health assessment method, device, and medium for multi-condition equipment, relating to the field of industrial equipment assessment technology. The method includes: performing density clustering on operational index data to determine the number of operating conditions for the industrial equipment; based on the number of operating conditions, performing multi-condition switching judgments for each operating condition within the corresponding assessment period to obtain the operating condition segments corresponding to each operating condition within the assessment period; constructing operating condition feature vectors corresponding to the operating condition segments, and performing operating condition matching on the operating condition feature vectors through trend similarity calculation to determine the operating condition type corresponding to the operating condition segment; calculating the kernel density adaptability score corresponding to each operating condition segment based on health index data; and calculating the industrial equipment health corresponding to each kernel density adaptability score through kernel density differential bandwidth matching. This application solves the technical problems of inaccurate operating condition switching judgments and lack of specific differences in operating conditions in existing multi-condition equipment health assessments.

Description

Technical Field

[0001] This application relates to the field of industrial equipment evaluation technology, and in particular to an adaptive health assessment method, device and medium for multi-condition equipment. Background Technology

[0002] In the industrial production sector, equipment health assessment is one of the core technologies for ensuring production continuity and reducing operation and maintenance costs. It is widely used in industries such as petrochemicals, power, and water treatment, and is particularly suitable for general rotating machinery such as centrifugal pumps, fans, and motors. With the development of the Industrial Internet and intelligent manufacturing, equipment health assessment is shifting from traditional periodic maintenance to predictive maintenance. Its core trend is to quantify the health status of equipment based on real-time operating data through algorithmic models, and to provide early warnings of potential failures.

[0003] In existing technologies, the health assessment of industrial equipment is mainly divided into two categories: single-condition fixed threshold method and multi-index fusion assessment method. First, the single-condition fixed threshold method is based on historical data of a single stable operating condition of the equipment, sets a fixed health threshold, and judges the health status of the equipment by comparing real-time data with the threshold. Second, the multi-index fusion assessment method integrates multi-dimensional health indicators through machine learning and other methods to calculate a comprehensive health score, thereby reducing the limitations of single-index assessment. Because equipment such as centrifugal pumps in industrial equipment often needs to switch operating conditions (flow rate, load, etc.) according to production needs, the distribution of health indicators varies significantly under different operating conditions. Specific analysis is needed for the classification and switching of operating conditions, the judgment of indicator anomalies, and the differences in operating condition specificity. Existing technologies rely on manual experience for operating condition classification, suffer from serious misjudgments during operating condition switching, and lack the representation of specific differences in operating conditions, failing to meet the needs of today's complex industrial production environment. Summary of the Invention

[0004] This application provides an adaptive health assessment method, device, and medium for multi-condition equipment, which solves the technical problems in the prior art where the health assessment of multi-condition equipment is inaccurate in judging the switching of operating conditions and lacks specific differences in operating conditions.

[0005] In a first aspect, embodiments of this application provide an adaptive health assessment method for multi-condition equipment. The method includes: acquiring operational indicator data of industrial equipment and performing density clustering on the operational indicator data to determine the number of operational conditions of the industrial equipment; based on the number of operational conditions, performing multi-condition switching judgment on each operational condition within the corresponding assessment period of the industrial equipment to obtain the operational condition segment corresponding to each operational condition within the assessment period; constructing operational condition feature vectors corresponding to operational condition segments, and performing operational condition matching on the operational condition feature vectors through trend similarity calculation to determine the operational condition type corresponding to the operational condition segment; acquiring health indicator data within the operational condition segment corresponding to the operational condition type, and calculating the kernel density adaptability score corresponding to each operational condition segment based on the health indicator data; and calculating the health of the industrial equipment corresponding to each kernel density adaptability score through kernel density differential bandwidth matching.

[0006] In one implementation of this application, density clustering is performed on the operating index data to determine the number of operating conditions of industrial equipment. Specifically, this includes: setting an initial cluster number corresponding to the operating index data and determining the number of samples of other operating index data in the ε-neighborhood of the operating index data; traversing the operating index data, and marking the corresponding operating index data as noise points when the number of samples is less than a preset minimum number of samples; constructing a first cluster when the number of samples is greater than or equal to the preset minimum number of samples, and adding other operating index data in the ε-neighborhood of the corresponding operating index data to the first cluster to obtain a second cluster; traversing each core point in the second cluster and adding the operating index data in the ε-neighborhood of the core point to the second cluster until no new operating index data is added, thus obtaining an equipment operating condition cluster; and determining the number of operating conditions based on the equipment operating condition cluster and the noise points.

[0007] In one implementation of this application, based on the number of operating conditions, a multi-condition switching determination is performed on each operating condition within the corresponding evaluation period of the industrial equipment to obtain the operating condition segment corresponding to each operating condition within the evaluation period. Specifically, this includes: calculating the operating condition discrimination degree corresponding to the operating index data based on the number of operating conditions to obtain the index weight of each operating condition of the industrial equipment; determining the variance threshold and verifying the operating condition mutation for the operating conditions within the corresponding evaluation period of the industrial equipment to determine the operating condition mutation characteristics; calculating the operating condition mutation confidence degree corresponding to the operating condition mutation characteristics based on the number of operating conditions and the index weight, and obtaining the mutation completion point of the operating condition switching through the operating condition switching determination based on the operating condition mutation confidence degree; and performing a weighted average of the mutation completion points corresponding to each operating condition within the evaluation period to determine the operating condition segment.

[0008] In one implementation of this application, variance threshold determination and operational condition mutation verification are performed on the operating conditions within the corresponding evaluation period of the industrial equipment to determine the operational condition mutation characteristics. Specifically, this includes: obtaining the mean of standardized indicators within the corresponding evaluation period of the industrial equipment, and calculating the variance corresponding to each operating condition based on the mean of standardized indicators; comparing the variance with a preset variance threshold to determine the mutation verification status corresponding to each operating condition; wherein, the mutation verification status includes: suspected mutation and non-mutation; performing operational condition mutation anomaly detection on the operating conditions suspected of mutation, and calculating the continuous window mean fluctuation of the operating conditions that pass the operational condition mutation anomaly detection to determine the operational condition mutation characteristics.

[0009] In one implementation of this application, the working condition feature vector is matched with the working condition feature vector by calculating the trend similarity to determine the working condition type corresponding to the working condition segment. Specifically, this includes: calculating the weighted Euclidean distance between the working condition feature vector and each working condition in the preset working condition library, and filtering working conditions whose weighted Euclidean distance is less than or equal to a preset distance parameter to determine the matching working condition set; calculating the trend similarity corresponding to each working condition in the matching working condition set, and performing a stability check on the trend similarity to determine the working condition type corresponding to the working condition segment.

[0010] In one implementation of this application, the kernel density adaptability score corresponding to each working condition segment is calculated based on health indicator data. Specifically, this includes: preprocessing the health indicator data to determine the basic information of the working condition and the health level range; and calculating the kernel density adaptability score corresponding to each working condition segment based on the basic information of the working condition and the health level range.

[0011] In one implementation of this application, after calculating the kernel density adaptability score corresponding to each working condition segment based on the basic working condition information and health level range, the method further includes: if the kernel density adaptability score and the amount of single indicator data do not meet the preset passing threshold, supplementing the amount of single indicator data until the kernel density adaptability score and the amount of single indicator data meet the passing threshold.

[0012] In one implementation of this application, the health of industrial equipment corresponding to each kernel density adaptability score is calculated by kernel density differential bandwidth matching. Specifically, this includes: calculating the first industrial equipment health corresponding to the kernel density adaptability score using a first bandwidth formula when the kernel density adaptability score is greater than a first differential threshold; calculating the second industrial equipment health corresponding to the kernel density adaptability score using a second bandwidth formula when the kernel density adaptability score is less than the first differential threshold but greater than or equal to a second differential threshold; and performing interval solving on the first and / or second industrial equipment health to obtain the industrial equipment health; wherein, the industrial equipment health includes: equipment health probability, equipment sub-health probability, and equipment failure probability.

[0013] Secondly, embodiments of this application also provide an adaptive health assessment device for multi-condition equipment, characterized in that the device includes: a data acquisition module for acquiring operational index data and health index data within the corresponding operating condition segments of the industrial equipment; an operating condition analysis module for performing density clustering on the operational index data to determine the number of operating conditions of the industrial equipment; the operating condition analysis module is further used to perform multi-condition switching determination on each operating condition within the corresponding assessment period of the industrial equipment based on the number of operating conditions, so as to obtain the operating condition segments corresponding to each operating condition within the assessment period; the operating condition analysis module is further used to construct the operating condition feature vector corresponding to the operating condition segment, and perform operating condition matching on the operating condition feature vector through trend similarity calculation to determine the operating condition type corresponding to the operating condition segment; a health assessment module for calculating the kernel density adaptability score corresponding to each operating condition segment based on the health index data; and the health assessment module is further used to calculate the health of the industrial equipment corresponding to each kernel density adaptability score through kernel density differential bandwidth matching.

[0014] Thirdly, embodiments of this application also provide a non-volatile computer storage medium for adaptive health assessment of multi-condition equipment, storing computer-executable instructions, characterized in that the computer-executable instructions, when executed, can implement an adaptive health assessment method for multi-condition equipment.

[0015] This application provides an adaptive health assessment method, device, and medium for multi-condition equipment. By calculating the number of operating conditions of multi-condition industrial equipment, dividing the operating conditions of industrial equipment within the required health assessment time period, matching the operating conditions corresponding to each operating condition segment, and assessing the health of industrial equipment through an improved kernel density comprehensive score, this method solves the technical problems of inaccurate judgment of operating condition switching and lack of specific differences in operating conditions in the existing multi-condition equipment health assessment. It realizes the self-identification of operating conditions of multi-condition industrial equipment and reduces the probability of misjudgment across operating conditions. Attached Figure Description

[0016] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 A flowchart of an adaptive health assessment method for multi-condition equipment provided in this application embodiment; Figure 2 This application provides a schematic diagram of a centrifugal water pump operating condition clustering method. Figure 3 This is a schematic diagram of an adaptive health assessment device for multi-condition equipment provided in an embodiment of this application. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0018] This application provides an adaptive health assessment method, device, and medium for multi-condition equipment. By calculating the number of operating conditions of multi-condition industrial equipment, dividing the operating conditions of industrial equipment within the required health assessment time period, matching the operating conditions corresponding to each operating condition segment, and assessing the health of industrial equipment through an improved kernel density comprehensive score, this method solves the technical problems of inaccurate judgment of operating condition switching and lack of specific differences in operating conditions in the existing multi-condition equipment health assessment. It realizes the self-identification of operating conditions of multi-condition industrial equipment and reduces the probability of misjudgment across operating conditions.

[0019] The technical solutions proposed in the embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0020] Figure 1 A flowchart illustrating an adaptive health assessment method for multi-condition equipment provided in this application embodiment. Figure 1 As shown in the figure, the adaptive health assessment method for multi-condition equipment provided in this application embodiment specifically includes the following steps: Step 101: Obtain the operating index data of industrial equipment and perform density clustering on the operating index data to determine the number of operating conditions of industrial equipment.

[0021] For example, traditional clustering algorithms rely on manually setting parameters (such as ε-neighborhood radius, minimum number of samples, etc.), which cannot achieve adaptive calculation of the number of operating conditions and clustering parameters. This application determines the number of operating conditions of industrial equipment by performing density clustering on operational index data, and automatically calculates the minimum number of samples based on the operating condition dimension and the number of samples, without the need for manual parameter adjustment, thus realizing adaptive calculation of the number of operating conditions for different industrial equipment under multiple operating conditions.

[0022] Specifically, density clustering is performed on the operational indicator data to determine the number of operating conditions for industrial equipment. This includes: setting the initial cluster number corresponding to the operational indicator data and determining the number of samples of other operational indicator data in the ε-neighborhood of the operational indicator data; traversing the operational indicator data, and marking the corresponding operational indicator data as noise points when the number of samples is less than the preset minimum number of samples; constructing the corresponding first cluster when the number of samples is greater than or equal to the preset minimum number of samples, and adding other operational indicator data in the ε-neighborhood of the corresponding operational indicator data to the first cluster to obtain the second cluster; traversing each core point in the second cluster and adding the operational indicator data in the ε-neighborhood of the core point to the second cluster until no new operational indicator data is added, thus obtaining the equipment operating condition cluster; and determining the number of operating conditions based on the equipment operating condition cluster and the noise points.

[0023] In one embodiment, firstly, operational indicator data of industrial equipment is collected through acquisition devices such as industrial sensors, PLCs, and DCS systems. The operational indicator data needs to include the corresponding operating condition characteristic data of the equipment, that is, the core operating condition parameters of the equipment (such as the flow rate and pressure difference of centrifugal pumps; the air volume and air pressure of fans, etc.), and needs to cover the multi-dimensional operating condition characteristics required for clustering processing. The operational indicator data is then subjected to data standardization processing.

[0024] Then, based on the collected and standardized operational metric data, the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) algorithm, an improved version of machine learning, is used to classify different operating conditions. The DBSCAN algorithm is a density-based clustering algorithm that groups samples with connected densities into the same cluster. It can automatically identify clusters of arbitrary shapes and label samples in low-density areas as noise points.

[0025] The key parameters and explanations for the DBSCAN algorithm used in this application are as follows: Adaptive Neighborhood, based on samples (i.e., operational metric data). Centered on, with radius as The circular region. Based on operational index data, this application constructs an adaptive neighborhood radius, which is explained by the following formula.

[0026] (1) in, This is the scene calibration coefficient (0.15~0.2 for industrial equipment, default 0.15). For operating condition characteristics (such as flow rate plus pressure difference), ); For the first The standard deviation of a standardized operating condition characteristic. Due to The radius of the neighborhood is adaptively constructed. The radius of the neighborhood does not need to be manually set, providing a data foundation for adaptive density clustering.

[0027] By collecting historical data (5 minutes / time, 8640 records) from centrifugal pumps over 30 days, operating characteristics included flow rate ( ) and pressure difference ( Two indicators, therefore Scene calibration coefficient The default value of 0.15 is used, and the data has already been preprocessed. The standard deviation of the flow rate is calculated. Standard deviation of pressure difference (After standardization), the corresponding adaptive neighborhood radius .

[0028] Minimum number of samples This is used to ensure the statistical validity of clusters and avoid single-point clusters. This application sets... .

[0029] in, This represents the total number of samples. Indicates rounding up; when hour, Therefore, it can be simplified to .

[0030] The key point is that if the sample of The neighborhood contains at least For each sample, then The core point. If the sample of Number of samples in the neighborhood ,but The core point.

[0031] Boundary point, at the core point Within the neighborhood, but itself Number of samples in the neighborhood The sample.

[0032] Noise points are samples that are neither core points nor boundary points; the proportion of noise points. Automatically assigned to the nearest neighbor condition without affecting the overall classification.

[0033] Density can be achieved, sample from Density is achievable if and only if a sample chain exists. , , ,and exist of Within the neighborhood, from Density can be achieved.

[0034] Density connected, if samples and All from the core point If the density is achievable, then and Density connected.

[0035] In the process of density clustering of operational indicator data in this application, it is necessary to first collect all operational indicator data (samples). Initialize to an unmarked state, with the corresponding cluster number set to [value]. .

[0036] Iterate through each unlabeled sample ,turn up of All samples within the neighborhood; if the number of neighborhood samples ,mark If the sample is a noise point, proceed to the next sample; if the number of neighboring samples is... Create a new cluster (number) ),Will Add all unlabeled samples in the neighborhood to the cluster; traverse each core point within the cluster. ,turn up of For neighboring samples, add unlabeled samples to the current cluster until no new samples can be added. After the traversal is complete, output all clusters and noise points.

[0037] like Figure 2 As shown, Figure 2 This is a schematic diagram illustrating the division of operating conditions of a centrifugal water pump into clusters, as provided in an embodiment of this application.

[0038] For density clustering of centrifugal water pumps, two typical operating conditions were identified: low flow rate and high flow rate, which are two clustered distributions. The blue cluster represents the low flow rate condition (20-30 m³ / h, average flow rate 25.03 m³ / h, average pressure difference 0.401 MPa), with a total of 5184 data points. The orange cluster represents the high flow rate condition (40-50 m³ / h, average flow rate 42.98 m³ / h, average pressure difference 0.698 MPa), with a total of 3456 data points. The two identified operating conditions do not have significant overlap.

[0039] Step 102: Based on the number of operating conditions, perform multi-condition switching judgment for each operating condition within the corresponding evaluation period of the industrial equipment to obtain the operating condition segment corresponding to each operating condition within the evaluation period.

[0040] For example, equipment health assessment is typically a periodic task, and the equipment may undergo multiple operating condition switches within the assessment period. If a uniform health assessment standard is used, misjudgments are highly likely. Even if the equipment is within the healthy range in two states, applying a threshold for a single operating condition might mistakenly classify one normal operating condition as a fault. Therefore, before conducting a health assessment, it is necessary to detect whether multiple operating conditions coexist within the assessment period and accurately identify the equipment operating condition type corresponding to each period based on a preset operating condition classification standard. This application achieves the identification of operating condition switching points by performing multi-condition switching determination on each operating condition within the corresponding assessment period of industrial equipment, thus avoiding the mixing of cross-condition data.

[0041] Specifically, based on the number of operating conditions, multi-condition switching is determined for each operating condition within the corresponding evaluation period of the industrial equipment to obtain the corresponding operating condition segment within the evaluation period. This includes: calculating the operating condition discrimination degree corresponding to the operating index data based on the number of operating conditions to obtain the index weight of each operating condition of the industrial equipment; determining the variance threshold and verifying the operating condition mutation for the operating conditions within the corresponding evaluation period of the industrial equipment to determine the operating condition mutation characteristics; calculating the operating condition mutation confidence degree corresponding to the operating condition mutation characteristics based on the number of operating conditions and index weights, and obtaining the mutation completion point of the operating condition switching through operating condition switching determination based on the operating condition mutation confidence degree; and performing a weighted average of the mutation completion points corresponding to each operating condition within the evaluation period to determine the operating condition segment.

[0042] Furthermore, variance threshold determination and operational condition mutation verification are performed on the operating conditions of the industrial equipment within the corresponding evaluation period to determine the characteristics of operational condition mutations. Specifically, this includes: obtaining the mean of standardized indicators within the corresponding evaluation period of the industrial equipment, and calculating the variance corresponding to each operating condition based on the mean of standardized indicators; comparing the variance with the preset variance threshold to determine the mutation verification status corresponding to each operating condition; wherein, the mutation verification status includes: suspected mutation and non-mutation; performing operational condition mutation anomaly detection on the operating conditions suspected of mutation, and calculating the continuous window mean fluctuation of the operating conditions that pass the operational condition mutation anomaly detection to determine the characteristics of operational condition mutations.

[0043] In one embodiment, when there are multiple indicators for determining operating conditions, such as flow rate and pressure difference for determining the operating conditions of a centrifugal pump, the determination needs to consider assigning weights based on the distinguishability of each operating condition characteristic; the higher the distinguishability, the greater the weight.

[0044] Furthermore, a change in operating condition is only determined when two or more core indicators are detected to change simultaneously and their trends are consistent, in order to ensure the compliance of the determination. For a single indicator to change, it is determined to be an outlier rather than a change in operating condition, thereby reducing misjudgments caused by sensor failure.

[0045] Based on the above judgment rules, firstly, the contribution (i.e., the distinguishability) of each working condition feature to the working condition differentiation is quantified, and explained by the following formula.

[0046] (2) in, For the first The weights of each indicator, after normalization ; For the first The variance of each indicator under different working conditions (the larger the value, the stronger the ability to distinguish working conditions). For the first The variance of an indicator under the same operating conditions (the smaller the value, the more stable the indicator). This is used to avoid the denominator being 0.

[0047] Given that the operating conditions of a centrifugal pump are determined by two indicators, flow rate and pressure difference, a segment of the collected historical data, containing 1000 data points, is extracted. The actual operating conditions of the extracted segment are represented in sequence as follows: 1-600 points (low flow rate condition), 601-650 points (gradual switching), and 651-1000 points (high flow rate condition).

[0048] Based on the data characteristics of low and high flow rates, the contribution of flow rate and pressure difference to the operating conditions is calculated and represented in Table 1 below.

[0049] Table 1. Working Condition Contribution Parameter Table

[0050] Then, abrupt changes are detected for the characteristic indicators of each working condition within the time period by configuring a sliding window (setting the window size W=10 data points and the step size L=1 data point) and calculating the variance within the window.

[0051] Among them, for indicators The The variance of a window is calculated as follows: (3) in, This represents the average of standardized indicators within the window.

[0052] It is important to note that the variance threshold is set to... ;in, As a global benchmark, the initial test adaptable to all operating conditions refers to the mean of the variance of this index under various pure operating conditions, obtained through statistical analysis of historical data. The trigger condition is if... If two consecutive windows meet this condition, it is marked as a suspected mutation.

[0053] Since centrifugal pumps operate in two modes: low flow and high flow, under the low flow condition... , =0.007; Under high flow conditions, 7, =0.007.

[0054] Therefore, the mean variance of the pure operating condition ,index For traffic, For pressure difference: , ; corresponding variance threshold , .

[0055] Furthermore, the window variance of the 600th data point =0.021>0.0195, =0.020>0.018, and the window variances of the subsequent 601st and 602nd data points also satisfy this condition. The situation is as follows: if two consecutive windows meet the conditions, it is suspected to be a sudden change, that is, the 600th data point is suspected to be the operating condition switching point.

[0056] The formula for calculating the direction of mean change in the suspected mutation stage is: (The difference between the mean values ​​of the two windows before and after).

[0057] Furthermore, the following trend verification rules based on physical principles can be set to assist in the judgment: the direction of the mean change must conform to the physical logic of the operating condition switch, such as when switching from low flow to high flow. >0 and >0; When switching from high to low traffic. <0 and <0. If the trends are contradictory (e.g., the average flow rate increases while the average pressure difference decreases), the indicator is judged as abnormal, and the judgment of the sudden change of the indicator is terminated; if the trends are consistent, the subsequent judgment is carried out. If the indicator value exceeds the physical range (e.g., flow rate <0, pressure difference <0), it is directly marked as sensor fault filtering and will not participate in the subsequent calculation.

[0058] Starting at 600 points, according to The calculations all yielded >0 and If the value is >0, it can be preliminarily determined that the operation has switched from a low flow rate to a high flow rate, and the subsequent step of determining the completion time of the sudden change can be taken.

[0059] Determining whether the mutation is complete requires continuous monitoring of the mean value within a window. If there are N=3 consecutive windows and the mean fluctuation is ≤5%, it indicates that the indicator has stabilized within the new range. The judgment logic is as follows: .

[0060] Record the point at which the mutation of this indicator is completed. (i.e., the starting point of the 3rd stable window), marked as the mutation of index i ( For indicator Q, if the mean fluctuation of the indicator is ≤5% for three consecutive windows starting from the 645th data point, then the mutation point is considered complete. =645, =1.

[0061] For indicator P, if the mean fluctuation of the indicator is ≤5% for three consecutive windows starting from the 650th data point, then the mutation point is complete. =650, =1.

[0062] Finally, the switching of operating conditions is determined by the synergy of multiple indicators, and the corresponding switching confidence level is calculated. The following formula will be used to explain this.

[0063] (4) in, The number of operating condition indicators; As an indicator The state (mutated = 1, non-mutated = 0); Set confidence threshold For example, taking 50%, the following switching judgment rules are used to determine whether the working condition has changed: like ,and If an indicator is marked as a sudden change, it is determined to be a change in actual operating conditions; like or not all If all indicators change abruptly, it is determined that the indicators that caused the changes are all single-indicator abnormalities (not due to operating condition switching).

[0064] The corresponding normalized weights have been solved. ,in 0.55, 0.45; both flow rate and pressure difference indicators were judged to be abrupt changes. and Switch confidence level Furthermore, both indicators showed sudden changes, indicating a switch to a real operating condition.

[0065] Furthermore, take The weighted average of the points where each indicator mutation occurs is used as the final switching point. , The transition phase is the intermediate process from the stable state of the old operating condition to the stable state of the new operating condition, where the starting point of the transition phase is... The earliest triggering window start point for suspected mutations in the two indicators; the transition segment termination point. final switching point (The starting point where both indicators are stable); the transition range [ , This indicates that data within this range is not included in health calculations and threshold updates due to unstable operating conditions.

[0066] For repeated mutations, filtering is performed, for example, ensuring that the working condition switch is not repeatedly determined within 100 data points, to avoid misjudgments caused by frequent fluctuations. The corresponding normalized weights have been calculated. ;in 0.55, 0.45. For the index Q, the mutation completion point... =645; For index P, the mutation completion point =650.

[0067] Determine the switching point (647th data point); transition section start point =600, which is the earliest trigger point for the suspected sudden change in flow; the transition section termination point. =647, which is the final switching point; the transition range is 600-647 points, a total of 48 data points. Data in this range is not included in the health calculation.

[0068] Step 103: Construct the working condition feature vector corresponding to the working condition segment, and perform working condition matching on the working condition feature vector through trend similarity calculation to determine the working condition type corresponding to the working condition segment.

[0069] For example, for any segment of pure operating conditions... This requires calculating the mean vector of multiple indicators for all data points within the segment, and then performing condition matching on the condition feature vector through potential similarity calculation, so as to provide a data foundation for the comprehensive health assessment of each condition segment.

[0070] Specifically, by calculating trend similarity, the working condition feature vector is matched to determine the working condition type corresponding to the working condition segment. This includes: calculating the weighted Euclidean distance between the working condition feature vector and each working condition in the preset working condition library, and filtering working conditions whose weighted Euclidean distance is less than or equal to a preset distance parameter to determine the matching working condition set; calculating the trend similarity corresponding to each working condition in the matching working condition set, and performing a stability check on the trend similarity to determine the working condition type corresponding to the working condition segment.

[0071] In one embodiment, firstly, the arbitrary pure operating condition segments are divided... Calculate the multi-indicator mean vector of all data points within this segment (representing the core features of this segment), and interpret it using the following formula.

[0072] (5) in, ; This represents the number of data points within the operating condition segment r. For the first The first data point Individual indicator values, The total number of indicators, Indicates the first [unit / item] in this working condition section The average value of each indicator.

[0073] Taking the segmented data to be matched within a certain time period as an example, the working condition segment r=1 (points 1-599) is a purely low-flow segment. =599, The operating condition segment r=2 (648-1000 points) is a purely high flow segment. =353, .

[0074] Then, calculate the feature vector of the current segment. The weighted Euclidean distance between the load condition and the load condition in the load condition database. ;in, This refers to the working conditions obtained through clustering. Lower operating condition judgment indicators The value of the cluster centers.

[0075] Filter out those that meet the requirements Candidate working condition set .like Enter the final matching stage; if If so, it is determined to be a suspected fault, and the equipment or sensor needs to be inspected.

[0076] For the 1000 points extracted, the operating condition is C1 (low flow). , For operating condition C2 (high flow rate): , .

[0077] In a cluster of 8640 points, the cluster centers for low-flow conditions are... for Clustering centers for high-flow operating conditions for For segment 1, the distance to C1 is The distance from C2 is The initial matching result is Similarly, for segment 2, the distance to C1 is... The distance from C2 is Initial matching results: .

[0078] Finally, to ensure that the fluctuations in the current segment data conform to the inherent stability law of the target working condition, the variance of each indicator within the current segment must be less than or equal to the mean variance within the corresponding candidate working condition, expressed as: ;in, That is, the first Section Working Condition No. The variance of each indicator.

[0079] If the candidate working condition set It includes multiple working conditions. The trend similarity between the current segment and each candidate working condition is calculated. The closer the similarity is to 1, the more closely the indicator trend matches. This is explained by the following formula.

[0080] (6) Pick The maximum operating condition that passes the stability check is taken as the final matching result; if... It has only one operating condition and has passed the stability check, so it is directly matched to that operating condition.

[0081] For segment 1, variance , The verification passed; With only one operating condition, there is no need to calculate similarity; the final matching result is that segment 1 matches C1 (low flow operating condition).

[0082] For segment 2, variance , The verification passed; the final matching result shows that segment 2 matches C2 (high flow condition).

[0083] Step 104: Obtain health indicator data within the working condition segment corresponding to the working condition type, and calculate the kernel density adaptability score corresponding to each working condition segment based on the health indicator data.

[0084] For example, this application calculates the kernel density adaptability score corresponding to each operating condition segment to determine the adaptability of the corresponding operating condition segment to kernel density calculation, thereby realizing the preliminary judgment of health analysis of each operating condition segment of industrial equipment and providing a data foundation for kernel density calculation of the health of industrial equipment.

[0085] Specifically, based on health indicator data, the kernel density adaptability score corresponding to each working condition segment is calculated, including: preprocessing the health indicator data to determine the basic information of the working condition and the health level range; and calculating the kernel density adaptability score corresponding to each working condition segment based on the basic information of the working condition and the health level range.

[0086] Furthermore, after calculating the kernel density adaptability score corresponding to each working condition segment based on the basic working condition information and health level range, the method also includes: if the kernel density adaptability score and the amount of single indicator data do not meet the preset passing threshold, supplementing the amount of single indicator data until the kernel density adaptability score and the amount of single indicator data meet the passing threshold.

[0087] In one embodiment, firstly, through condition-specific data input, the health scores of all key indicators within the pure operating condition segment (such as centrifugal pump spindle vibration, temperature, etc.) are calculated; basic operating condition information, such as the initial thresholds of C1 / C2 (health... ,Fault Health level range: The health range is greater than Sub-health is to The fault is less than (For example, a centrifugal pump in a low-flow state scores 81-100 points for healthy, 40-80 points for sub-healthy, and ≤39 points for faulty).

[0088] Then, calculate the KDE applicability comprehensive score for this working condition. The nuclear density suitability assessment is explained using the following formula.

[0089] (7) (8) in, The amount of data for key indicators. This represents the ratio of the data volume range. Score based on the amount of data for a single indicator. The score is based on the balance of the data volume. For the number of indicators, A comprehensive score is given for the applicability of kernel density.

[0090] It should be noted that the data volume for each individual indicator is required. ,and If the data is positive, proceed to the industrial equipment health status step; otherwise, additional data is required.

[0091] Step 105: Calculate the health of industrial equipment corresponding to each kernel density adaptability score by using kernel density differential bandwidth matching.

[0092] For example, this application calculates the health of industrial equipment corresponding to each kernel density adaptability score by kernel density differential bandwidth matching, thereby realizing the health assessment of multi-condition industrial equipment based on kernel density calculation. This solves the technical problems of inaccurate judgment of condition switching and lack of condition-specific differences in the health assessment of multi-condition equipment in the prior art, realizes the self-identification of the condition of multi-condition industrial equipment, and reduces the probability of misjudgment across conditions.

[0093] Specifically, through kernel density differential bandwidth matching, the health of industrial equipment corresponding to each kernel density adaptability score is calculated, including: when the kernel density adaptability score is greater than the first differential threshold, the health of the first industrial equipment corresponding to the kernel density adaptability score is calculated using the first bandwidth formula; when the kernel density adaptability score is less than the first differential threshold but greater than or equal to the second differential threshold, the health of the second industrial equipment corresponding to the kernel density adaptability score is calculated using the second bandwidth formula; the health of the first industrial equipment and / or the health of the second industrial equipment is solved by interval calculation to obtain the health of the industrial equipment; wherein, the health of the industrial equipment includes: equipment health probability, equipment sub-health probability, and equipment failure probability.

[0094] In one embodiment, KDE modeling and density calculation are performed under different load conditions. First, according to... Value selection differential bandwidth formula.

[0095] when When using this formula, the following method should be employed: (9) in, The interquartile range (75th percentile) of the r-th indicator Subtract the 25th percentile ), The standard deviation of the r-th indicator when When using the adaptive bandwidth formula: (10) in, Let r be the sample size for the r-th indicator; The average sample size for all indicators. ; This serves as a weighting index to balance differences in data volume. It is usually taken as 0.5.

[0096] Furthermore, using the Epanechnikov kernel, the health probability density curve for the current operating condition segment is generated according to the following formula. : (11) in: ; ; ; ; ; .

[0097] Finally, based on the preset comprehensive health score threshold, the probability of each health interval is calculated. For example, for those in the sample interval The probability that the equipment is healthy is: The probability of equipment being in a sub-optimal state is... The probability of equipment failure is Take the health level with the highest probability as the final state of the working condition segment and output the result.

[0098] In one embodiment, in the unweighted health assessment based on 1000 data points for different operating conditions, the health assessment for the low-flow operating condition (C1) is as follows: The time period is set from 1 to 599 points; the operating condition is low flow rate (20-30 m³ / h); key indicators include: vibration acceleration, abnormal value of bearing housing vibration spectrum, bearing temperature, motor winding temperature, shaft seal leakage, impeller wear, three-phase current imbalance, and insulation resistance.

[0099] KDE Applicability Overall Score The calculation is based on the data array .

[0100] , , Scores for the amount of data for each indicator. Further .

[0101] because Therefore, using the standard bandwidth formula, the health level thresholds for low-traffic operating conditions are as follows: Healthy: 81-100 points; Sub-healthy: 40-80 points; Faulty: 0-39 points.

[0102] Calculate the probability of each health level: P(healthy) = P(X ≥ 81) = 62.3%; P(sub-healthy) = P(81>X ≥ 40) = 31.5%; P(faulty) = P(X< 40) = 6.2%.

[0103] Therefore, the assessment results for low-flow operating conditions are: 62.3% probability of being healthy, 31.5% probability of being sub-healthy, and 6.2% probability of being faulty.

[0104] The health assessment for high-flow conditions (C2) is similar and will not be elaborated here.

[0105] The above are embodiments of the method proposed in this application. Based on the same inventive concept, embodiments of this application also provide an adaptive health assessment device for multi-condition equipment, the structure of which is as follows: Figure 3 As shown.

[0106] Figure 3 This is a schematic diagram of the internal structure of an adaptive health assessment device for multi-condition equipment provided in an embodiment of this application. Figure 3 As shown, the device includes: The data acquisition module 301 is used to acquire the operating index data of industrial equipment and the health index data within the corresponding operating condition segment of the operating condition type; the operating condition analysis module 302 is used to perform density clustering on the operating index data to determine the number of operating conditions of the industrial equipment; the operating condition analysis module 302 is also used to perform multi-operating condition switching judgment on each operating condition within the corresponding evaluation period of the industrial equipment based on the number of operating conditions to obtain the operating condition segment corresponding to each operating condition within the evaluation period; the operating condition analysis module 302 is also used to construct the operating condition feature vector corresponding to the operating condition segment, and perform operating condition matching on the operating condition feature vector through trend similarity calculation to determine the operating condition type corresponding to the operating condition segment; the health assessment module 303 is used to calculate the kernel density adaptability score corresponding to each operating condition segment based on the health index data; the health assessment module 303 is also used to calculate the health of the industrial equipment corresponding to each kernel density adaptability score through kernel density differential bandwidth matching.

[0107] Some embodiments of this application provide corresponding to Figure 1 A non-volatile computer storage medium storing computer-executable instructions, which, when executed, can implement an adaptive health assessment method for multi-condition equipment.

[0108] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments for IoT devices and media are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0109] The systems, media, and methods provided in this application are one-to-one correspondences. Therefore, the systems and media also have similar beneficial technical effects as their corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the systems and media will not be repeated here.

[0110] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0111] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0112] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0113] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0114] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0115] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0116] Computer-readable media include both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0117] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0118] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for adaptive health assessment of a multi- condition device, the method comprising: The method includes: Obtain operational indicator data of industrial equipment and perform density clustering on the operational indicator data to determine the number of operating conditions of the industrial equipment; Based on the number of operating conditions, a multi-condition switching determination is performed on each operating condition within the corresponding evaluation period of the industrial equipment to obtain the operating condition segment corresponding to each operating condition within the evaluation period. Construct the working condition feature vector corresponding to the working condition segment, and perform working condition matching on the working condition feature vector through trend similarity calculation to determine the working condition type corresponding to the working condition segment; Obtain health indicator data within the working condition segment corresponding to the working condition type, and calculate the kernel density adaptability score corresponding to each working condition segment based on the health indicator data. By matching the differential bandwidth of kernel density, the health of industrial equipment corresponding to each kernel density fitness score is calculated.

2. The adaptive health assessment method for multi-condition equipment according to claim 1, characterized in that, Density clustering is performed on the operational indicator data to determine the number of operating conditions for the industrial equipment, specifically including: Set the initial cluster number corresponding to the operational indicator data, and determine the number of samples of other operational indicator data in the ε-neighborhood corresponding to the operational indicator data; Traverse the operational indicator data, and if the number of samples is less than the preset minimum number of samples, mark the corresponding operational indicator data as noise points; If the number of samples is greater than or equal to the preset minimum number of samples, construct the corresponding first cluster and record the corresponding operational indicator data. Other operational metrics data from the neighborhood are added to the first cluster to obtain the second cluster; Traverse each core point in the second cluster and add the operation index data in the ε neighborhood of the core point to the second cluster until no new operation index data is added, thus obtaining the equipment condition cluster; The number of operating conditions is determined based on the equipment operating condition cluster and the noise point.

3. The adaptive health assessment method for multi-condition equipment according to claim 1, characterized in that, Based on the number of operating conditions, a multi-condition switching determination is performed on each operating condition within the corresponding evaluation period of the industrial equipment to obtain the operating condition segment corresponding to each operating condition within the evaluation period, specifically including: Based on the number of operating conditions, the operating condition discrimination degree corresponding to the operating indicator data is calculated to obtain the indicator weight of each operating condition of the industrial equipment. Variance threshold determination and operating condition mutation verification are performed on the operating conditions of the industrial equipment within the corresponding evaluation period to determine the characteristics of operating condition mutations. Based on the number of operating conditions and the weight of the indicators, the confidence level of the operating condition change corresponding to the operating condition change feature is calculated, and based on the confidence level of the operating condition change, the change completion point of the operating condition switch is obtained through the operating condition switch determination. The working condition segment is determined by weighted averaging of the abrupt change completion points corresponding to each working condition within the evaluation period.

4. The adaptive health assessment method for multi-condition equipment according to claim 3, characterized in that, The variance threshold for the operating conditions of the industrial equipment within the corresponding evaluation period is determined, and the characteristics of sudden changes in operating conditions are verified, specifically including: Obtain the mean of standardized indicators for the industrial equipment within the corresponding evaluation period, and calculate the variance for each operating condition based on the mean of the standardized indicators. By comparing the variance with a preset variance threshold, the mutation verification status corresponding to each working condition is determined; wherein, the mutation verification status includes: suspected mutation and non-mutation; The suspected abnormal operating conditions are subjected to anomaly detection, and the mean fluctuation of the continuous window of the operating conditions that pass the anomaly detection is calculated to determine the abnormal operating condition characteristics.

5. The adaptive health assessment method for multi-condition equipment according to claim 1, characterized in that, By calculating trend similarity, the operating condition feature vector is matched to determine the operating condition type corresponding to the operating condition segment, specifically including: Calculate the weighted Euclidean distance between the working condition feature vector and each working condition in the preset working condition library, and filter the working conditions whose weighted Euclidean distance is less than or equal to the preset distance parameter to determine the matching working condition set; Calculate the trend similarity corresponding to each working condition in the matching working condition set, and perform a stability check on the trend similarity to determine the working condition type corresponding to the working condition segment.

6. The adaptive health assessment method for multi-condition equipment according to claim 1, characterized in that, Based on the aforementioned health indicator data, calculate the kernel density adaptability score corresponding to each operating condition segment, specifically including: The health indicator data are preprocessed to determine the basic working condition information and health level range; Based on the aforementioned basic operating condition information and the aforementioned health level range, the kernel density adaptability score corresponding to each operating condition segment is calculated.

7. The adaptive health assessment method for multi-condition equipment according to claim 6, characterized in that, After calculating the kernel density adaptability score corresponding to each working condition segment based on the aforementioned basic working condition information and the aforementioned health level intervals, the method further includes: If the kernel density fitness score and the amount of single-indicator data do not meet the preset passing threshold, the amount of single-indicator data is supplemented until the kernel density fitness score and the amount of single-indicator data meet the passing threshold.

8. The adaptive health assessment method for multi-condition equipment according to claim 1, characterized in that, By using kernel density differential bandwidth matching, the health of industrial equipment corresponding to each kernel density adaptability score is calculated, specifically including: If the kernel density adaptability score is greater than the first differentiation threshold, the first industrial equipment health level corresponding to the kernel density adaptability score is calculated using the first bandwidth formula. If the kernel density adaptability score is less than the first differentiation threshold and greater than or equal to the second differentiation threshold, the second industrial equipment health level corresponding to the kernel density adaptability score is calculated using the second bandwidth formula. The health status of the first industrial equipment and / or the health status of the second industrial equipment is solved by interval calculation to obtain the health status of the industrial equipment; wherein, the health status of the industrial equipment includes: equipment health probability, equipment sub-health probability, and equipment failure probability.

9. An adaptive health assessment device for multi-condition equipment, characterized in that, The device includes: The data acquisition module is used to acquire the operating index data of industrial equipment and the health index data within the corresponding operating condition segment of the operating condition type. The operating condition analysis module is used to perform density clustering on the operating indicator data to determine the number of operating conditions of the industrial equipment. The operating condition analysis module is also used to determine the multi-operating condition switching of each operating condition within the corresponding evaluation period of the industrial equipment based on the number of operating conditions, so as to obtain the operating condition segment corresponding to each operating condition within the evaluation period. The working condition analysis module is also used to construct the working condition feature vector corresponding to the working condition segment, and to perform working condition matching on the working condition feature vector through trend similarity calculation in order to determine the working condition type corresponding to the working condition segment. The health assessment module is used to calculate the kernel density adaptability score corresponding to each working condition segment based on the health indicator data. The health assessment module is also used to calculate the health of industrial equipment corresponding to each kernel density adaptability score by matching the kernel density differential bandwidth.

10. A non-volatile computer storage medium for adaptive health assessment of multi-condition equipment, storing computer-executable instructions, characterized in that, When the computer-executable instructions are executed, they can implement the adaptive health assessment method for multi-condition equipment as described in any one of claims 1-8.

Images