Equipment state monitoring and anomaly analysis strategy method and system based on PMS

Abstract

The invention discloses an equipment state monitoring and anomaly analysis strategy method and system based on a PMS, and belongs to the technical field of equipment management.The method comprises the steps that equipment basic information, real-time operation state data, historical maintenance records and risk level parameters in an equipment management system are obtained, a comprehensive equipment state data set is generated, and a dynamic health attention index is calculated; and generating a strategy chain instruction set to schedule the sensor group to execute data acquisition operation, generating a target monitoring data set, inputting the target monitoring data set into the multi-stage adaptive analysis tree in combination with an analysis strategy instruction, generating a structured analysis report, and matching maintenance decision suggestions from the maintenance knowledge base and outputting the maintenance decision suggestions when a preset condition of a decision triggering strategy instruction is met. According to the method, the dynamic health attention index is adopted to drive strategy chain instruction generation, and the multi-stage adaptive analysis tree hierarchical diagnosis and closed-loop parameter optimization mechanism is combined, so that accurate distribution of monitoring resources, accurate positioning of fault root causes, economic optimization of maintenance decision and adaptive continuous upgrading of the system can be realized.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment management, and in particular to a method and system for equipment status monitoring and abnormality analysis based on PMS. Background Art

[0002] Equipment management systems play a core role in modern industrial operations and maintenance, enabling status monitoring by integrating basic equipment information, real-time operating data, and historical maintenance records. Existing technologies generally rely on preset thresholds or fixed cycles for anomaly detection and maintenance decisions, falling under the category of static management.

[0003] Current mainstream solutions use a unified monitoring strategy to collect data from all devices, analyze real-time data using a single diagnostic model, and trigger maintenance recommendations based on fixed rules. Some systems incorporate basic risk assessments, but these weighting factors are often static and lack dynamic adjustment mechanisms.

[0004] However, existing solutions suffer from rigid monitoring resource allocation, resulting in wasted resources for low-risk equipment and insufficient monitoring of high-risk equipment. Single diagnostic models struggle to address complex fault scenarios, resulting in high rates of false positives and missed negatives. Maintenance decisions lack real-time adaptability to operating conditions, leading to over-maintenance and delayed response. Fixed system parameters prevent automatic optimization based on operational performance, leading to decreased accuracy over time. Summary of the Invention

[0005] To address the above issues, the present invention provides a PMS-based equipment status monitoring and anomaly analysis strategy method and system. This system uses a dynamic health concern index to drive strategy chain instruction generation, combined with a multi-level adaptive analysis tree hierarchical diagnosis and closed-loop parameter optimization mechanism. This enables precise allocation of monitoring resources, accurate location of root causes of faults, economic optimization of maintenance decisions, and continuous adaptive upgrades of the system.

[0006] The above objectives can be achieved through the following solutions: A PMS-based equipment status monitoring and anomaly analysis strategy method and system include obtaining basic equipment information, real-time operating status data, historical maintenance records and preset risk level parameters in the equipment management system to generate a comprehensive equipment status data set; calculating a dynamic health concern index based on the comprehensive equipment status data set; generating a policy chain instruction set including monitoring strategy instructions, analysis strategy instructions and decision triggering strategy instructions based on the dynamic health concern index; scheduling a preset sensor group to perform data collection operations according to the monitoring strategy instructions to generate a target monitoring data set; inputting the target monitoring data set and the analysis strategy instructions into a preset multi-level adaptive analysis tree, performing hierarchical diagnosis operations, and generating a structured analysis report; when the structured analysis report meets the preset conditions of the decision triggering strategy instructions, matching and outputting maintenance decision suggestions from a preset maintenance knowledge base.

[0007] Optionally, the calculation of the dynamic health concern index based on the comprehensive equipment status data set includes: extracting abnormal characteristic parameters from the real-time operating status data, extracting failure frequency parameters from the historical maintenance records, and extracting equipment value parameters from the equipment basic information; using the abnormal characteristic parameters and the failure frequency parameters to generate a risk weight factor; fusing the equipment value parameter and the risk level parameter to generate a consequence severity coefficient; and performing weighted calculation on the risk weight factor and the consequence severity coefficient to generate a dynamic health concern index.

[0008] Optionally, the policy chain instruction set generated based on the dynamic health concern index, which includes monitoring strategy instructions, analysis strategy instructions and decision trigger strategy instructions, includes: matching the preset monitoring intensity rule library according to the dynamic health concern index to generate monitoring strategy instructions including data acquisition frequency parameters, sensor accuracy parameters and monitoring target parameters; activating the preset analysis depth rule library according to the dynamic health concern index to generate analysis strategy instructions including diagnostic model combination identifiers and analysis path identifiers; calling the preset decision trigger rule library based on the device type identifier in the device basic information to generate decision trigger strategy instructions including confidence thresholds and fault mode matching conditions; integrating the monitoring strategy instructions, the analysis strategy instructions and the decision trigger strategy instructions to obtain a policy chain instruction set.

[0009] Optionally, inputting the target monitoring data set and the analysis strategy instructions into a preset multi-level adaptive analysis tree, performing a hierarchical diagnosis operation, and generating a structured analysis report includes: selecting a primary analysis node or an advanced analysis node in the multi-level adaptive analysis tree according to the analysis path identifier in the analysis strategy instructions; performing a threshold verification operation on the target monitoring data set through the primary analysis node to generate a primary diagnosis identifier; when the primary diagnosis identifier meets a preset depth trigger condition, activating the advanced analysis node to perform multi-model fusion diagnosis to generate the structured analysis report.

[0010] Optionally, when the primary diagnostic identifier meets the preset deep trigger condition, the advanced analysis node is activated to perform multi-model fusion diagnosis, and the generation of the structured analysis report includes: when the primary diagnostic identifier meets the preset deep trigger condition, the advanced analysis node is activated and the preset machine learning model group is called according to the diagnostic model combination identifier; the target monitoring data set is input into the machine learning model group respectively to perform parallel analysis to generate a model output result set; a confidence-weighted fusion operation is performed on the model output result set to generate the fault root cause parameters and impact assessment parameters in the structured analysis report.

[0011] Optionally, the matching of maintenance decision suggestions from a preset maintenance knowledge base includes: extracting fault mode codes and equipment type identifiers from the structured analysis report; searching in a preset maintenance knowledge base based on the fault mode codes to obtain a set of candidate maintenance solutions; extracting operating condition parameters from the equipment basic information, and performing cost optimization screening using the operating condition parameters and the set of candidate maintenance solutions to generate maintenance decision suggestions.

[0012] Optionally, the method also includes: collecting the policy chain instruction set, the target monitoring data set, the structured analysis report and the maintenance execution record to generate a policy execution chain data set; extracting the equipment status improvement parameters from the maintenance execution record; calculating the policy effectiveness evaluation value based on the equipment status improvement parameters; when the policy effectiveness evaluation value is lower than the preset optimization trigger threshold, adjusting the weight coefficient of the risk weight factor and the consequence severity coefficient.

[0013] Optionally, adjusting the weight coefficients of the risk weight factor and the consequence severity coefficient includes: extracting false alarm cases and missed alarm cases from the policy execution chain data set to generate an optimized training sample set; correcting the weight coefficient of the risk weight factor based on the optimized training sample set; extracting actual maintenance cost parameters from the maintenance execution record; and using the actual maintenance cost parameters to correct the weight coefficient of the consequence severity coefficient.

[0014] Optionally, the method includes: feeding back the strategy effectiveness evaluation value to the multi-level adaptive analysis tree; adjusting the activation priority of the machine learning model group in the advanced analysis node according to the strategy effectiveness evaluation value; and optimizing the confidence threshold in the decision-triggering strategy instruction based on the fault verification result in the maintenance execution record.

[0015] Based on the same inventive concept, the present invention also provides a PMS-based equipment status monitoring and abnormality analysis system, which includes: a status data acquisition module, which is used to obtain equipment basic information, real-time operation status data, historical maintenance records and preset risk level parameters in the equipment management system to generate a comprehensive equipment status data set; a health indicator calculation module, which is used to calculate the dynamic health concern index based on the comprehensive equipment status data set; a policy instruction generation module, which is used to generate a policy chain instruction set including monitoring policy instructions, analysis policy instructions and decision triggering policy instructions based on the dynamic health concern index; a monitoring data acquisition module, which is used to schedule a preset sensor group to perform data acquisition operations according to the monitoring policy instructions to generate a target monitoring data set; a report generation module, which is used to input the target monitoring data set and the analysis policy instructions into a preset multi-level adaptive analysis tree, perform hierarchical diagnostic operations, and generate a structured analysis report; a decision output module, which is used to match and output maintenance decision suggestions from a preset maintenance knowledge base when the structured analysis report meets the preset conditions of the decision triggering policy instructions.

[0016] Compared with the prior art, the present invention has the following advantages: 1. This invention quantifies the risk level of equipment through a dynamic health concern index and automatically matches differentiated monitoring strategies, significantly reducing resource usage of low-risk equipment while ensuring refined monitoring of high-risk equipment, effectively optimizing sensor and computing resource utilization. 2. This invention uses a multi-level adaptive analysis tree to perform hierarchical diagnosis. Primary nodes quickly screen for anomalies, and advanced nodes fuse multiple models for in-depth analysis. This significantly reduces real-time computing load and avoids misjudgments and missed judgments, ensuring accurate location and root cause analysis of complex faults. 3. This invention combines the real-time operating conditions of the equipment with the maintenance knowledge base to perform cost optimization screening, generating maintenance recommendations that balance reliability and economy, avoiding over-maintenance or under-maintenance. At the same time, it dynamically controls output timing through decision-making triggering strategies to ensure timely response to critical faults. 4. The present invention dynamically modifies the health indicator weight coefficient, model priority and confidence threshold based on the strategy effectiveness evaluation value and maintenance execution record, continuously improves the risk prediction accuracy and decision reliability, and forms a complete feedback loop from execution to optimization.

[0017] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures pointed out in the description, claims and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0019] Figure 1 1 is a flow chart of a PMS-based device status monitoring and abnormality analysis strategy method according to an embodiment of the present invention.

[0020] Figure 2 3 is a schematic diagram of the structure of the PMS-based equipment status monitoring and abnormality analysis system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0022] Reference Figure 1 One embodiment of the present invention proposes a PMS-based equipment status monitoring and anomaly analysis strategy method, which uses a dynamic health concern index to drive the generation of strategy chain instructions. Combined with a multi-level adaptive analysis tree hierarchical diagnosis and closed-loop parameter optimization mechanism, it can achieve precise allocation of monitoring resources, accurate location of fault root causes, economic optimization of maintenance decisions, and continuous adaptive upgrades of the system.

[0023] The method of this embodiment specifically includes: Obtain basic equipment information, real-time operating status data, historical maintenance records, and preset risk level parameters from the equipment management system to generate a comprehensive equipment status data set; Specifically, the system extracts device model parameters, installation date parameters, and device value parameters from the device basic information database of the equipment management system; obtains time-series data on device operating status, including vibration amplitude, temperature, and current parameters, from the real-time data interface; retrieves historical fault count parameters, maintenance timestamp sequences, and maintenance action description text from the maintenance record database; and reads the risk level parameter from the preset configuration file, which is predefined as a discrete numeric identifier based on the importance of the device's function. These four types of data are then mapped based on the device's unique identifier. The device basic information is directly matched to the risk level parameter using the device code, real-time operating status data is aligned with the device's physical location coordinates and timestamps, and historical maintenance records are linked to the maintenance event index using the device serial number. During data fusion, non-numeric fields are converted to standardized encodings, and numeric fields are dimensionally normalized. This generates a structured data table based on the device. Each record in the table contains a device identification field, a set of basic information fields, a real-time status data matrix, an array of historical maintenance records, and a risk level field. This operation ensures the integration of heterogeneous data from multiple sources under a unified spatiotemporal reference, forming a comprehensive device status dataset that can be directly accessed by subsequent modules.

[0024] calculating a dynamic health concern index based on the comprehensive device status dataset; Generate a policy chain instruction set including monitoring policy instructions, analysis policy instructions and decision triggering policy instructions based on the dynamic health concern index; Scheduling a preset sensor group to perform data collection operations according to the monitoring strategy instructions to generate a target monitoring data set; Specifically, the monitoring strategy directive is parsed to obtain the data acquisition frequency parameter, sensor accuracy level parameter, and a list of monitoring target parameters. The data acquisition frequency parameter, expressed in Hertz (Hz), represents the number of samples per second and is directly read from the monitoring strategy directive. The sensor accuracy level parameter is a preset level identifier corresponding to sensors of varying accuracy. The specific sensor model is determined by querying a preset sensor accuracy mapping table. The monitoring target parameter list contains identifiers of the physical quantities to be monitored, such as vibration and temperature. These identifiers are bound to the device's measurement point code. Based on the sensor accuracy level parameter, sensor instances that meet the accuracy requirements are selected from a preset sensor group. The sensor group consists of sensor nodes deployed at different physical locations on the device, each with a unique device code and accuracy attributes. Based on the monitoring target parameter list, a subset of sensors in the sensor group that match the target physical quantity is selected. A control command containing the data acquisition frequency parameter and a start timestamp is sent to this selected sensor subset. The sensor nodes collect raw data according to the frequency and perform timestamp alignment on the raw data using a hardware clock synchronization protocol to generate a time series dataset. The time series dataset is normalized based on the raw sensor readings and the upper limit of the sensor range. Finally, the normalized data of each sensor node are merged according to the timestamp to generate a structured target monitoring dataset containing device code, timestamp sequence, monitored physical quantity identifier and the corresponding normalized time series dataset value.

[0025] Inputting the target monitoring data set and the analysis strategy instructions into a preset multi-level adaptive analysis tree, performing hierarchical diagnostic operations, and generating a structured analysis report; When the structured analysis report meets the preset conditions of the decision triggering policy instruction, maintenance decision suggestions are matched from a preset maintenance knowledge base and output.

[0026] Specifically, the system uses a comprehensive quantitative assessment of the equipment's status to drive adaptive adjustments to the monitoring strategy, and employs a hierarchical diagnostic mechanism to accurately locate anomalies. Specifically, the system first integrates the equipment's static attributes, real-time operating characteristics, historical maintenance records, and preset risk parameters to generate a dynamic health concern index, which quantifies the equipment's risk level in real time. Secondly, monitoring strategy instructions are dynamically generated based on the index value to precisely control the sensor group's acquisition frequency, accuracy, and monitoring targets. The collected data is then input into a multi-level analysis tree for hierarchical diagnosis. Primary nodes perform rapid threshold screening, while advanced nodes trigger multi-model fusion analysis. Finally, when the diagnostic results meet the preset trigger conditions, the system combines the equipment's operating conditions to match the most economically efficient maintenance decision from the knowledge base, forming a closed-loop control system from status perception to decision execution.

[0027] Optionally, calculating a dynamic health concern index based on the comprehensive device status data set includes: Extracting abnormal characteristic parameters from the real-time operating status data, extracting fault frequency parameters from the historical maintenance records, and extracting equipment value parameters from the equipment basic information; Generating a risk weight factor using the abnormal characteristic parameter and the fault frequency parameter; Fusion of the equipment value parameter and the risk level parameter to generate a consequence severity coefficient; The risk weight factor and the consequence severity coefficient are weighted and calculated to generate a dynamic health concern index.

[0028] Specifically, it is necessary to first extract specific data parameters from different sources, and extract abnormal characteristic parameters from real-time operating status data that can reflect whether the current operation of the equipment is normal, such as an abnormal increase in the vibration amplitude of the equipment, or the degree to which the operating temperature exceeds the normal range. At the same time, extract the failure frequency parameter from the historical maintenance records, which quantifies the frequency of failure of a specific equipment in the past period of time. Finally, extract the equipment value parameter from the basic information of the equipment, which represents the economic value of the equipment or its importance in the production process. The next step is to generate a risk weight factor, which aims to quantify the possibility of equipment failure. The calculation method is to combine the abnormal characteristic parameters monitored in real time with the historical failure frequency parameters. For the risk weight factor ,have: ; Where, It is the normalized abnormal characteristic parameter, reflecting the severity of the deviation of the current operating status from the normal baseline. This value is obtained from the real-time operating status data. It is the normalized fault frequency parameter, which reflects the frequency of historical equipment failures. This value is obtained from historical maintenance records. and It is a preset weight used to adjust the relative importance of real-time anomalies and historical failures in risk assessment. Then a consequence severity coefficient is generated to quantify the loss or impact that may be caused if a device fails. It is calculated by integrating the value of the device with the preset risk level. The risk level parameter here is a pre-set value, usually defined by experts based on the criticality of the device in the entire system, such as general, important, critical, etc. ,have: ; Where, It is the normalized equipment value parameter, representing the replacement cost of the equipment or the production loss caused by downtime. This value is obtained from the basic equipment information; It is the preset risk level parameter after normalization, reflecting the functional importance of the equipment; and It is a preset weight used to balance the economic value of the equipment itself and its functional criticality. Finally, the calculated risk weight factor and consequence severity coefficient are weighted to generate the final dynamic health concern index. This index combines the possibility of equipment failure and the severity of its consequences to form a comprehensive health status assessment indicator. ,have: ; Where, and A pre-set weighting factor balances the likelihood of a risk occurring with the severity of its consequences, typically summing the two to ensure consistent results. Using this index, the system dynamically assesses and prioritizes all equipment requiring attention. Rather than relying solely on real-time data thresholds for alarms, the system integrates multiple dimensions, including real-time status, historical performance, equipment value, and functional importance, to construct a more comprehensive and insightful equipment health assessment model. This approach accurately distinguishes between equipment exhibiting the same abnormal signal but of varying importance. For example, a minor vibration in an auxiliary cooling fan versus a minor pressure fluctuation in a core reactor would receive a significantly higher dynamic health attention index due to its significantly higher severity coefficient. This allows maintenance resources to be precisely directed to equipment with the highest current risk and the greatest potential for loss, enabling a shift from reactive response to proactive prevention, and from extensive management to precise decision-making. This significantly improves equipment management efficiency and the overall reliability of the production system.

[0029] Optionally, generating a policy chain instruction set including monitoring policy instructions, analysis policy instructions, and decision triggering policy instructions based on the dynamic health concern index includes: According to the dynamic health concern index, a preset monitoring intensity rule library is matched to generate a monitoring strategy instruction including data collection frequency parameters, sensor accuracy parameters and monitoring target parameters; Activating a preset analysis depth rule library according to the dynamic health concern index to generate an analysis strategy instruction including a diagnosis model combination identifier and an analysis path identifier; Based on the device type identifier in the device basic information, a preset decision trigger rule library is called to generate a decision trigger strategy instruction including a confidence threshold and a fault mode matching condition; The monitoring strategy instructions, the analysis strategy instructions and the decision triggering strategy instructions are integrated to obtain a strategy chain instruction set.

[0030] Specifically, the dynamic health concern index is first matched from a preset monitoring intensity rule library. This rule library is a two-dimensional mapping table that stores data acquisition frequency parameters, sensor accuracy parameters, and monitoring target parameters corresponding to different index intervals. The data acquisition frequency parameter is in Hz, the sensor accuracy parameter is a percentage value, and the monitoring target parameter is a string array that identifies key monitoring points. The matching process uses linear interpolation to determine the specific parameter values ​​to ensure the continuity of the index and parameter mapping. Next, the preset analysis depth rule library is activated. This rule library has a decision tree structure and stores diagnostic model combination identifiers and analysis path identifiers corresponding to different index thresholds. The diagnostic model combination identifier is a coded string pointing to a preset machine learning model group such as a random forest or neural network, and the analysis path identifier is a Boolean value indicating whether to select a primary or advanced node. During activation, the index is compared with the preset threshold set and the corresponding identifier is selected. Subsequently, the device type identifier (a classification code such as motor or transformer) is extracted from the basic device information. A preset decision trigger rule library, a hash table structure that stores confidence thresholds and fault mode matching conditions corresponding to different device types, is called. The confidence threshold is a floating-point number between 0 and 1, and the fault mode matching condition is a string pattern. The calling process directly retrieves the identifier to obtain parameters. Finally, the system integrates the generated monitoring strategy instructions, analysis strategy instructions, and decision trigger strategy instructions to form a policy chain instruction set. This integration method uses data structure serialization operations to encapsulate the instruction parameters as JSON-formatted objects. The health concern index dynamically quantifies device risks, adaptively adjusts monitoring and analysis resource allocation, optimizes device management efficiency, avoids resource waste on low-risk devices, and ensures timely diagnosis and accurate decision-making for high-risk devices, improving system response speed and maintenance accuracy.

[0031] Optionally, inputting the target monitoring data set and the analysis strategy instructions into a preset multi-level adaptive analysis tree, performing hierarchical diagnosis operations, and generating a structured analysis report includes: selecting a primary analysis node or a high-level analysis node in the multi-level adaptive analysis tree according to the analysis path identifier in the analysis strategy instruction; Performing a threshold check operation on the target monitoring data set by the primary analysis node to generate a primary diagnosis identifier; When the primary diagnosis identifier meets the preset depth trigger condition, the advanced analysis node is activated to perform multi-model fusion diagnosis and generate the structured analysis report.

[0032] Specifically, first parse the analysis path identifier in the analysis strategy instruction. This identifier is a Boolean variable. When the value is true, the advanced analysis node is selected, and when the value is false, the primary analysis node is selected. In the data structure of the multi-level adaptive analysis tree, the primary analysis node stores the preset threshold rule library, and the advanced analysis node stores the machine learning model group interface. When the primary analysis node is selected, the real-time monitoring parameters in the target monitoring data set are compared with the warning threshold in the threshold rule library. The warning threshold is determined by the equipment technical manual and historical operation data statistics. The comparison process uses piecewise function processing. For the temperature parameter , when satisfied Overheat alarm indicator, for vibration parameters , when satisfied Mechanical failure identification, and These are the upper temperature and vibration thresholds, respectively. All alarm flags are combined to form a primary diagnostic flag. When the severity parameter of a primary diagnostic flag exceeds the preset depth trigger threshold, the advanced analysis node is activated. The severity parameter is weighted based on the number and type of alarms. The depth trigger threshold is obtained from the decision trigger rule base based on the device type. This layered diagnosis enables dynamic resource allocation, allowing primary nodes to quickly screen for anomalies and advanced nodes to accurately locate faults, reducing real-time computing load while ensuring the accuracy of in-depth analysis of critical faults.

[0033] Optionally, when the primary diagnosis identifier satisfies a preset depth trigger condition, activating the advanced analysis node to perform multi-model fusion diagnosis and generating the structured analysis report includes: When the primary diagnosis identifier satisfies a preset deep trigger condition, the advanced analysis node is activated and a preset machine learning model group is called according to the diagnosis model combination identifier; Inputting the target monitoring data set into the machine learning model group respectively to perform parallel analysis and generate a model output result set; A confidence-weighted fusion operation is performed on the model output result set to generate fault root cause parameters and impact assessment parameters in the structured analysis report.

[0034] Specifically, first verify whether the severity level parameter of the primary diagnostic identifier is greater than the deep trigger threshold. If the condition is met, activate the advanced analysis node; according to the diagnostic model combination identifier in the analysis strategy instruction, call the corresponding model group from the preset machine learning model library. The identifier is a string array, such as "CNN-SVM" pointing to the combination of convolutional neural network and support vector machine. The target monitoring data set is synchronously input into each model for parallel analysis. The convolutional neural network processes the vibration spectrum image data and outputs the fault probability. , support vector machine processes temperature time series data to output fault probability ; All model output results form the model output result set. Confidence weighted fusion is performed on the result set, and the weight of each model Based on historical verification accuracy, , for the probability of root cause of failure ,have: ; Where, is the weight of the i-th model, is the fault probability output by the i-th model. The fault root cause parameter is set to the fault type code corresponding to the maximum probability, and the impact assessment parameter is set to the probability value itself. The structured analysis report encapsulates these parameters in XML format, including fields for fault location and confidence. Multi-model complementary analysis improves diagnostic robustness, reduces the risk of misjudgment by a single model, and enhances the ability to identify complex faults and the reliability of the results.

[0035] Optionally, matching maintenance decision suggestions from a preset maintenance knowledge base includes: extracting a failure mode code and a device type identifier from the structured analysis report; Searching a preset maintenance knowledge base according to the failure mode code to obtain a set of candidate maintenance solutions; Operating condition parameters are extracted from the basic equipment information, and cost optimization screening is performed using the operating condition parameters and the candidate maintenance solution set to generate maintenance decision recommendations.

[0036] Specifically, first, the system parses the structured analysis report to extract the fault mode code and equipment type identification; the fault mode code is an alphanumeric combination in the standard fault classification system, such as BRG001 for bearing wear, and the equipment type identification is the equipment classification code, such as TM for transformer. According to the fault mode code, a full-text search is performed in the maintenance knowledge base. The maintenance knowledge base is stored in a graph database, with nodes as fault modes and edges connecting candidate maintenance plans; the search returns a set of candidate maintenance plans, including plan descriptions, standard working hours, spare parts lists, and historical validity parameters. Subsequently, operating condition parameters are extracted from the basic equipment information, including quantitative indicators such as load rate, ambient temperature, and continuous operating time; cost optimization screening is performed on the candidate maintenance plan set to calculate the comprehensive cost of each plan. : ; in Get real-time quotes for spare parts costs based on spare parts lists. Labor costs are calculated based on standard working hours and unit labor hour rates. The downtime loss is estimated based on the equipment capacity and downtime duration; 、 、 is the adaptive weight coefficient of the working condition, which is determined by the load rate Normalized calculation: ; Ultimately, the solution with the lowest overall cost is selected to generate maintenance decision recommendations. By integrating fault characteristics with real-time operating conditions for economic evaluation, resource allocation is optimized while ensuring maintenance reliability, avoiding secondary losses caused by over- or under-maintenance.

[0037] Optionally, the method further includes: Collecting the policy chain instruction set, the target monitoring data set, the structured analysis report and the maintenance execution record to generate a policy execution chain data set; equipment status improvement parameters extracted from the maintenance execution record; Calculating a strategy effectiveness evaluation value based on the equipment status improvement parameter; When the strategy effectiveness evaluation value is lower than a preset optimization trigger threshold, the weight coefficients of the risk weight factor and the consequence severity coefficient are adjusted.

[0038] Optionally, the weight coefficient for adjusting the risk weight factor and the consequence severity coefficient includes: Extracting false positive cases and missed negative cases from the policy execution chain dataset to generate an optimized training sample set; Modifying the weight coefficient of the risk weight factor according to the optimized training sample set; extracting actual maintenance cost parameters from the maintenance execution record; The actual maintenance cost parameter is used to modify the weight coefficient of the consequence severity coefficient.

[0039] Specifically, first collect the policy chain instruction set, target monitoring data set, structured analysis report and maintenance execution record, and serialize these data through the data integration module to generate the policy execution chain data set; the serialization process is encapsulated in JSON format to ensure that the data fields include instruction parameters, monitoring values, report content and maintenance logs. Then, extract the equipment status improvement parameters from the maintenance execution record, including the percentage of failure rate reduction and the average trouble-free operation time increment. The policy effectiveness evaluation value is obtained by calculating the difference between the equipment status improvement parameters before and after maintenance; the policy effectiveness evaluation value is defined as the normalized average value of multiple equipment status improvement parameters. For the policy effectiveness evaluation value ,have: ; in Indicates the status value after maintenance, Indicates the status value before maintenance. Improve the number of parameters for device status, and Extracted directly from the maintenance execution records, all values ​​are normalized to the range of [0,1]. When the strategy effectiveness evaluation value is lower than the preset optimization trigger threshold, false alarm cases and missed cases are extracted from the strategy execution chain data set. The optimization trigger threshold is a preset constant in the range of [0.6,0.8]. False alarm cases are defined as records that indicate faults in the structured analysis report but are actually fault-free, and missed cases are defined as records that are not reported but actually have faults. The extraction method is based on data field comparison to generate an optimized training sample set. The optimized training sample set contains case feature vectors and labels. The weight coefficient of the risk weight factor is corrected according to the optimized training sample set. The correction process includes calculating the false alarm rate FP and missed alarm rate FN of the optimized training sample set. The false alarm rate is the number of false alarm cases divided by the total number of cases, and the missed alarm rate is the number of missed alarm cases divided by the total number of cases. For the weight coefficient of the updated risk weight factor ,have: ; Where, is the weight coefficient of the current risk weight factor, is the false alarm rate, is the underreporting rate, The learning rate is preset in the range of [0.01, 0.1]. At the same time, the actual maintenance cost parameter is extracted from the maintenance execution record. This parameter is a normalized comprehensive cost value, including spare parts cost, labor cost and downtime loss. For the actual maintenance cost parameter ,have: ; Where, 、 、 Extracted from the maintenance execution record fields, is the preset maximum cost benchmark value. Finally, the weight coefficient of the consequence severity coefficient is modified using the actual maintenance cost parameter. The weight coefficient of the modified consequence severity coefficient is ,have: ; in is the current weight coefficient, The preset adjustment coefficient range is [0.05, 0.1]. Normalized. Dynamically optimize the weight coefficients through a closed-loop feedback mechanism to improve the accuracy of the dynamic health concern index; enhance the system's adaptability, reduce the incidence of false positives and missed positives, optimize maintenance resource allocation, improve equipment failure prediction accuracy, and reduce overall maintenance costs.

[0040] Optionally, the method comprises: Feeding back the strategy effectiveness evaluation value to the multi-level adaptive analysis tree; Adjusting the activation priority of the machine learning model group in the advanced analysis node according to the strategy effectiveness evaluation value; A confidence threshold in the decision triggering strategy instruction is optimized based on the fault verification result in the maintenance execution record.

[0041] Specifically, when feeding back the strategy effectiveness evaluation value to the multi-level adaptive analysis tree, it is necessary to open a shared memory area in the analysis tree data structure to store the strategy effectiveness evaluation value; this value is transmitted to the analysis tree control module through the data bus, and the control module parses the data according to the preset feedback interface protocol and updates the status register of the analysis tree. When adjusting the activation priority of the machine learning model group in the advanced analysis node according to the strategy effectiveness evaluation value, the following operations are performed: read the strategy effectiveness evaluation value from the status register of the analysis tree and compare it with the preset priority mapping table; the priority mapping table stores the correspondence between the strategy effectiveness evaluation value interval and the model group weight coefficient, and uses linear interpolation to calculate the new weight coefficient of each model. Updated model group weight coefficient ,have: ; in is the current weight coefficient of the i-th model (the initial value comes from the model historical accuracy database), is the strategy effectiveness evaluation value (calculated from the equipment status improvement parameters before and after maintenance), is the preset learning rate adjustment factor (range [0.01, 0.1]). The updated model group weight coefficients are used to reorder the model group calls in descending order, with high-weight models being prioritized for analysis. When optimizing the confidence threshold in the decision triggering strategy instruction based on the fault verification results in the maintenance execution record, the actual fault type and the diagnostic results of the structured analysis report are analyzed from the maintenance execution record, and the false alarm rate and missed alarm rate are calculated. For the optimized confidence threshold ,have: ; Where, is the current confidence threshold (the initial value comes from the decision trigger rule base), To optimize the step size coefficient (range [0.05, 0.2]), and By comparing actual maintenance records with diagnostic reports, ensure reasonable range.

[0042] Based on the same inventive concept, Figure 2 As shown, the present invention also provides a PMS-based equipment status monitoring and abnormality analysis system, the system comprising: The status data acquisition module is used to obtain basic equipment information, real-time operating status data, historical maintenance records, and preset risk level parameters from the equipment management system to generate a comprehensive equipment status data set; a health index calculation module, configured to calculate a dynamic health concern index based on the comprehensive device status data set; A policy instruction generation module, configured to generate a policy chain instruction set including monitoring policy instructions, analysis policy instructions, and decision triggering policy instructions based on the dynamic health concern index; A monitoring data acquisition module is used to schedule a preset sensor group to perform data acquisition operations according to the monitoring strategy instructions to generate a target monitoring data set; A report generation module, configured to input the target monitoring data set and the analysis strategy instructions into a preset multi-level adaptive analysis tree, perform hierarchical diagnostic operations, and generate a structured analysis report; The decision output module is used to match and output maintenance decision suggestions from a preset maintenance knowledge base when the structured analysis report meets the preset conditions of the decision triggering policy instruction.

[0043] Example 1, in order to verify the effectiveness of the method described in the present invention, this example takes a regional natural gas distribution station as the application scenario. The distribution station undertakes the key task of stably delivering natural gas to downstream urban pipelines and industrial users, and the safety and reliability of its operation are of vital importance. Any equipment failure, especially the unplanned shutdown of the core compressor unit, may lead to large-scale gas supply interruption and bring serious safety hazards. The traditional operation and maintenance mode of the station relies on fixed maintenance cycles and manual judgment based on coarse-grained alarms of the SCADA system, which makes it difficult to effectively prevent emergencies caused by early hidden faults. The present invention is intended to be applied to the status monitoring and anomaly analysis of the core equipment of the distribution station to achieve predictive maintenance and ensure the safe operation of the pipeline network.

[0044] In this example, the system first uses the status data acquisition module to obtain basic information, including historical maintenance records (including a previous incident of reduced gas transmission efficiency due to piston ring wear within the past 24 months), and pre-set risk level parameters for a core reciprocating compressor unit (model: Ariel-JGK-4, asset value: RMB 8 million, criticality level: Highest) from the station control SCADA system. Furthermore, sensors deployed on the compressor body and inlet and outlet pipes collect real-time operating status data, including body vibration, inlet and outlet pressures, outlet temperature, and motor current, to form a comprehensive equipment status dataset.

[0045] The health indicator calculation module calculates a dynamic health concern index based on this data set. For example, during a monitoring session at 3:30 PM on September 20, 2024, the system extracts from real-time operating status data an abnormal, slight fluctuation in the compressor outlet pressure, with the pressure ratio to the inlet pressure deviating by 3% from the normal operating curve. This serves as the abnormal characteristic parameter. From historical maintenance records, the system also extracts a failure frequency of 0.5 times per year as the failure frequency parameter. Furthermore, the system extracts the equipment value parameter (normalized to 0.95) and the risk level parameter (normalized to 1.0) from basic equipment information. Using the abnormal characteristic parameter (normalized to 0.5) and the failure frequency parameter (normalized to 0.6), the system calculates a risk weight factor (R) of 0.5*0.5+0.5*0.6=0.55. Furthermore, by integrating the equipment value parameter and the risk level parameter, the system generates a consequence severity coefficient (C) of 0.6*0.95+0.4*1.0=0.97. Finally, a weighted calculation of the two yields a dynamic health concern index (H) of 0.6*R+0.4*C=0.6*0.55+0.4*0.97=0.718. This index is significantly higher than that of other equipment on the station (a spare filter has a dynamic health concern index of only 0.18), indicating that this compressor unit is the equipment that currently requires the most attention.

[0046] Based on the health concern index of 0.718, the strategy instruction generation module generated a strategy chain instruction set. Monitoring strategy instruction: match the monitoring intensity rule base, increase the data collection frequency of the compressor outlet pressure and fuselage vibration sensor from the conventional once per minute to once per second, and dispatch high-frequency acoustic sensors deployed in the valve area for special monitoring. Analysis strategy instruction: activate the analysis depth rule base, select the advanced analysis path, and generate the diagnostic model combination identifier "Wavelet+GRU" for deep fault diagnosis. Decision trigger strategy instruction: call the decision trigger rule base of the compressor group, set the confidence threshold of fault diagnosis to 0.90, and focus the fault mode matching conditions on "valve internal leakage, piston ring wear, and bearing failure."

[0047] Based on the monitoring strategy, the monitoring data acquisition module dispatched relevant sensors to perform 15 minutes of high-frequency data collection, generating a high-resolution target monitoring dataset containing pressure, vibration, and acoustic signals. The report generation module then fed this dataset and analysis strategy into a multi-level adaptive analysis tree. For hierarchical diagnosis, since the analysis path was identified as advanced, the system directly activated the advanced analysis node. For multi-model fusion diagnosis, the system invoked a pre-configured "Wavelet + GRU" machine learning model suite. The Wavelet transform (Wavelet) performs time-frequency analysis on high-frequency pressure and acoustic signals to extract subtle leakage signatures, while the Gated Recurrent Unit (GRU) network models the time series of vibration and pressure ratio to analyze their dynamic evolution trends. By applying confidence-weighted fusion to the model outputs, the system generated a structured analysis report, concluding: "Root Cause: Minor internal leakage in the compressor's second-stage outlet valve; Confidence Level: 0.92; Impact Assessment: Currently causing a 4% decrease in gas transmission efficiency. If unaddressed, the leakage is expected to worsen within 15 days, resulting in an efficiency drop of over 10% and the risk of valve fracture."

[0048] Because the reported confidence level of 0.92 exceeded the threshold of 0.90 set by the decision triggering strategy instruction, the decision output module was activated. Standard operating procedures related to "reciprocating compressor valve internal leakage" were retrieved from the maintenance knowledge base. Combined with the downstream gas load forecast provided by the SCADA system (the next week is a low-peak gas demand period), cost optimization screening was performed, and the final maintenance decision recommendation was output: "It is recommended to handle it during the next scheduled maintenance window (estimated in 7 days). Recommended maintenance plan: Replace the secondary outlet valve assembly (spare part number V-341-B), with an estimated downtime of 2 hours. During this period, maintain high-frequency monitoring. If the efficiency drops by more than 7%, immediately start the backup compressor and perform emergency maintenance."

[0049] In addition, the present invention also demonstrates closed-loop self-optimization capabilities. After the valve replacement maintenance was completed, the system collected maintenance execution records, confirmed that the fault was consistent with the diagnosis, and the equipment's gas transmission efficiency returned to normal. The system calculated that the strategy effectiveness evaluation value of this strategy was 0.95, and this high score was fed back to the system to improve the priority of the "Wavelet+GRU" model in diagnosing valve internal leaks. In another incident, the system issued an abnormal warning for the pressure sensor of a certain section of the pipeline, but after manual verification on site, it was confirmed that it was a false alarm caused by sensor zero drift. The system recorded this as a false alarm case, and automatically fine-tuned the calculation model of the risk weight factor based on the data, increased the weight of cross-validation of multi-sensor data, and effectively improved the robustness of the future health attention index.

[0050] Data comparisons show that the application of this invention has significantly improved the safety management and operational efficiency of this natural gas distribution station. The system can provide 15 days' advance warning of hidden compressor unit failures, transforming potential unplanned downtime into planned maintenance and avoiding major gas supply incidents. By dynamically adjusting monitoring strategies, while maintaining the accuracy of critical equipment monitoring, the overall data transmission and computing overhead is reduced by approximately 50%. Accurate maintenance decisions not only reduce the cost of each repair but also minimize the impact on downstream gas supply by optimizing maintenance timing.

[0051] Table 1 Dynamic health attention index of sub-transmission station equipment: ; Table 2 Policy chain instruction adaptive adjustment: ; Table 3 Fault diagnosis and decision-making suggestions: ;

[0052] As can be seen from the data in Tables 1 to 3 above, the present invention achieves differentiated intelligent monitoring of different equipment within a distribution station through dynamic risk quantification. Table 1 clearly demonstrates that core compressor units showing early signs of failure receive the highest attention. Table 2 illustrates how the system adaptively allocates monitoring and analysis resources accordingly. The results in Table 3 demonstrate the present invention's high accuracy in diagnosing complex industrial equipment faults and its forward-looking decision-making. It can provide specific, optimized maintenance plans tailored to operating conditions, successfully upgrading the operation and maintenance model from passive response to active prediction, and providing solid technical support for the safe, stable, and efficient operation of the natural gas pipeline network.

[0053] It should be noted that the formulas appearing above can translate physical quantities of different properties into unitless standard values ​​or superimposable parameters of the same dimension through the principle of dimensional consistency and mathematical standardization means (such as normalization, dimensionless parameter conversion or unit system unification), thereby eliminating the interference of different dimensions on the operation logic, so that the formulas have mathematical operation rationality and objective law adaptability while retaining the distribution characteristics of the original data. It is a conventional technical means and will not be elaborated here. The electrical connection between the above-mentioned units does not necessarily mean a direct connection of the circuit. The indirect connection method can be applied to the embodiments of the present invention as long as the purpose of the present invention is achieved. The above is only an exemplary embodiment of the present invention and the scope of the present invention cannot be limited thereto.

[0054] That is, any equivalent changes and modifications made according to the teachings of the present invention are still within the scope of the present invention. Those skilled in the art will readily conceive of other embodiments of the present invention after considering the disclosure of the specification and practical truths. This application is intended to cover any variations, uses, or adaptations of the present invention that follow the general principles of the present invention and include common knowledge or customary technical means in the art not described herein.

Claims

1. The equipment status monitoring and abnormality analysis strategy method based on PMS is characterized by: The method comprises: Obtain basic equipment information, real-time operating status data, historical maintenance records, and preset risk level parameters from the equipment management system to generate a comprehensive equipment status data set; calculating a dynamic health concern index based on the comprehensive device status dataset; Generate a policy chain instruction set including monitoring policy instructions, analysis policy instructions and decision triggering policy instructions based on the dynamic health concern index; Scheduling a preset sensor group to perform data collection operations according to the monitoring strategy instructions to generate a target monitoring data set; Inputting the target monitoring data set and the analysis strategy instructions into a preset multi-level adaptive analysis tree, performing hierarchical diagnostic operations, and generating a structured analysis report; When the structured analysis report meets the preset conditions of the decision triggering policy instruction, maintenance decision suggestions are matched from a preset maintenance knowledge base and output.

2. The PMS-based equipment status monitoring and abnormality analysis strategy method according to claim 1 is characterized in that: Calculating the dynamic health concern index according to the comprehensive device status data set includes: Extracting abnormal characteristic parameters from the real-time operating status data, extracting fault frequency parameters from the historical maintenance records, and extracting equipment value parameters from the equipment basic information; Generating a risk weight factor using the abnormal characteristic parameter and the fault frequency parameter; Fusion of the equipment value parameter and the risk level parameter to generate a consequence severity coefficient; The risk weight factor and the consequence severity coefficient are weighted and calculated to generate a dynamic health concern index.

3. The PMS-based equipment status monitoring and abnormality analysis strategy method according to claim 2 is characterized in that: The generation of a policy chain instruction set including monitoring policy instructions, analysis policy instructions, and decision triggering policy instructions based on the dynamic health concern index includes: According to the dynamic health concern index, a preset monitoring intensity rule library is matched to generate a monitoring strategy instruction including data collection frequency parameters, sensor accuracy parameters and monitoring target parameters; Activating a preset analysis depth rule library according to the dynamic health concern index to generate an analysis strategy instruction including a diagnosis model combination identifier and an analysis path identifier; Based on the device type identifier in the device basic information, a preset decision trigger rule library is called to generate a decision trigger strategy instruction including a confidence threshold and a fault mode matching condition; The monitoring strategy instructions, the analysis strategy instructions and the decision triggering strategy instructions are integrated to obtain a strategy chain instruction set.

4. The PMS-based equipment status monitoring and abnormality analysis strategy method according to claim 3 is characterized in that: The step of inputting the target monitoring data set and the analysis strategy instructions into a preset multi-level adaptive analysis tree, performing hierarchical diagnosis operations, and generating a structured analysis report includes: selecting a primary analysis node or a high-level analysis node in the multi-level adaptive analysis tree according to the analysis path identifier in the analysis strategy instruction; Performing a threshold check operation on the target monitoring data set by the primary analysis node to generate a primary diagnosis identifier; When the primary diagnosis identifier meets the preset depth trigger condition, the advanced analysis node is activated to perform multi-model fusion diagnosis and generate the structured analysis report.

5. The PMS-based equipment status monitoring and abnormality analysis strategy method according to claim 4 is characterized in that: When the primary diagnosis identifier satisfies a preset depth trigger condition, activating the advanced analysis node to perform multi-model fusion diagnosis and generating the structured analysis report includes: When the primary diagnosis identifier satisfies a preset deep trigger condition, the advanced analysis node is activated and a preset machine learning model group is called according to the diagnosis model combination identifier; Inputting the target monitoring data set into the machine learning model group respectively to perform parallel analysis and generate a model output result set; A confidence-weighted fusion operation is performed on the model output result set to generate fault root cause parameters and impact assessment parameters in the structured analysis report.

6. The PMS-based equipment status monitoring and abnormality analysis strategy method according to claim 1 is characterized in that: The matching of maintenance decision suggestions from a preset maintenance knowledge base includes: extracting a failure mode code and a device type identifier from the structured analysis report; Searching a preset maintenance knowledge base according to the failure mode code to obtain a set of candidate maintenance solutions; Operating condition parameters are extracted from the basic equipment information, and cost optimization screening is performed using the operating condition parameters and the candidate maintenance solution set to generate maintenance decision recommendations.

7. The PMS-based equipment status monitoring and abnormality analysis strategy method according to claim 4 is characterized in that: The method further comprises: Collecting the policy chain instruction set, the target monitoring data set, the structured analysis report and the maintenance execution record to generate a policy execution chain data set; equipment status improvement parameters extracted from the maintenance execution record; Calculating a strategy effectiveness evaluation value based on the equipment status improvement parameter; When the strategy effectiveness evaluation value is lower than a preset optimization trigger threshold, the weight coefficients of the risk weight factor and the consequence severity coefficient are adjusted.

8. The PMS-based equipment status monitoring and abnormality analysis strategy method according to claim 7 is characterized in that: The weight coefficients for adjusting the risk weight factor and the consequence severity coefficient include: Extracting false positive cases and missed negative cases from the policy execution chain dataset to generate an optimized training sample set; Modifying the weight coefficient of the risk weight factor according to the optimized training sample set; extracting actual maintenance cost parameters from the maintenance execution record; The actual maintenance cost parameter is used to modify the weight coefficient of the consequence severity coefficient.

9. The PMS-based equipment status monitoring and abnormality analysis strategy method according to claim 7 is characterized in that: The method comprises: Feeding back the strategy effectiveness evaluation value to the multi-level adaptive analysis tree; Adjusting the activation priority of the machine learning model group in the advanced analysis node according to the strategy effectiveness evaluation value; A confidence threshold in the decision triggering strategy instruction is optimized based on the fault verification result in the maintenance execution record.

10. A PMS-based equipment status monitoring and anomaly analysis system, applying the PMS-based equipment status monitoring and anomaly analysis strategy method according to any one of claims 1 to 9, characterized in that: The system comprises: The status data acquisition module is used to obtain basic equipment information, real-time operating status data, historical maintenance records, and preset risk level parameters from the equipment management system to generate a comprehensive equipment status data set; a health index calculation module, configured to calculate a dynamic health concern index based on the comprehensive device status data set; A policy instruction generation module, configured to generate a policy chain instruction set including monitoring policy instructions, analysis policy instructions, and decision triggering policy instructions based on the dynamic health concern index; A monitoring data acquisition module is used to schedule a preset sensor group to perform data acquisition operations according to the monitoring strategy instructions to generate a target monitoring data set; A report generation module, configured to input the target monitoring data set and the analysis strategy instructions into a preset multi-level adaptive analysis tree, perform hierarchical diagnostic operations, and generate a structured analysis report; The decision output module is used to match and output maintenance decision suggestions from a preset maintenance knowledge base when the structured analysis report meets the preset conditions of the decision triggering policy instruction.

Images