A fault early warning method, system, product and medium of a dry gas seal system

Abstract

A fault early warning method, system, product, and medium for a dry gas sealing system are disclosed, relating to the field of condition monitoring technology. The method includes: acquiring multiple monitoring parameters of the target dry gas sealing system and classifying them into first, second, and third categories; determining a more sensitive fluctuation allowable range within a safe operating range based on historical normal operating data; triggering a fault diagnosis process if a first-category parameter is detected to exceed the fluctuation allowable range; inputting the multiple monitoring parameters into a fault tree model to match candidate failure modes, and calling a target operator to perform mechanistic verification of the physical correlation between the time-series data of the multiple categories of parameters; finally, adjusting the confidence level of the candidate failure modes based on the verification results and outputting an early warning. This application effectively solves the problems of difficulty in detecting early, weak faults and easy false alarms due to fluctuations in normal operating conditions by combining sensitive range triggering with multi-dimensional parameter time-series correlation cross-validation, thus improving the accuracy of early fault early warning for dry gas sealing systems.

Description

Technical Field

[0001] This application relates to the field of condition monitoring technology, and in particular to a fault early warning method, system, product and medium for a dry gas sealing system. Background Technology

[0002] Dry gas seals, as a core non-contact sealing technology for shaft ends in large rotating machinery (such as centrifugal compressors), are directly related to the effective isolation of flammable, explosive, or toxic process media. Real-time monitoring and fault early warning of the operating status of dry gas seal systems are of vital engineering significance for preventing catastrophic leakage accidents, ensuring long-term safe operation of equipment, and reducing the risk of unplanned shutdowns.

[0003] Currently, monitoring and early warning for dry gas seals are mainly achieved by setting fixed thresholds in the control system. Specifically, this technology pre-sets absolute alarm thresholds for key monitoring parameters such as the differential pressure between the sealing gas and the process gas or isolation gas, as well as the pressure or flow rate of the leaking gas. During equipment operation, the control system compares the collected process parameters with the set alarm thresholds in real time. When the value of a process parameter reaches or exceeds the alarm threshold, the control system determines that there is an operational risk to the equipment and outputs an alarm signal or executes a shutdown interlock command.

[0004] However, in actual production scenarios, the unit frequently adjusts according to production load (such as changes in speed and process pressure), causing normal fluctuations in sealing parameters. Since the parameter variations caused by early faults are extremely weak, they usually fail to trigger the preset absolute alarm threshold. If the redundancy range of the alarm threshold is forcibly narrowed to meet early warning requirements (i.e., increasing the sensitivity of the alarm action), reasonable parameter fluctuations generated when the unit performs normal operating condition adjustments will easily exceed the narrowed threshold boundary. This increases the probability that the control system will misjudge normal operating condition fluctuations as equipment faults, reducing the accuracy of the warning results. Summary of the Invention

[0005] This application provides a fault early warning method, system, product, and medium for dry gas sealing systems, which can improve the accuracy of early fault warning for dry gas sealing systems.

[0006] In a first aspect, this application provides a fault early warning method for a dry gas sealing system, applied to a condition monitoring system, comprising: acquiring multiple monitoring parameters of a target dry gas sealing system, and dividing the multiple monitoring parameters into a first type of parameter, a second type of parameter, and a third type of parameter; the first type of parameter is a parameter characterizing the operating state of the target dry gas sealing system itself, the second type of parameter is a parameter characterizing the operating state of the unit to which the target dry gas sealing system belongs, and the third type of parameter is a parameter characterizing the state of the process medium transported by the target dry gas sealing system; based on the historical normal operation data of the target dry gas sealing system, determining a fluctuation allowable range within a preset safe operating range, wherein the width of the fluctuation allowable range is smaller than the width of the safe operating range; when the real-time monitored first... When the current value of a parameter exceeds the allowable fluctuation range, a fault diagnosis process is triggered. In response to the triggering of the fault diagnosis process, the first, second, and third types of parameters are input into a preset fault tree model for logical reasoning. From the multiple failure modes contained in the fault tree model, at least one candidate failure mode is matched. Based on the candidate failure mode, a target operator is called from a preset operator library, and the correlation between the time-series data of the first type of parameter and the time-series data of the second and / or third type of parameter is verified through the target operator to obtain the verification result. Based on the verification result, the confidence level of the candidate failure mode is adjusted, and based on the adjusted confidence level, the fault warning information corresponding to the target dry gas sealing system is output.

[0007] By adopting the above technical solution, the condition monitoring system can determine a more sensitive fluctuation allowable range based on multiple monitoring parameters of the target dry gas sealing system itself, its affiliated unit, and the process medium being transported, within the safe operating range defined by its historical normal operation data. When the first type of parameter exceeds this fluctuation allowable range, a fault diagnosis process is triggered. Then, a fault tree model is used to match candidate failure modes, and operators are invoked to verify the physical mechanism level of the correlation between the time-series data of various parameters. Based on the verification results, the confidence level is adjusted and a warning message is output. This solution, without triggering traditional absolute alarms that lead to false shutdowns, improves the accuracy of equipment condition warning results by reasoning about the causal and correlation relationships between multi-dimensional parameters and locating subtle early faults from abnormal fluctuations of a single parameter.

[0008] In conjunction with some embodiments of the first aspect, in some embodiments, based on the historical normal operation data of the dry gas sealing system, a fluctuation allowable range is determined within a preset safe operating range. Specifically, this includes: acquiring historical data of the first type of parameters collected during normal operation of the target dry gas sealing system under various historical operating conditions; grouping the historical data according to the operating condition type to obtain a historical data set corresponding to each operating condition type; for each historical data set, determining the high alarm value and low alarm value of the first type of parameters under the operating condition type based on the distribution range of the first type of parameters in the historical data set; and using the interval between the high alarm value and the low alarm value as the fluctuation allowable range corresponding to the operating condition type.

[0009] By adopting the above technical solution, the condition monitoring system acquires historical data of the first type of parameters of the target dry gas sealing system under various historical operating conditions, and groups them according to the operating condition type. Based on the distribution range of each historical data set, high and low alarm values ​​are determined for the corresponding operating conditions, thereby constructing a fluctuation allowable range specific to each operating condition type. This solution refines a unified, broad threshold into a dynamic, narrow-band benchmark strongly correlated with specific operating conditions, eliminating the interference of operating condition differences on the normal fluctuation range of parameters, and enabling the condition monitoring system to capture anomalies with sensitivity dynamically matched to the current operating condition.

[0010] In some embodiments, in conjunction with the first aspect, the method further includes: during real-time monitoring, calculating the rate of change of the second type of parameter within a preset time window; when the rate of change exceeds a preset rate threshold, determining that the target dry gas sealing system is in a condition switching process; determining the source condition type and adjacent condition type of the condition switching process based on the current direction of change of the second type of parameter; the source condition type is the condition type to which the second type of parameter belonged before the switch, and the adjacent condition type is the condition type adjacent to the source condition type in the direction of change; merging the allowable fluctuation range corresponding to the source condition type and the allowable fluctuation range corresponding to the adjacent condition type to obtain a transitional allowable fluctuation range; during the condition switching process, replacing the allowable fluctuation range corresponding to the source condition type with the transitional allowable fluctuation range as the judgment criterion for triggering the fault diagnosis process.

[0011] By adopting the above technical solution, the condition monitoring system calculates the rate of change of the second type of parameter in real-time monitoring, and identifies that the unit is in the process of switching operating conditions when the rate exceeds the limit. Then, based on the direction of parameter change, it determines the source operating condition type and the adjacent operating condition type, and merges the allowable fluctuation ranges of the two into a transitional fluctuation allowable range to replace the original baseline. This solution fully considers the transitional characteristics of parameter linkage when operating conditions change drastically. By dynamically relaxing the diagnostic trigger threshold through merging the range, it effectively avoids false alarms caused by reasonable exceedances of the first type of parameter during normal unit operations such as load or speed adjustments.

[0012] In conjunction with some embodiments of the first aspect, in some embodiments, after replacing the allowable fluctuation range corresponding to the source operating condition type with the allowable fluctuation range of the transition fluctuation range, the method further includes: determining that the operating condition switching process ends when the rate of change is less than or equal to a preset rate threshold and continues to exceed a preset stable duration; in response to the end of the operating condition switching process, determining the operating condition type to which the second type of parameter actually belongs at the end of the operating condition switching process as the current operating condition type; and switching the allowable fluctuation range of the transition fluctuation range to the allowable fluctuation range corresponding to the current operating condition type.

[0013] By adopting the above technical solution, after the condition monitoring system determines the actual condition type to which the second type of parameter belongs after the condition switch is completed, it uses this as the current condition type and smoothly switches the allowable fluctuation range for the transition to the allowable fluctuation range corresponding to the current condition type. This solution enables the monitoring benchmark to adaptively and accurately follow the transition of the equipment's steady-state operating mode, eliminating the benchmark lag problem after the transition period ends and preventing misjudgments caused by actual back-cutting or cross-level switching. This improves the consistency and reliability of the fault early warning logic throughout the entire condition switch cycle.

[0014] In conjunction with some embodiments of the first aspect, in some embodiments, there are multiple preset fault tree models, each corresponding one-to-one with an operational phase of the target dry gas sealing system. The operational phase includes at least a startup phase, a stable operation phase, and a shutdown phase. The first type of parameters, the second type of parameters, and the third type of parameters are input into the preset fault tree models for logical reasoning. Specifically, this includes: determining the operational phase of the target dry gas sealing system when the fault diagnosis process is triggered based on the second type of parameters; matching the fault tree model corresponding to the operational phase from the multiple fault tree models as the target fault tree model; and inputting the first type of parameters, the second type of parameters, and the third type of parameters into the target fault tree model for logical reasoning.

[0015] By adopting the above technical solution, the condition monitoring system determines the current operating stage based on the second type of parameters, and then matches the target fault tree model from multiple pre-set fault tree models corresponding to the corresponding stages. This model is then used to input various parameters for logical reasoning. This solution takes into account the significant differences in the physical state and failure mechanism of dry gas seals during start-up, shutdown, and stable operation. By introducing a time-dimensional stage matching mechanism, it eliminates interference from failure modes outside the current stage, improving the specificity of the logical reasoning process and the accuracy of candidate failure mode selection.

[0016] In conjunction with some embodiments of the first aspect, in some embodiments, the operator library pre-stores multiple operators, each used to characterize a specific correlation between multiple time-series data; a mapping relationship is pre-established between the failure modes in the fault tree model and the operators in the operator library, the mapping relationship being used to characterize the expected correlation that should be satisfied between the time-series data of the first type of parameter and the time-series data of the second type of parameter and / or the third type of parameter when the failure mode is met; according to the candidate failure mode, the target operator is called from the preset operator library, and the correlation between the time-series data of the first type of parameter and the time-series data of the second type of parameter and / or the third type of parameter is verified by the target operator to obtain the verification result, specifically including: according to the mapping relationship, calling the target operator corresponding to the candidate failure mode from the operator library; calculating the actual correlation between the time-series data of the first type of parameter and the time-series data of the second type of parameter and / or the third type of parameter through the target operator; comparing the actual correlation with the corresponding expected correlation to obtain the verification result.

[0017] By adopting the above technical solution, the condition monitoring system calls the target operator corresponding to the candidate failure mode according to the mapping relationship, calculates the actual correlation between the first type of parameters and the second and third types of parameters, and compares it with the expected correlation under the failure mode. This solution goes beyond the traditional judgment mode that only relies on a single parameter exceeding the limit. It extends the static logic of the fault tree to a dynamic mechanism verification that includes multi-dimensional time-series features such as trends and correlations. This allows candidate failure modes to be cross-verified by a multi-path time-series coupling mechanism, reducing the false judgment rate caused by occasional data fluctuations.

[0018] In conjunction with some embodiments of the first aspect, in some embodiments, before calculating the actual correlation between the timing data of the first type of parameters and the timing data of the second type of parameters and / or the third type of parameters through the target operator, the method further includes: determining the driving timing data and the response timing data in the timing data involved by the target operator; the driving timing data is the timing data of the second type of parameters and / or the timing data of the third type of parameters, and the response timing data is the timing data of the first type of parameters; in the driving timing data, identifying the moment when the numerical change exceeds a preset change threshold as the driving change moment; within a preset search window after the driving change moment, identifying the moment when the numerical change in the response timing data exceeds the corresponding preset change threshold as the response change moment; determining the time difference between the response change moment and the driving change moment as the propagation delay time between the driving timing data and the response timing data; and performing a time shift on the response timing data according to the propagation delay time to obtain aligned response timing data.

[0019] By adopting the above technical solution, before calculating the actual correlation, the condition monitoring system first identifies the driving change moment and the response change moment that exceed the limit from the driving time series data and the response time series data, respectively, calculates the propagation delay time, and then performs time shifting and alignment on the response time series data accordingly. This solution accurately compensates for the physical delay caused by mechanical conduction or process medium flow, enabling the causal time series data to achieve synchronization at the real physical mechanism level on the time axis, thereby improving the objectivity of subsequent operator correlation calculations and the accuracy of verification results.

[0020] In a second aspect, this application provides a state monitoring system, comprising: one or more processors and a memory; the memory is coupled to one or more processors, the memory is used to store computer program code, the computer program code including computer instructions, and the one or more processors call the computer instructions to cause the state monitoring system to perform the method described in the first aspect and any possible implementation thereof.

[0021] Thirdly, this application provides a computer-readable storage medium storing computer instructions that, when executed on a state monitoring system, cause the state monitoring system to perform the method described in the first aspect and any possible implementation thereof.

[0022] Fourthly, this application provides a computer program product, including a computer program or instructions that, when run on a state monitoring system, cause the state monitoring system to perform the method described in the first aspect and any possible implementation thereof.

[0023] Understandably, the status monitoring system provided in the second aspect, the computer-readable storage medium provided in the third aspect, and the computer program product provided in the fourth aspect are all used to execute the methods provided in the embodiments of this application. Therefore, the beneficial effects they can achieve can be referred to the beneficial effects in the corresponding methods, and will not be repeated here. Attached Figure Description

[0024] Figure 1 This is a schematic diagram of a scenario for a fault early warning method for a dry gas sealing system according to an embodiment of this application;

[0025] Figure 2 This is a flowchart illustrating the fluctuation allowable range triggering diagnostic process in a fault early warning method for a dry gas sealing system according to an embodiment of this application.

[0026] Figure 3 This is a flowchart illustrating the dynamic adjustment of the working condition switching transition range in a fault early warning method for a dry gas sealing system according to an embodiment of this application.

[0027] Figure 4 This is a flowchart illustrating the operation phase matching and timing alignment verification in a fault early warning method for a dry gas sealing system according to an embodiment of this application.

[0028] Figure 5 This is a schematic diagram of the physical device structure of a status monitoring system in an embodiment of this application.

[0029] in, Figure 5 The annotations in the attached figures are explained as follows:

[0030] 501. CPU (Central Processing Unit); 502. ROM (Read-Only Memory); 503. RAM (Random Access Memory); 504. Bus; 505. I / O Interface; 506. Input Section; 507. Output Section; 508. Storage Section; 509. Communication Section; 510. Driver; 511. Removable Media. Detailed Implementation

[0031] The terminology used in the following embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification of this application, the singular expressions “a,” “an,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this application refers to any or all possible combinations including one or more of the listed items.

[0032] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0033] This application provides a fault early warning method, system, product, and medium for a dry gas sealing system. The following describes the method in conjunction with... Figure 1 This section introduces application scenarios for embodiments of this application. Please refer to [link / reference]. Figure 1 This is a schematic diagram of a fault early warning method for a dry gas sealing system in an embodiment of this application.

[0034] In related technologies, the status monitoring and early warning of dry gas sealing systems can be achieved by setting fixed absolute alarm thresholds for key monitoring parameters (such as sealing differential pressure and leakage gas flow rate) in the control system. Specifically, during equipment operation, the control system compares the collected process parameters with the absolute alarm threshold in real time, and outputs an alarm signal when the parameter value reaches the threshold. However, in this scenario, the parameter variations caused by early faults are extremely weak and usually cannot trigger the absolute threshold. If the threshold range is forcibly narrowed to improve sensitivity, reasonable parameter fluctuations generated when the unit performs normal operating condition adjustments will easily exceed the boundary, causing the control system to misjudge normal fluctuations as equipment faults.

[0035] The fault early warning method for the dry gas sealing system in this application involves acquiring and classifying multiple monitoring parameters of the target dry gas sealing system, and determining a more sensitive fluctuation allowable range within the safe operating range based on historical normal operation data. When the first type of parameter exceeds this range, a fault diagnosis process is triggered, multiple types of parameters are input into the fault tree model to match candidate failure modes, and the target operator is called to perform mechanism verification on the correlation relationship of multiple time series data to adjust the confidence level, thereby achieving early fault warning for the dry gas sealing system. This method can not only sensitively capture early and weak fault characteristics, but also effectively eliminate interference caused by fluctuations in normal operating conditions through cross-validation of multi-dimensional parameter time series relationships.

[0036] It is evident that the fault early warning method for the dry gas sealing system in this application embodiment can not only achieve accurate early warning of minor faults in the dry gas sealing system, but also effectively solve the contradiction between the increase in alarm sensitivity and the increase in false alarm rate in traditional monitoring and early warning technologies, thereby achieving the engineering goal of preventing leakage accidents and ensuring the long-term safe and stable operation of large units.

[0037] The following describes the process of the method provided in this implementation. Please refer to [link / reference]. Figure 2 This is a flowchart illustrating a fault early warning method for a dry gas sealing system in an embodiment of this application.

[0038] S201. Obtain multi-channel monitoring parameters of the target dry gas sealing system and classify the multi-channel monitoring parameters into three categories: Category 1 parameters, Category 2 parameters, and Category 3 parameters. Category 1 parameters characterize the operating status of the target dry gas sealing system itself; Category 2 parameters characterize the operating status of the unit to which the target dry gas sealing system belongs; and Category 3 parameters characterize the status of the process medium transported by the target dry gas sealing system.

[0039] The target dry gas seal system refers to the dry gas seal device that currently requires status monitoring and fault early warning. It is typically a series dry gas seal installed on the shaft end of large rotating machinery such as turbine compressors, including primary and secondary seals. Multi-channel monitoring parameters refer to multiple time-series data collected by the DCS (Distributed Control System) or local instruments that reflect the operating status of the dry gas seal and its associated unit, such as leakage gas pressure, leakage gas flow rate, unit speed, shaft displacement, unit inlet pressure, and process gas temperature. The first type of parameter represents macroscopic parameters that directly reflect the microscopic operating status of the dry gas seal end face, such as primary seal leakage (including leakage gas pressure, leakage gas flow rate, or pressure difference between the sealing gas and the leakage gas) and secondary seal leakage. The second type of parameter represents the operating condition parameters of the associated unit, such as unit speed, unit shaft displacement, and unit vibration. The third type of parameter represents the state parameters of the process medium transported or contacted by the unit, such as unit inlet pressure, unit outlet pressure, process gas temperature, process gas composition, and flare back pressure.

[0040] This step is executed after the condition monitoring system starts its monitoring task and establishes a communication link with the data source. Specifically, the condition monitoring system first acquires multiple time-series data corresponding to the target dry gas seal system periodically from the DCS system or field instruments through communication interfaces such as OPC UA (Open Platform Communications Unified Architecture) and Modbus TCP (Modbus Transmission Control Protocol). The system then preprocesses the raw data, including timestamp alignment, missing value interpolation, outlier removal, and dimension normalization. Subsequently, the condition monitoring system maps each monitoring parameter to its corresponding parameter category according to a pre-configured parameter classification mapping table (which pre-marks each parameter based on its measurement point label, physical meaning, and correlation with the physical mechanism of the sealing end face). The first category of parameters serves as the primary judgment parameter for assessing changes in the seal's own state, directly reflecting the operation of the seal end face gap and gas film. The second category of parameters serves as the mechanical-side driving parameters directly coupled to the seal. The third category of parameters serves as the driving parameters that indirectly affect the seal through the process environment. The status monitoring system stores the categorized parameters into corresponding time-series data buffers according to their categories.

[0041] Optionally, in some embodiments, the condition monitoring system can also load different parameter classification mapping templates according to the type of the target dry gas sealing system (e.g., single-end face seal, series seal, series with intermediate labyrinth seal, double-end face seal, etc.) and the seal design pressure level (low pressure, medium pressure, high pressure) to adapt to the differentiated composition of monitoring parameters under different sealing structures.

[0042] S202. Based on the historical normal operation data of the dry gas sealing system, determine the allowable fluctuation range within the preset safe operation range. The width of the allowable fluctuation range is smaller than the width of the safe operation range.

[0043] Historical normal operation data refers to the monitoring parameter data collected by the target dry gas sealing system during past operation, either manually confirmed or marked as fault-free by the system. The safe operating range refers to the parameter value range determined by the hardware interlock alarm thresholds of the first type of parameter, such as low-low alarm, low alarm, high alarm, and high-high alarm. This range is usually preset by the engineering designer or user according to equipment safety specifications and is the absolute threshold range used by traditional control systems to trigger shutdown interlocks. For example, the safe operating range for leaked gas flow can be set to [low alarm value, high alarm value]. The allowable fluctuation range refers to the value range within the safe operating range, obtained based on statistical analysis of historical normal operation data, where the first type of parameter should not exceed the normal fluctuation range. The width of this range is strictly smaller than the width of the safe operating range, serving as a sensitive triggering benchmark for capturing subtle early fault characteristics.

[0044] For reference, the width of the allowable fluctuation range can be set to approximately 20% of the width of the safe operating range, i.e. α is a proportionality coefficient, typically ranging from 0.1 to 0.3. It can be dynamically adjusted through training with historical data. The value of α is determined based on specific working conditions and sensor accuracy.

[0045] This step is performed after the condition monitoring system completes parameter acquisition and classification, but before entering the real-time monitoring phase. Specifically, the condition monitoring system first reads a historical data sequence of the first type of parameters marked as being in normal operating condition from the historical database over a sufficiently long period (e.g., the past 3 to 12 months). It then performs statistical analysis on this historical data sequence to obtain its mean, standard deviation, quantiles, and other statistical characteristics. Next, using the center value of the safe operating range or the mean of the historical data as the center, the condition monitoring system determines a narrower allowable fluctuation range within the safe operating range according to a preset range width ratio coefficient α (e.g., using the mean as the center, taking half the set width above and below it, or using...). The upper and lower boundaries are determined by a method where k is a configurable coefficient, typically 2 to 3.

[0046] It should be noted that the fluctuation allowable range determined in this step may have different implementation forms in different embodiments: in some simplified embodiments, only a unified fluctuation allowable range may be set; in other embodiments, as described in steps S302 to S305, a corresponding fluctuation allowable range may be determined for each operating condition type; during the operating condition switching process, the transitional fluctuation allowable range described in step S309 may also be used. The 'fluctuation allowable range' mentioned in the subsequent step S203 refers to the currently effective fluctuation allowable range, which the condition monitoring system can adaptively select according to the actual operating conditions.

[0047] S203. When the current value of the first type of parameter monitored in real time exceeds the allowable fluctuation range, the fault diagnosis process is triggered.

[0048] Real-time monitoring refers to the process by which the condition monitoring system continuously acquires the first type of parameters of the target dry gas sealing system at the current moment according to a preset sampling period. The current value refers to the real-time value of the first type of parameter collected by the condition monitoring system at the current sampling moment during real-time monitoring. The fault diagnosis process refers to a series of diagnostic actions initiated by the condition monitoring system to locate the potential causes of faults in the target dry gas sealing system. This includes the fault tree reasoning, operator verification, and confidence adjustment processes involved in subsequent steps S204 to S206.

[0049] This step is continuously executed in a loop after the condition monitoring system determines the allowable fluctuation range. Specifically, the condition monitoring system acquires the current value of the first type of parameter in real time during each sampling period and compares the current value with the upper and lower boundaries of the pre-determined allowable fluctuation range: if the current value is within the allowable fluctuation range, the condition monitoring system determines that the target dry gas sealing system is currently within the normal fluctuation range, does not trigger the fault diagnosis process, and only continues to maintain routine real-time monitoring; if the current value exceeds the allowable fluctuation range (i.e., greater than the upper boundary or less than the lower boundary), but is still within the safe operating range (i.e., the absolute hardware alarm at the DCS level has not yet been triggered), the condition monitoring system determines that the target dry gas sealing system may have early abnormal potential, and then triggers the fault diagnosis process, calling subsequent steps to conduct in-depth analysis of potential fault causes; if the current value has exceeded the safe operating range, the condition monitoring system can simultaneously output high-level alarm information and start the fault diagnosis process to accurately locate the fault cause.

[0050] Optionally, in some embodiments, the status monitoring system may also adopt a continuous determination strategy, that is, the fault diagnosis process is triggered only when the current value of the first type of parameter exceeds the fluctuation allowable range in N consecutive sampling points (e.g., N=3), or when the average value in a preset sliding window exceeds the fluctuation allowable range, so as to further reduce the interference of sensor instantaneous outliers on the diagnosis process.

[0051] S204. In response to the fault diagnosis process being triggered, the first type of parameters, the second type of parameters and the third type of parameters are input into the preset fault tree model for logical reasoning, and at least one candidate failure mode is matched from the multiple failure modes contained in the fault tree model.

[0052] The fault tree model refers to a tree-like structure model pre-constructed by the condition monitoring system based on the FTA (Fault Tree Analysis) method to characterize the logical relationship between dry gas seal failure events and their causes. The top event is dry gas seal failure, intermediate events represent various failure mechanisms, and the bottom event represents abnormal changes in monitoring parameters. Events at each level are connected via logic gates such as AND gates, OR gates, and voting gates. A failure mode refers to a pre-defined event node in the fault tree model that characterizes a specific type or cause of dry gas seal failure, such as increased air intake in the secondary seal, wear on the primary seal face, stuck floating seal ring, cross-flow gas in the unit, or backpressure intrusion of process media. A candidate failure mode refers to one or more possible failure modes selected by the condition monitoring system from multiple failure modes through logical reasoning in the fault tree model that match the characteristics of the current input parameters.

[0053] This step is executed immediately after the condition monitoring system determines that the fault diagnosis process has been triggered. Specifically, the condition monitoring system first loads the fault tree model corresponding to the target dry gas sealing system type (such as a series dry gas seal) from a pre-set fault tree model library. Each failure mode in this fault tree model is pre-associated with corresponding characteristic conditions, including the direction of change of the first type of parameter (such as pressure increase or decrease, flow rate increase or decrease), the change characteristics of the second type of parameter, and the change characteristics of the third type of parameter. Subsequently, the condition monitoring system inputs the first type of parameter, the second type of parameter, and the third type of parameter at the current moment into the corresponding bottom event node of the fault tree model, and compares the parameter values ​​of each bottom event node with the pre-set characteristic conditions to determine whether the bottom event is valid. The condition monitoring system performs logical reasoning on the fault tree model in a bottom-up manner, sequentially passing the determination results of the bottom events upward according to the rules of logic gates such as AND gates, OR gates, and voting gates, until it determines which intermediate events and failure modes are activated. Finally, the condition monitoring system selects all activated failure modes from the multiple failure modes contained in the fault tree model as at least one candidate failure mode.

[0054] Optionally, in some embodiments, the condition monitoring system can also pre-set differentiated fault tree models for different working pressure levels (such as low pressure <1MPa, medium pressure 1~4MPa, high pressure ≥4MPa), different sealing sizes, and different sealing design concepts. The condition monitoring system adaptively selects the corresponding fault tree model and loads it into the inference engine according to the actual working conditions of the target dry gas sealing system, so as to improve the pertinence and accuracy of fault diagnosis.

[0055] The fault tree model construction process is as follows: The model building module of the condition monitoring system or relevant technical experts first identify the top event of the fault tree, that is, "the target dry gas sealing system fails or has serious abnormal hidden dangers" as the top logical starting point. Around this top event, based on the Failure Mode and Effects Analysis (FMEA) method, the top event is decomposed layer by layer into multiple intermediate events representing different dimensions of failure, such as abnormal sealing gas supply, mechanical damage to the sealing system, process fluid disturbance of the unit, and thermoelastic-hydrodynamic instability of the sealing surface. For each intermediate event, the specific physical mechanism and failure mode that caused the event are further traced down, such as wear of the main sealing end face, jamming of the sealing dynamic ring, aging and damage of the O-ring seal, gas leakage in the interstage labyrinth seal, or back pressure of the process medium, and these specific failure causes are defined as candidate failure mode nodes in the fault tree model.

[0056] After establishing the failure mode nodes, the core of the construction process lies in establishing the underlying mapping relationship between these theoretical failure modes and actual multi-channel monitoring parameters, that is, defining the bottom events of the fault tree. The condition monitoring system directly associates the bottom events with the specific abnormal characteristics of the first type of parameters (seale parameters), the second type of parameters (unit mechanical parameters), and the third type of parameters (process medium parameters). For example, "abnormal increase in the flow rate of the primary seal leakage gas," "violent fluctuation in unit speed," and "abnormal sealing gas inlet temperature" are used as the underlying discrimination conditions for triggering the "seale end face wear" failure mode. In this process, the system configures clear judgment rules for each bottom event, including the direction of parameter change (such as sudden increase or sudden decrease), the rate of change, and the magnitude of exceeding the safe range. In order to accurately express the causal and concurrent relationships between events at each level, various Boolean logic gates, including OR gates, AND gates, and voting gates, are flexibly configured between the top event, intermediate event, failure mode node, and bottom event during the pre-configuration process. For example, when multiple specific parameter anomalies must occur simultaneously to confirm a certain failure mode, an AND gate is used for strong constraint connection; when any one of multiple independent causes can lead to an anomaly in the higher-level system, an OR gate is used for connection.

[0057] Finally, all the tree-like topology, node characteristic attributes, and logical gate operation rules constructed based on engineering mechanisms are compiled and transformed into structured data models (such as directed acyclic graph data structures or rule configuration files in XML / JSON format) that can be directly parsed by the state monitoring system's inference engine. These digital models are persistently pre-stored in the state monitoring system's local or cloud-based rule base, so that when the fault diagnosis process is triggered, they can be quickly loaded and perform efficient automated logical reasoning based on the three types of parameters input in real time, ensuring accurate matching of candidate failure modes that conform to the current physical representation.

[0058] S205. Based on the candidate failure mode, call the target operator from the preset operator library, and use the target operator to verify the correlation between the time series data of the first type of parameter and the time series data of the second type of parameter and / or the time series data of the third type of parameter, and obtain the verification result.

[0059] The operator library refers to a pre-configured set of functions containing multiple standardized operators in the condition monitoring system. Each operator performs specific transformations, calculations, or judgments on the input time-series data, such as linear trend judgment operators, correlation judgment operators, and fluctuation amplitude judgment operators. The output of each operator is a binary result of 0 or 1. The target operator refers to one or more specific operators called by the condition monitoring system from the operator library based on the characteristic requirements of the current candidate failure mode, used to perform correlation verification. Time-series data refers to a sequence of parameter sample values ​​arranged in chronological order, including time-series data of the first type of parameters, the second type of parameters, and the third type of parameters. Correlation refers to the physical mechanism relationship between the time-series data of the first type of parameters and the time-series data of the second and / or third type of parameters in terms of change time points, change trends, and change amplitudes, such as synchronous change, linear correlation, and delayed response. The verification result refers to the judgment result obtained by the condition monitoring system through the target operator, used to evaluate whether the actual correlation relationship conforms to the expected physical mechanism of the candidate failure mode.

[0060] This step is executed after the condition monitoring system acquires at least one candidate failure mode. Specifically, the condition monitoring system first queries a preset mapping table for one or more target operators corresponding to the candidate failure mode. This mapping table records the expected correlation between the first type of parameters and the second type of parameters and / or the third type of parameters, as well as the corresponding operator type, when each failure mode is valid.

[0061] Subsequently, the condition monitoring system calls the determined target operator from the operator library and uses the timing data of the first type of parameters, as well as the timing data of the second type of parameters and / or the third type of parameters related to the candidate failure mode, as input to the target operator.

[0062] The status monitoring system calculates the correlation between the aforementioned time series data using target operators. For example, it calculates the Pearson correlation coefficient, Spearman rank correlation coefficient, or mutual information between two time series data using a correlation judgment operator, calculates the slope change characteristics (such as rising, falling, stable, or abrupt changes) of the time series data using a linear trend judgment operator, or calculates the range or variance of the time series data within a preset time window using a fluctuation amplitude judgment operator.

[0063] Finally, the state monitoring system compares the calculated indicators with preset thresholds to obtain a binary result of 0 or 1 from the target operator output, which serves as the verification result and is used to characterize whether the actual correlation conforms to the expected physical mechanism.

[0064] S206. Based on the verification results, adjust the confidence level of the candidate failure modes, and output the fault warning information corresponding to the target dry gas sealing system based on the adjusted confidence level.

[0065] Here, confidence level refers to the numerical index assigned by the condition monitoring system to each candidate failure mode, used to quantify the probability that the candidate failure mode corresponds to the actual cause of failure in the target dry gas sealing system. The value typically ranges from 0 to 1 or 0 to 100. Fault warning information refers to the structured message generated by the condition monitoring system based on the adjusted confidence level, used to inform the user of the possible causes, severity, and handling suggestions for failures in the target dry gas sealing system. Examples include warning level, failure mode name, fault description, and recommended handling measures.

[0066] This step is performed after the condition monitoring system obtains the verification results. Specifically, the condition monitoring system first assigns an initial confidence level to each candidate failure mode. The initial confidence level can be determined based on factors such as the activation depth of the fault tree inference path or the historical matching frequency.

[0067] The status monitoring system then adjusts the confidence level of each candidate failure mode based on the verification results.

[0068] If the verification result indicates that the actual correlation matches the expected correlation (i.e., the verification result is 1), the status monitoring system increases the confidence of the candidate failure mode according to the preset reward strategy.

[0069] If the verification result indicates that the actual correlation does not conform to the expected correlation (i.e., the verification result is 0), the status monitoring system reduces the confidence of the candidate failure mode according to the preset penalty strategy.

[0070] The condition monitoring system then normalizes and sorts the adjusted confidence scores of all candidate failure modes and compares them with preset warning thresholds. Candidate failure modes with confidence scores higher than the warning thresholds are identified as the final cause of failure. The condition monitoring system generates a fault warning message based on the failure description, fault level, and preset handling suggestions associated with the candidate failure mode and outputs it to the user interface. If the confidence scores of all candidate failure modes are lower than the warning thresholds, the condition monitoring system determines that the failure is due to an unknown cause and outputs a corresponding prompt, while also indicating the possibility of abnormal sensor data.

[0071] Specifically, the condition monitoring system initializes the initial confidence level for each candidate failure mode based on the proportion of the number of activated base events in the fault tree model to the total number of base events associated with that failure mode. The initial confidence level can be calculated using the following formula: ,in This represents the actual number of bottom events that were activated. The initial confidence level is the total number of underlying events associated with the failure mode. For example, if a failure mode is associated with 5 underlying events, and 3 of them are activated, then the initial confidence level is... .

[0072] Optionally, in some embodiments, the state monitoring system may also update the confidence of candidate failure modes according to the following adjustment strategy based on the verification results: when the verification result is consistent (i.e., the output is 1), the reward strategy is executed. ,in This indicates the adjusted confidence level. Indicates the confidence level before adjustment. The reference value is 0.2; when the verification result is inconsistent (i.e., the output is 0), the penalty policy is applied. ,in The reference value is 0.3. When multiple target operators verify the same candidate failure mode, the state monitoring system accumulates and adjusts the verification results of each verification. The candidate failure mode with a confidence level higher than the preset confidence level threshold (reference value is 0.7) is determined as the final failure mode and an early warning is output.

[0073] In this embodiment, since the allowable fluctuation range is determined within the safe operating range based on historical normal operating data, and when the first type of parameter exceeds the range, multiple types of parameters are input into the fault tree model to match candidate failure modes, and the correlation between the time series data of each parameter is verified by the target operator to adjust the confidence level, the time series mechanism of multi-dimensional parameters can be cross-validated without relying on absolute thresholds. This effectively solves the problem that early weak faults are difficult to detect and that fluctuations in normal operating conditions can easily lead to misjudgments, thereby improving the accuracy of early fault warning for the dry gas sealing system.

[0074] The above embodiments mainly introduce a scheme to determine the allowable fluctuation range based on historical normal operation data and trigger fault diagnosis when the first type of parameter exceeds the range. In practical applications, the units to which the dry gas sealing system belongs often operate under various different operating conditions and frequently need to perform operating condition switching operations such as load or speed adjustments. If only a single allowable fluctuation range is used as the judgment benchmark, the normal linkage change of parameters during drastic operating condition switching may lead to false triggering of fault warnings.

[0075] The following provides a more detailed description of the process of the method provided in this implementation. Please refer to [link / reference]. Figure 3 This is another flowchart illustrating a fault early warning method for a dry gas sealing system in an embodiment of this application.

[0076] S301. Obtain the multi-channel monitoring parameters of the target dry gas sealing system and classify the multi-channel monitoring parameters into three categories: first-class parameters, second-class parameters, and third-class parameters.

[0077] This step is similar to the description of step S201 in the above embodiment, and will not be repeated here.

[0078] S302. Obtain historical data of the first type of parameters collected when the target dry gas sealing system is operating normally under various historical operating conditions.

[0079] Historical operating conditions refer to the different operating states or modes experienced by the target dry gas sealing system during its past operation. The classification of operating conditions is usually based on the value ranges of the second type of parameters (such as unit speed, unit load, etc.) and the third type of parameters (such as unit inlet pressure, process gas temperature, etc.), for example, low load operating conditions, rated load operating conditions, and high load operating conditions. Normal operation refers to the stable operating state of the target dry gas sealing system under these historical operating conditions, without any faults, alarms triggered, or abnormal fluctuations. Historical data refers to the time-series data of the first type of parameters collected and stored by the condition monitoring system from the historical database during the normal operation of the target dry gas sealing system under the aforementioned historical operating conditions, such as the time-series data of primary seal leakage gas flow recorded over several past operating cycles.

[0080] This step is executed before the condition monitoring system initially configures the allowable fluctuation range for the target dry gas seal system, or when the condition monitoring system receives a historical data update instruction to periodically update the monitoring benchmark. Specifically, the condition monitoring system first accesses the internal or external historical database and retrieves the historical operation records corresponding to the target dry gas seal system from the database based on the unique identifier of the target dry gas seal system; then, the condition monitoring system filters out data segments belonging to the normal operation period based on the status markers in the operation log (such as tags for no alarm, no downtime events, no maintenance intervention, etc.); finally, the condition monitoring system extracts the time-series data of the first type of parameter from the data of the above-mentioned normal operation period as historical data and loads the historical data into memory.

[0081] S303. Group the historical data according to the working condition type to obtain the historical data set corresponding to each working condition type.

[0082] Among them, "operating condition type" refers to the category label obtained by classifying the operating status of the target dry gas sealing system based on the value characteristics of the second type of parameters and / or the third type of parameters. For example, low-speed operating condition, medium-speed operating condition, and high-speed operating condition are classified according to the unit speed, or low-pressure operating condition, medium-pressure operating condition, and high-pressure operating condition are classified according to the unit inlet pressure. "Historical data set" refers to the subset of historical data of the first type of parameters corresponding to each operating condition type, obtained by grouping historical data according to operating condition type by the condition monitoring system.

[0083] This step is performed after the condition monitoring system completes the historical data extraction. Specifically, the condition monitoring system first determines the criteria and boundaries for classifying operating conditions. The criteria are usually second-type parameters and / or third-type parameters. The boundaries can be pre-set by expert knowledge or obtained by automatically clustering the second-type and third-type parameters in the historical data using clustering algorithms (such as K-Means, DBSCAN, or Gaussian mixture models). Subsequently, the condition monitoring system classifies each sampling point in the historical data into the corresponding operating condition type based on the actual values ​​of the second-type and / or third-type parameters at that sampling time. The condition monitoring system then summarizes the classification results according to the operating condition type to obtain a historical data set corresponding to each operating condition type, and stores each historical data set separately using the operating condition type as an index.

[0084] After the condition monitoring system completes the clustering of operating condition types, it further sorts all operating condition types in ascending order according to the mean of the second type of parameter (such as unit speed) in each operating condition type, forming an ordered sequence of operating condition types along the axis of the second type of parameter value. ,in The operating condition type corresponding to the second type of parameter with the smallest mean, This corresponds to the operating condition type with the largest average parameter value in the second category. For example, if clustering yields three operating condition types with average unit speeds of 3000 rpm, 6000 rpm, and 9000 rpm respectively, the condition monitoring system will categorize them as {low speed operating condition}. Medium speed operating conditions High-speed operating conditions Adjacent operating condition types refer to two immediately adjacent operating condition types in the sequence (e.g., ...). and Adjacent, and (Adjacent). The condition monitoring system stores the obtained ordered sequence of operating conditions and the boundary values ​​of the second type of parameter corresponding to each operating condition into a preset operating condition classification table.

[0085] When operating condition types are jointly classified from multiple dimensions, the condition monitoring system can establish multiple ordered value axes according to different dimensions of the second type of parameters, and select the corresponding value axis to query adjacent operating conditions based on the dimensions of the second type of parameters that actually change during the operating condition switching process.

[0086] Optionally, in some embodiments, the condition monitoring system can also classify operating conditions based on a multi-dimensional joint approach, that is, classify operating conditions based on the combined characteristics of a second type of parameter (such as unit speed) and a third type of parameter (such as unit inlet pressure), for example, forming composite operating condition types such as "low speed-low pressure" and "high speed-high pressure" to more precisely characterize the operating state of the target dry gas sealing system.

[0087] S304. For each historical data set, based on the distribution range of the first type of parameter in the historical data set, determine the high alarm value and low alarm value of the first type of parameter under the operating condition type.

[0088] The distribution range refers to the statistically significant numerical distribution of the first type of parameter in the historical data set. It is typically characterized by statistical features such as mean, standard deviation, quantiles, and upper and lower bounds. For example, the data for the first type of parameter under a certain operating condition might exhibit an approximately normal distribution with a mean of μ and a standard deviation of σ. The high alarm value refers to the upper boundary of the normal fluctuation range of the first type of parameter, determined by the condition monitoring system for a specific operating condition. When the first type of parameter exceeds this value, the condition monitoring system will determine that it has deviated from the normal fluctuation range under that operating condition. The low alarm value refers to the lower boundary of the normal fluctuation range of the first type of parameter, determined by the condition monitoring system for a specific operating condition. When the first type of parameter falls below this value, the condition monitoring system will determine that it has deviated from the normal fluctuation range under that operating condition.

[0089] This step is performed separately for each historical data set after the condition monitoring system completes the grouping of the historical data sets. Specifically, the condition monitoring system first performs statistical analysis on the first type of parameters in the current historical data set, calculating its mean, standard deviation, maximum value, minimum value, and quantiles, and constructs a distribution range model for the historical data set based on the statistical characteristics. Subsequently, based on this distribution range model, the condition monitoring system calculates high alarm values ​​and low alarm values ​​according to preset boundary determination rules. The boundary determination rules can be, for example, by adding or subtracting K times the standard deviation from the mean (K is usually 2 or 3) to determine the upper and lower boundaries, or by using the quantile method to take the upper quantile (such as the 95th or 99th quantile) of the historical data as the high alarm value and the lower quantile (such as the 5th or 1st quantile) as the low alarm value. The condition monitoring system associates the determined high alarm values ​​and low alarm values ​​with the corresponding operating condition types and stores them in the monitoring baseline parameter table.

[0090] Optionally, in some embodiments, the status monitoring system may also perform outlier removal processing on the historical data set before calculating the high alarm value and low alarm value (such as using the 3σ criterion or box plot method) to avoid individual abnormal sampling points causing deviations in the statistical distribution range, thereby improving the accuracy of the high alarm value and low alarm value.

[0091] S305. The interval between the high alarm value and the low alarm value shall be taken as the allowable fluctuation interval corresponding to the operating condition type.

[0092] In this embodiment, the allowable fluctuation range refers to the numerical range defined for each operating condition type, which is enclosed by the high alarm value and the low alarm value under that operating condition type. It is used as the judgment criterion for determining whether the first type of parameter exceeds the normal fluctuation range under that operating condition.

[0093] This step is executed after the condition monitoring system determines the high and low alarm values ​​corresponding to each operating condition type. Specifically, the condition monitoring system uses the low alarm value for each operating condition type as the lower boundary of the allowable fluctuation range for that operating condition, and the high alarm value as the upper boundary of the allowable fluctuation range for that operating condition, thereby constructing a one-to-one allowable fluctuation range corresponding to that operating condition type. The condition monitoring system stores the allowable fluctuation ranges corresponding to each operating condition type in the monitoring benchmark parameter table in the form of key-value pairs, where the key is the identifier of the operating condition type, and the value is the upper and lower boundary values ​​of the allowable fluctuation range corresponding to that operating condition.

[0094] S306. During real-time monitoring, calculate the rate of change of the second type of parameter within a preset time window.

[0095] The real-time monitoring process refers to the continuous process by which the condition monitoring system continuously acquires multiple monitoring parameters of the target dry gas sealing system according to a preset sampling cycle and performs online analysis and judgment. The preset time window refers to the sliding time interval set by the condition monitoring system for calculating the rate of change of the second type of parameter. The window length can be set according to the change characteristics of the second type of parameter, with reference values ​​such as 1 minute, 5 minutes, or 10 minutes. The rate of change refers to how quickly the second type of parameter changes with time, as calculated by the condition monitoring system within the preset time window. It is usually expressed as the change in parameter per unit time, such as the rate of change of unit speed (rpm / min) or the rate of change of unit inlet pressure (MPa / min).

[0096] This step is continuously executed in a loop after the state monitoring system enters the real-time monitoring phase. Specifically, the state monitoring system first maintains a sliding buffer queue in memory with a length equal to a preset time window to store the time-series data of the most recently acquired second-type parameters. Then, in each sampling period, the state monitoring system adds the latest acquired second-type parameter value to the tail of the sliding buffer queue, while removing the oldest data point from the head of the queue to keep the queue length constant. Next, the state monitoring system calculates the rate of change of the second-type parameters based on the data in the sliding buffer queue. The specific calculation method can be the first-to-last difference method (i.e., the difference between the last value and the first value of the window divided by the window duration) or the linear regression method (i.e., performing least squares fitting on the data within the window to obtain the slope).

[0097] S307. When the rate of change exceeds the preset rate threshold, it is determined that the target dry gas sealing system is in the process of switching operating conditions.

[0098] The preset rate threshold refers to a critical value pre-set by the condition monitoring system to determine whether the rate of change of the second type of parameter is a rapid change. This threshold can be set based on the dynamic characteristics of the unit to which the target dry gas seal system belongs and the statistical analysis of historical operating condition switching data. For example, a reference value of 100 rpm / min can be used for the unit speed. The operating condition switching process refers to the dynamic process by which the unit to which the target dry gas seal system belongs transitions from one operating condition type to another due to reasons such as production load adjustment, start-up and shutdown operations, and process parameter adjustment.

[0099] This step is executed after the condition monitoring system calculates the rate of change of the second type of parameter. Specifically, the condition monitoring system compares the rate of change obtained in step S306 with a preset rate threshold: if the rate of change is less than or equal to the preset rate threshold, the condition monitoring system determines that the target dry gas sealing system is currently in a steady-state operation and maintains the current allowable fluctuation range as the judgment benchmark; if the rate of change exceeds the preset rate threshold, the condition monitoring system determines that the target dry gas sealing system is in the process of switching operating conditions and immediately triggers the subsequent source operating condition and adjacent operating condition identification process, while setting the operating condition switching flag bit inside the system.

[0100] S308. Based on the current change direction of the second type of parameter, determine the source operating condition type and adjacent operating condition type of the operating condition switching process.

[0101] Here, the current direction of change refers to the direction of change determined by the condition monitoring system based on the changing trend of the second type of parameter time series data within a preset time window, including both upward and downward directions. The source operating condition type refers to the operating condition type to which the second type of parameter belonged before the operating condition switch occurred, i.e., the initial operating condition of the operating condition switch. The adjacent operating condition type refers to the operating condition type adjacent to the source operating condition type along the current direction of change on the operating condition type value axis, i.e., the target operating condition of the operating condition switch.

[0102] This step is executed immediately when the condition monitoring system determines that the target dry gas seal system is in the process of switching operating conditions. Specifically, the condition monitoring system first determines the current direction of change based on the first and last values ​​of the second type of parameter or the sign of the slope of the fitted straight line within the preset time window: a positive slope indicates an upward direction, and a negative slope indicates a downward direction.

[0103] Subsequently, the condition monitoring system queries the preset condition type classification table to determine the condition type to which the second type of parameter belongs based on the steady-state value of the parameter before the condition switching trigger time, and uses it as the source condition type.

[0104] Subsequently, the condition monitoring system searches for the next working condition type that is immediately adjacent to the source working condition type on the value axis of the working condition type according to the current direction of change. Specifically, if the current direction of change is upward, the adjacent working condition type is the working condition above the source working condition type on the value axis; if the current direction of change is downward, the adjacent working condition type is the working condition below the source working condition type on the value axis.

[0105] In some embodiments, if the source operating condition type is already located at the end of the ordered sequence of operating condition types (i.e., the lowest or highest operating condition type), and the direction of change points outward from the sequence, the status monitoring system determines that there is no adjacent operating condition type. In this case, the fluctuation allowable interval corresponding to the source operating condition type is directly expanded outward in one side along the direction of change by a preset proportional coefficient (e.g., 1.2 to 1.5 times) as the transition fluctuation allowable interval, so as to adapt to the operating condition switching scenario under the end operating condition.

[0106] S309. Merge the allowable fluctuation range corresponding to the source operating condition type with the allowable fluctuation range corresponding to the adjacent operating condition type to obtain the allowable transitional fluctuation range.

[0107] Among them, the transient fluctuation allowable range refers to the temporary fluctuation allowable range constructed by the condition monitoring system for the working condition switching process. It is formed by merging the fluctuation allowable range corresponding to the source working condition type and the fluctuation allowable range corresponding to the adjacent working condition type. It is used as a relaxed benchmark for judging the first type of parameter anomaly during the working condition switching process.

[0108] This step is executed after the condition monitoring system determines the source operating condition type and the adjacent operating condition type. Specifically, the condition monitoring system first retrieves the allowable fluctuation range corresponding to the source operating condition type and the allowable fluctuation range corresponding to the adjacent operating condition type from the monitoring reference parameter table. Then, the condition monitoring system performs a merging operation on these two allowable fluctuation ranges. The merging rule is to take the larger value of the upper boundaries of the two allowable fluctuation ranges as the upper boundary of the transitional allowable fluctuation range, and take the smaller value of the lower boundaries of the two allowable fluctuation ranges as the lower boundary of the transitional allowable fluctuation range, thereby forming a union range that can simultaneously cover the normal fluctuation range of the source operating condition and the adjacent operating condition. The condition monitoring system temporarily stores the obtained transitional allowable fluctuation range in the current monitoring context.

[0109] S310. During the switching of operating conditions, the allowable fluctuation range of the transition fluctuation is replaced with the allowable fluctuation range corresponding to the source operating condition type as the judgment benchmark for triggering the fault diagnosis process.

[0110] This step is executed after the condition monitoring system obtains the allowable range for transitional fluctuations and continues to apply throughout the entire operating condition switching process. Specifically, the condition monitoring system switches the currently effective allowable fluctuation range from the allowable range corresponding to the source operating condition type to the allowable range for transitional fluctuations, and uses the allowable range for transitional fluctuations as the benchmark for subsequent comparison calculations between the current value of the first type of parameter and the judgment benchmark. During the operating condition switching process, the condition monitoring system continuously judges the current value of the first type of parameter according to the allowable range for transitional fluctuations. The condition monitoring system triggers the fault diagnosis process only when the current value of the first type of parameter exceeds the upper and lower boundaries of the allowable range for transitional fluctuations, thereby effectively avoiding misjudging normal linkage fluctuations of the first type of parameter during the operating condition switching process as fault anomalies.

[0111] S311. When the rate of change is less than or equal to the preset rate threshold and continues to exceed the preset stable duration, the working condition switching process is determined to be over.

[0112] The preset stabilization time refers to the minimum duration required for the condition monitoring system to determine that the second type of parameter change has returned to steady state. This duration can be set according to the dynamic response characteristics of the unit, with reference values ​​such as 30 seconds, 1 minute, or 5 minutes. The end of the operating condition switching process refers to the moment when the condition monitoring system determines that the unit to which the target dry gas sealing system belongs has completed the transition from the source operating condition to the target operating condition and has re-entered steady-state operation.

[0113] This step is continuously executed in a loop during the real-time monitoring process following step S310. Specifically, the condition monitoring system continuously calculates the rate of change of the second type of parameter within a preset time window according to the method in step S306, and compares the rate of change with a preset rate threshold in each sampling period; when the rate of change falls back to less than or equal to the preset rate threshold, the condition monitoring system starts or updates a duration timer and continues to monitor the rate of change in subsequent sampling periods; if the rate of change exceeds the preset rate threshold again before the cumulative duration of the duration timer reaches a preset stable duration, the condition monitoring system clears the duration timer and waits for the rate of change to fall back; if the rate of change remains less than or equal to the preset rate threshold and the cumulative duration of the duration timer exceeds the preset stable duration, the condition monitoring system determines that the working condition switching process has ended and clears the internal working condition switching flag.

[0114] S312. In response to the end of the operating condition switching process, determine the operating condition type to which the second type of parameter actually belongs at the end of the operating condition switching process as the current operating condition type.

[0115] The current operating condition type refers to the operating condition type to which the actual steady-state value of the second type of parameter belongs at the end of the operating condition switching process.

[0116] This step is executed immediately after the condition monitoring system determines that the operating condition switching process has ended. Specifically, the condition monitoring system first obtains the actual value of the second type of parameter at the end of the operating condition switching process. To ensure the stability of the value, the condition monitoring system can perform averaging or median processing on the second type of parameter within a preset time window before and after the end of the operating condition switching to obtain a representative steady-state value. Subsequently, the condition monitoring system queries the operating condition type to which the steady-state value belongs based on a preset operating condition type classification table and determines it as the current operating condition type.

[0117] It is worth noting that the current operating condition type is consistent with the adjacent operating condition type determined in step S308 in most cases; however, when complex situations such as reverse switching, mid-term termination, or cross-level switching occur during the operating condition switching process, the current operating condition type may be different from the adjacent operating condition type. The condition monitoring system determines the current operating condition type based on the actual value in this step.

[0118] S313. Switch the allowable range for transitional fluctuations to the allowable range for fluctuations corresponding to the current operating condition type.

[0119] The switching of the allowable fluctuation range refers to the process by which the condition monitoring system replaces the currently effective judgment benchmark from the transitional allowable fluctuation range with the allowable fluctuation range corresponding to a certain operating condition type.

[0120] This step is executed immediately after the condition monitoring system determines the current operating condition type. Specifically, the condition monitoring system first retrieves the allowable fluctuation range corresponding to the current operating condition type from the monitoring benchmark parameter table; then, the condition monitoring system replaces the currently effective judgment benchmark from the transitional allowable fluctuation range with the allowable fluctuation range corresponding to the current operating condition type, and uses this allowable fluctuation range as the judgment benchmark for determining whether the current value of the first type of parameter is abnormal during subsequent real-time monitoring; after the condition monitoring system completes the switch, it clears the temporary storage of the transitional allowable fluctuation range and records the timing information of this operating condition switch and the allowable fluctuation range switch in the monitoring log, including the start and end times of the switch, the source operating condition type, the current operating condition type, and the allowable fluctuation range used before and after the switch.

[0121] Optionally, in some embodiments, the condition monitoring system may also introduce a smooth transition mechanism when switching from the transitional fluctuation allowable range to the fluctuation allowable range corresponding to the current operating condition type. That is, at the moment of switching, it does not directly switch to the fluctuation allowable range corresponding to the current operating condition type, but gradually and linearly transitions the upper and lower boundaries of the judgment benchmark from the boundary of the transitional fluctuation allowable range to the boundary of the fluctuation allowable range corresponding to the current operating condition type within a preset smoothing time period, so as to avoid the risk of false alarm caused by the sudden narrowing of the judgment benchmark at the moment of switching.

[0122] S314. When the current value of the first type of parameter monitored in real time exceeds the allowable fluctuation range, the fault diagnosis process is triggered.

[0123] This step is similar to step S203 in the above embodiments, and will not be repeated here.

[0124] It should be noted that in this scheme, "operation phase" and "operating condition type" are two orthogonal judgment dimensions: the operation phase is used to characterize the macroscopic working period of the unit (start-up, steady state, shutdown), and its judgment is based on whether the unit speed is in the start-up ramp, steady state range, or shutdown ramp. This judgment is used to select the corresponding target fault tree model; while the operating condition type is used to characterize the subdivided operating state within the stable operation phase, and its judgment is based on the specific value range of the second type of parameter within the steady state range. This judgment is used to determine the corresponding allowable fluctuation range.

[0125] Specifically, before determining the operating condition switch, the condition monitoring system first performs an operation phase determination: if the target dry gas sealing system is determined to be in the start-up or shutdown phase, the condition monitoring system will suspend the use of the fluctuation allowable interval switching logic based on the ordered operating condition type sequence, and instead use a dynamic fluctuation allowable interval dedicated to the start-up or shutdown phase (this interval is obtained separately from the historical data of the start-up / shutdown phase, and usually has a looser boundary to accommodate the larger fluctuations of the first type of parameter during the start-up and shutdown process); only when the condition monitoring system determines that the target dry gas sealing system is in the stable operation phase will the operating condition switch detection and transition fluctuation allowable interval switching logic described in steps S306~S313 be activated.

[0126] In this embodiment, since the rate of change of the second type of parameter is calculated in real-time monitoring to determine the operating condition switching process, and the fluctuation allowable intervals corresponding to the source operating condition type and the adjacent operating condition type are merged into the transition fluctuation allowable interval as a temporary judgment benchmark, and then switched to the fluctuation allowable interval corresponding to the current operating condition type after the end, the trigger threshold during the operating condition adjustment period can be dynamically and adaptively relaxed. This effectively solves the problem of false early warning caused by normal linkage fluctuations of parameters when the unit frequently adjusts the production load, and thus realizes continuous and reliable monitoring of the status monitoring system under complex dynamic operating conditions.

[0127] The above embodiments mainly introduce how to effectively avoid false alarms during the normal operating condition adjustment phase by identifying the operating condition switching process and dynamically merging the allowable transition fluctuation range. In practical applications, after the fault diagnosis process is accurately triggered, due to the significant differences in the seal failure mechanism of the unit in different operating phases (such as the start-up and shutdown phases and the stable operation phase), and the objective physical delay in the transmission of changes in drive parameters to response parameters, if the original time series data is directly used for unified association rule verification, the verification results may fail and diagnostic errors may occur due to mismatch in phase mechanisms or time axis misalignment.

[0128] The following provides a more detailed description of the process of the method provided in this implementation. Please refer to [link / reference]. Figure 4 This is another flowchart illustrating a fault early warning method for a dry gas sealing system in an embodiment of this application.

[0129] S401. In response to the fault diagnosis process being triggered, the first type of parameters, the second type of parameters and the third type of parameters are input into the preset fault tree model for logical reasoning, and at least one candidate failure mode is matched from the multiple failure modes contained in the fault tree model.

[0130] This step specifically includes:

[0131] The second type of parameters determines the operating stage of the target dry gas sealing system when the fault diagnosis process is triggered.

[0132] The target fault tree model is selected from multiple fault tree models and matched with the fault tree model corresponding to the operational phase.

[0133] The first type of parameters, the second type of parameters, and the third type of parameters are input into the target fault tree model for logical reasoning.

[0134] The operation phase refers to the different working periods of the unit belonging to the target dry gas sealing system in a complete operating cycle, including at least the startup phase, the stable operation phase, and the shutdown phase. The startup phase refers to the period when the unit starts from the shutdown state and the speed gradually increases from zero to the speed at which the sealing end face opens or the rated speed. During this phase, the sealing end face is in a contact or semi-contact state. The stable operation phase refers to the period when the unit speed reaches above the end face opening speed and is in normal process operation. During this phase, the sealing end face is fully open and a stable gas film is formed. The shutdown phase refers to the period when the unit speed gradually decreases from the stable operation state to zero. During this phase, the sealing end face transitions from a non-contact state to a contact state.

[0135] There are multiple pre-set fault tree models, each corresponding to one of the operating stages of the target dry gas sealing system. That is, an independent fault tree model is constructed and pre-set for each operating stage. The composition of the top event, intermediate event, bottom event and failure mode of each fault tree model is customized according to the fault mechanism characteristics of the corresponding operating stage.

[0136] During the construction of the fault tree model, the condition monitoring system does not use a single static model. Instead, it independently constructs and pre-sets multiple fault tree models for different physical characteristics of the unit during startup, steady state, and shutdown. For example, in the fault tree model corresponding to the startup phase, considering that the sealing end face is in a state of dry friction or mixed friction at low unit speeds, and the leakage fluctuation is relatively large, the model will adaptively relax the trigger threshold of relevant low-level events and shield failure mode nodes such as gas film oscillation that only occur at high speeds. In the fault tree model corresponding to the stable operation phase, the full-range refined monitoring logic will be activated, introducing more stringent parameter deviation thresholds and complex multi-parameter cross-validation logic gates.

[0137] This step is executed immediately after the condition monitoring system determines that the fault diagnosis process has been triggered. Specifically, the condition monitoring system first extracts the second type of parameters at the current moment, focusing on parameters that characterize the operating phase, such as unit speed, and compares these second type of parameters with the preset operating phase determination rules:

[0138] If the unit speed is zero or close to zero, it is determined that the unit is currently in a shutdown state.

[0139] If the unit speed is between zero and the end face opening speed and is on the rise, it is determined that it is currently in the startup phase;

[0140] If the unit speed is higher than the end face opening speed and within a stable fluctuation range, it is determined that the unit is currently in a stable operation phase.

[0141] If the unit speed is between the end-face opening speed and zero and is decreasing, it is determined that the unit is currently in a shutdown phase.

[0142] Subsequently, the state monitoring system retrieves the corresponding fault tree model from multiple pre-set fault tree models based on the determined operating phase as the target fault tree model. Then, the state monitoring system inputs the first, second, and third type parameters of the current moment into the corresponding bottom event nodes of the target fault tree model, and performs logical reasoning on the target fault tree model in a bottom-up order according to the pre-set AND gate, OR gate, voting gate, and other logic gate rules in the target fault tree model, passing the judgment results of the bottom events upward layer by layer until it is determined which failure modes in the target fault tree model are activated. Finally, the state monitoring system selects all activated failure modes from the multiple failure modes contained in the target fault tree model as at least one candidate failure mode and outputs them to the subsequent steps.

[0143] Optionally, in some embodiments, the condition monitoring system can also pre-set multiple fault tree models for different pressure levels (such as low pressure, medium pressure, and high pressure), different sealing sizes, or different sealing design concepts under each operating stage, forming a multi-dimensional fault tree model library of operating stage × pressure level × sealing specification. The condition monitoring system adaptively selects the most matching target fault tree model according to the current comprehensive operating conditions to further improve the accuracy and relevance of candidate failure mode matching.

[0144] It should be noted that during the startup and shutdown phases, the physical characteristics of the first type of parameter (such as leakage gas flow rate) differ fundamentally from those during the stable operation phase because the sealing end face is in a contact or semi-contact state. In some embodiments, the condition monitoring system constructs independent allowable fluctuation ranges for the startup and shutdown phases, and the determination method is similar to steps S302 to S305, but the historical data sets used are respectively taken from the normal operation data of the historical startup process and the normal operation data of the historical shutdown process; the condition monitoring system automatically switches to the corresponding allowable fluctuation range according to the current operation phase as the judgment criterion for triggering the fault diagnosis process.

[0145] In some embodiments, the operator library contains multiple pre-stored operators, each of which is used to characterize a specific correlation between multiple time-series data.

[0146] In the fault tree model, a mapping relationship is pre-established between the failure modes and the operators in the operator library. The mapping relationship is used to characterize the expected correlation that should be satisfied between the time series data of the first type of parameter and the time series data of the second type of parameter and / or the third type of parameter when the failure mode is established. For example, the expected correlation corresponding to the failure mode "increased secondary seal air intake" is that the time series data of the primary leakage gas and the time series data of the secondary intake gas are completely consistent in terms of time points and trends. This expected correlation is characterized by the correlation judgment operator and verified by it.

[0147] S402. Based on the mapping relationship, call the target operator corresponding to the candidate failure mode from the operator library.

[0148] This step is executed after the condition monitoring system obtains at least one candidate failure mode. Specifically, the condition monitoring system first retrieves a pre-established mapping table from its internal storage. This mapping table uses the failure mode as the index key and stores the target operator identifier corresponding to each failure mode, the list of participating time-series data required for verification (including the first type of parameters and the corresponding second and / or third type of parameters), the expected correlation type (such as positive correlation, negative correlation, synchronous change, delayed response, linear increase, large-scale fluctuation, etc.), and related operator parameters (such as correlation coefficient threshold, trend judgment window length, fluctuation amplitude threshold, etc.).

[0149] Subsequently, the status monitoring system queries the corresponding target operator identifier in the mapping table based on the current candidate failure mode, and loads the corresponding target operator instance from the operator library according to the identifier. If there are multiple candidate failure modes, the status monitoring system executes the above mapping query and target operator call process for each candidate failure mode, and binds the called target operator with the corresponding candidate failure mode.

[0150] The following table shows a specific implementation of the mapping table used in the condition monitoring system:

[0151] Failure Mode Target operator Driving timing data Response timing data Expected related relationships Secondary seal increases air intake Correlation judgment operator Secondary intake volume time series data Time series data of primary leakage gas flow rate Positive correlation, correlation coefficient ≥ 0.8 Wear of primary sealing end face Linear trend judgment operator Unit speed timing data Time series data of primary leakage gas flow rate The flow rate of the first-stage leak gas shows a slow upward trend, and the fitted slope is greater than the preset slope threshold (e.g., 0.01 Nm³ / h·min). Insufficient sealing air pressure Correlation judgment operator Sealing gas inlet pressure timing data Level 1 Leakage Gas Pressure Time Series Data Positive correlation, correlation coefficient ≥ 0.85 Air film oscillation at the sealed end face Fluctuation Amplitude Judgment Operator Unit speed timing data Time series data of primary leakage gas flow rate When the rotational speed is stable, the fluctuation range of the primary leakage gas flow rate is greater than the preset amplitude threshold (e.g., 15% of the rated value). Process medium back pressure Linear trend judgment operator Unit inlet pressure time series data Time series data of primary leakage gas flow rate Negative correlation: when the inlet pressure increases, the flow rate of the primary leakage gas decreases. O-ring failure Fluctuation Amplitude Judgment Operator Sealing gas inlet temperature timing data Time series data of primary leakage gas flow rate The flow rate of the primary leakage gas fluctuates synchronously with temperature fluctuations, and the synchronicity of the fluctuations is ≥0.75.

[0152] S403. Determine the driving timing data and response timing data in the timing data involved in the target operator.

[0153] Driving time-series data refers to time-series data that serves as the cause of a causal relationship in the expected correlation corresponding to the failure mode. Changes in the corresponding physical quantities typically occur before the response time-series data. In this scheme, driving time-series data refers to time-series data of second-type parameters and / or third-type parameters, such as unit speed and unit inlet pressure. Response time-series data refers to time-series data that serves as the result of a causal relationship in the expected correlation corresponding to the failure mode. Changes in the corresponding physical quantities are typically triggered by changes in driving time-series data and occur subsequently. In this scheme, response time-series data refers to time-series data of first-type parameters, such as primary seal leakage gas flow rate and primary seal leakage gas pressure.

[0154] Specifically, this step is executed after the state monitoring system completes the target operator invocation. Specifically, the state monitoring system first reads the list of time-series data involved in the target operator and the role label of each time-series data in the expected correlation from the mapping table, based on the candidate failure modes bound to the current target operator. Then, the state monitoring system divides the time-series data involved in the verification into driving time-series data and response time-series data according to the role labels: time-series data labeled as second-type parameters and / or third-type parameters indicating causal causes are determined as driving time-series data, and time-series data labeled as first-type parameters indicating causal results are determined as response time-series data. The state monitoring system extracts the determined driving time-series data and response time-series data from the data cache and organizes them according to a unified sampling timestamp to form the input data pairs for the target operator.

[0155] S404. In the driving timing data, identify the moment when the numerical change exceeds the preset change threshold and use it as the driving change moment.

[0156] The preset change threshold for the drive timing data refers to the critical value pre-set by the condition monitoring system for determining whether the drive timing data value has changed significantly. This threshold can be set based on the dimensional characteristics, historical fluctuation statistics, and engineering experience of the drive timing data. For example, a reference value of 50 rpm can be used for unit speed, and a reference value of 0.05 MPa can be used for unit inlet pressure. It can also be set in the form of relative change (e.g., a reference value of 5% of the historical average of the drive timing data). The drive change moment refers to the moment when the value change in the drive timing data exceeds the preset change threshold corresponding to the drive timing data, that is, the starting moment when the drive physical quantity undergoes a significant change.

[0157] This step is executed after the condition monitoring system determines the drive timing data. Specifically, the condition monitoring system first calculates the numerical change of the drive timing data point by point according to a preset sliding window. The numerical change can be calculated by using the first difference between adjacent sampling points (i.e., the value of the later time step minus the value of the previous time step), the difference between the last value and the first value within the sliding window, or the standard deviation of the data within the sliding window. Subsequently, the condition monitoring system compares the calculated numerical change with a preset change threshold point by point, and filters out all sampling moments where the numerical change exceeds the preset change threshold. The condition monitoring system sorts the filtered moments in chronological order and, based on the fault diagnosis process trigger time, takes the most recent moment that exceeds the preset change threshold as the drive change moment.

[0158] S405. Within a preset search window after the driving change moment, identify the moment when the numerical change in the response time series data exceeds the corresponding preset change threshold, and use it as the response change moment.

[0159] The preset search window refers to the time interval set by the state monitoring system after the moment of the driving change, used to search for the moment of the response change. The window length can be set according to the physical transmission mechanism and historical statistical characteristics between the driving time series data and the response time series data, with reference values ​​such as 30 seconds, 1 minute, or 5 minutes. The moment of response change refers to the moment when the change in value in the response time series data exceeds the corresponding preset change threshold, which is identified within the preset search window; that is, the starting moment when the response physical quantity responds to the change in the driving physical quantity.

[0160] This step is executed after the condition monitoring system determines the moment of drive change. Specifically, the condition monitoring system first extracts data segments within the time interval of the response timing data, starting from the moment of drive change and ending at the end of a preset search window after the moment of drive change. Then, the condition monitoring system calculates the numerical change of each extracted response timing data segment point by point in a manner similar to step S404, and compares the calculated numerical change with the preset change threshold corresponding to the response timing data (this threshold is set independently of the preset change threshold corresponding to the drive timing data, and their values ​​are determined separately; for example, when the drive timing data is the unit speed, the corresponding preset change threshold reference value is 50 rpm, and when the response timing data is the first-level leakage gas flow rate, the corresponding preset change threshold reference value is 10% of the rated flow rate). The condition monitoring system selects the moment when the numerical change in the response timing data first exceeds the corresponding preset change threshold as the response change moment and outputs it to subsequent steps. If no moment when the numerical change in the response timing data exceeds the corresponding preset change threshold is found within the preset search window, the condition monitoring system determines that there is no valid response in the response timing data at the current moment of drive change.

[0161] If the state monitoring system does not identify the response change time within the preset search window, it determines that no valid response has appeared in the response timing data at the current driving change time. In this case, the state monitoring system no longer performs the conduction delay time calculation and timing shift operation in steps S406 to S407, but directly sets the verification result of this verification to 0 (i.e., determines that the actual correlation does not meet the expected correlation) and outputs the result to the subsequent confidence adjustment step; or, the state monitoring system uses the pre-calibrated default conduction delay time of the candidate failure mode to replace the measured delay time and continues to perform the subsequent correlation calculation.

[0162] S406. The time difference between the response change time and the drive change time is determined as the propagation delay time between the drive timing data and the response timing data.

[0163] The transmission delay time refers to the time lag caused by the change of the driving physical quantity in the driving time series data being transmitted to the response physical quantity in the response time series data through the actual physical transmission process. Its value is equal to the time difference between the response change time and the driving change time.

[0164] This step is executed after the state monitoring system determines the driving change time and the response change time, respectively. Specifically, the state monitoring system calculates the time difference between the timestamp of the response change time and the timestamp of the driving change time, and uses this time difference as the propagation delay time between the driving timing data and the response timing data; the state monitoring system stores this propagation delay time in the current verification context.

[0165] S407. Shift the response timing data according to the conduction delay time to obtain aligned response timing data.

[0166] Time shift refers to the process by which the condition monitoring system offsets the response timing data along the time axis by the propagation delay time. Aligned response timing data refers to the new timing data sequence obtained by aligning the response timing data with the causal triggering time of the driving timing data on the time axis through time shifting.

[0167] This step is performed after the state monitoring system obtains the conduction delay time. Specifically, the state monitoring system shifts the timestamp of each sampling point in the response timing data forward by the amount of conduction delay time, so that the response timing data, which originally lagged behind the driving timing data, is causally aligned with the driving timing data on the time axis. To ensure that the shifted response timing data is consistent with the driving timing data in terms of sampling timestamps, the state monitoring system can use interpolation methods such as linear interpolation or cubic spline interpolation to resample the shifted response timing data at the sampling timestamps of the driving timing data, resulting in aligned response timing data where the sampling timestamps are completely aligned with the driving timing data.

[0168] S408. Calculate the actual correlation between the aligned response timing data and the driving timing data (i.e., the timing data of the second type of parameters and / or the timing data of the third type of parameters) using the target operator.

[0169] The actual correlation refers to the quantitative indicators that the state monitoring system calculates based on the aligned response time series data and driving time series data through the target operator, reflecting the actual correlation characteristics between the two in the current monitoring period, including but not limited to correlation coefficient, trend consistency index, fluctuation synchronization index, etc.

[0170] Specifically, this step is executed after the state monitoring system obtains the aligned response time-series data. Specifically, the state monitoring system uses the aligned response time-series data and the driving time-series data as inputs to the target operator, and performs corresponding calculations based on the type of the target operator: If the target operator is a correlation judgment operator, the state monitoring system first normalizes the two time-series data (e.g., Z-score normalization or Min-Max normalization), and then calculates the Pearson correlation coefficient, Spearman rank correlation coefficient, or mutual information as the quantified value of the actual correlation; if the target operator is a linear trend judgment operator, the state monitoring system performs least-squares fitting on the two time-series data within a preset time window, calculates their respective slopes, compares the signs and magnitudes of the slopes, and outputs trend consistency characteristics; if the target operator is a fluctuation amplitude judgment operator, the state monitoring system calculates the range or variance of the two time-series data within a preset time window and compares their synchronicity, outputting fluctuation synchronicity characteristics; the state monitoring system uses the quantified indicators calculated by the target operator as the actual correlation.

[0171] S409. Perform a consistency comparison between the actual association and the corresponding expected association to obtain the verification result.

[0172] Consistency comparison refers to the process by which the condition monitoring system compares the actual correlation with the expected correlation corresponding to the candidate failure modes and determines whether the two are consistent. The verification result is a binary judgment result obtained by the condition monitoring system through consistency comparison, which reflects whether the actual correlation conforms to the expected correlation. An output of 1 indicates consistency (i.e., the actual correlation conforms to expectations), and an output of 0 indicates inconsistency (i.e., the actual correlation does not conform to expectations).

[0173] This step is executed after the condition monitoring system obtains the actual correlation. Specifically, the condition monitoring system first queries the mapping table for the expected correlation corresponding to the current candidate failure mode. The expected correlation is stored in the form of structured parameters, including the expected correlation type (such as positive correlation, negative correlation, synchronous rise, delayed response, etc.) and the corresponding judgment thresholds (such as the lower limit of the correlation coefficient, the sign of the trend slope, the fluctuation synchronicity threshold, etc.).

[0174] Subsequently, the status monitoring system compares the actual correlations calculated in step S408 with the expected correlations item by item:

[0175] If the quantified value of the actual correlation meets the threshold for determining the expected correlation (such as the actual correlation coefficient being greater than or equal to the lower limit of the expected correlation coefficient, the actual trend slope sign being consistent with the expected one, the actual fluctuation synchronicity meeting the expected threshold, etc.), the status monitoring system determines that the two are consistent and outputs a verification result of 1.

[0176] Otherwise, the status monitoring system will determine that the two are inconsistent and output a verification result of 0.

[0177] The condition monitoring system binds the verification results with the corresponding candidate failure modes for use in subsequent confidence adjustment and fault warning information output steps.

[0178] In this embodiment, the target fault tree model corresponding to the operating stage is matched according to the second type of parameters, and the driving change time and response change time are identified before operator verification to determine the conduction delay time. The response timing data is then time-shifted according to this time to obtain the aligned response timing data. Therefore, the interference of non-current operating stage mechanisms is eliminated and the physical conduction delay is accurately compensated. This effectively solves the problem of inaccurate timing verification of multiple parameters due to physical response lag and stage mechanism differences, thereby achieving accurate fault location and early warning based on real physical timing synchronization.

[0179] The status monitoring system in the embodiments of this invention is described below from the perspective of hardware processing. Please refer to [link / reference]. Figure 5 This is a schematic diagram of the physical device structure of a status monitoring system in an embodiment of this application.

[0180] It should be noted that, Figure 5 The structure of the status monitoring system shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.

[0181] like Figure 5 As shown, the status monitoring system includes a CPU 501, which can perform various appropriate actions and processes based on a program stored in the read-only memory ROM 502 or a program loaded from the storage section 508 into the random access memory RAM 503, such as performing the methods described in the above embodiments. The RAM 503 also stores various programs and data required for system operation. The CPU 501, ROM 502, and RAM 503 are interconnected via a bus 504. An I / O interface 505 is also connected to the bus 504.

[0182] The following components are connected to I / O interface 505: input section 506 including audio input devices, push-button switches, etc.; output section 507 including a liquid crystal display (LCD) and audio output devices, indicator lights, etc.; storage section 508 including a hard disk, etc.; and communication section 509 including a network interface card such as a LAN (Local Area Network) card, modem, etc. Communication section 509 performs communication processing via a network such as the Internet. Drive 510 is also connected to I / O interface 505 as needed. Removable media 511, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 510 as needed so that computer programs read from them can be installed into storage section 508 as needed.

[0183] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program / instructions carried on a computer-readable medium, the computer program / instructions containing computer program / instructions for performing the methods shown in the flowcharts. In such embodiments, the computer program / instructions can be downloaded and installed from a network via communication section 509, and / or installed from removable medium 511. When the computer program is executed by CPU 501, it performs the various functions defined in the present invention.

[0184] It should be noted that specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0185] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. Each block in a flowchart or block diagram may represent a module, program segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those shown in the drawings.

[0186] Specifically, the condition monitoring system of this embodiment includes a processor and a memory. The memory stores a computer program. When the computer program is executed by the processor, it implements a fault early warning method for a dry gas sealing system provided in the above embodiment.

[0187] In another aspect, the present invention also provides a computer-readable storage medium, which may be included in the condition monitoring system described in the above embodiments; or it may exist independently and not assembled into the condition monitoring system. The storage medium carries one or more computer programs that, when executed by a processor of the condition monitoring system, cause the condition monitoring system to implement a fault early warning method for a dry gas sealing system provided in the above embodiments.

[0188] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit it. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

[0189] As used in the above embodiments, depending on the context, the term "when..." can be interpreted as meaning "if...", "after...", "in response to determining...", or "in response to detecting...". Similarly, depending on the context, the phrase "when determining..." or "if (the stated condition or event) is interpreted as meaning "if determining...", "in response to determining...", "when (the stated condition or event) is detected", or "in response to detecting (the stated condition or event)".

[0190] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM or random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A fault early warning method for a dry gas sealing system, applied to a condition monitoring system, characterized in that, include: Acquire multi-channel monitoring parameters of the target dry gas sealing system, and classify the multi-channel monitoring parameters into a first category of parameters, a second category of parameters, and a third category of parameters; The first type of parameter is a parameter that characterizes the operating status of the target dry gas sealing system itself; the second type of parameter is a parameter that characterizes the operating status of the unit to which the target dry gas sealing system belongs; and the third type of parameter is a parameter that characterizes the status of the process medium transported by the target dry gas sealing system. Based on the historical normal operation data of the target dry gas sealing system, a fluctuation allowable range is determined within the preset safe operation range, wherein the width of the fluctuation allowable range is smaller than the width of the safe operation range. When the current value of the first type of parameter monitored in real time exceeds the allowable fluctuation range, the fault diagnosis process is triggered. In response to the fault diagnosis process being triggered, the first type of parameters, the second type of parameters and the third type of parameters are input into a preset fault tree model for logical reasoning, and at least one candidate failure mode is matched from the multiple failure modes contained in the fault tree model. According to the candidate failure mode, the target operator is called from the preset operator library, and the correlation between the time series data of the first type of parameter and the time series data of the second type of parameter and / or the time series data of the third type of parameter is verified by the target operator to obtain the verification result. Based on the verification results, the confidence level of the candidate failure modes is adjusted, and based on the adjusted confidence level, the fault warning information corresponding to the target dry gas sealing system is output.

2. The method according to claim 1, characterized in that, The determination of the allowable fluctuation range within a preset safe operating range, based on the historical normal operating data of the target dry gas sealing system, specifically includes: Historical data of the first type of parameters collected when the target dry gas sealing system is operating normally under various historical conditions; The historical data are grouped according to the working condition type to obtain a historical data set corresponding to each working condition type. For each set of historical data, based on the distribution range of the first type of parameters in the set of historical data, determine the high alarm value and low alarm value of the first type of parameters under the operating condition type; The interval between the high alarm value and the low alarm value is taken as the fluctuation allowable interval corresponding to the operating condition type.

3. The method according to claim 2, characterized in that, The method further includes: During real-time monitoring, the rate of change of the second type of parameter within a preset time window is calculated; When the rate of change exceeds a preset rate threshold, it is determined that the target dry gas sealing system is in the process of switching operating conditions. Based on the current change direction of the second type of parameter, the source operating condition type and adjacent operating condition type of the operating condition switching process are determined; the source operating condition type is the operating condition type to which the second type of parameter belongs before the switching, and the adjacent operating condition type is the operating condition type that is adjacent to the source operating condition type in the change direction. The fluctuation allowable range corresponding to the source operating condition type is merged with the fluctuation allowable range corresponding to the adjacent operating condition type to obtain the transition fluctuation allowable range; During the switching of operating conditions, the allowable fluctuation range of the transition fluctuation is replaced by the allowable fluctuation range corresponding to the source operating condition type, and used as the judgment criterion for triggering the fault diagnosis process.

4. The method according to claim 3, characterized in that, After replacing the allowable fluctuation range corresponding to the source operating condition type with the allowable transition fluctuation range, the method further includes: When the rate of change is less than or equal to the preset rate threshold and continues for more than the preset stable duration, the working condition switching process is determined to be over. In response to the end of the working condition switching process, the working condition type to which the second type of parameter actually belongs at the end of the working condition switching process is determined as the current working condition type; Switch the allowable range of transitional fluctuations to the allowable range of fluctuations corresponding to the current operating condition type.

5. The method according to claim 1, characterized in that, The preset fault tree model is multiple, and each of the multiple fault tree models corresponds one-to-one with the operation stage of the target dry gas sealing system. The operation stage includes at least the startup stage, the stable operation stage, and the shutdown stage. The step of inputting the first type of parameters, the second type of parameters, and the third type of parameters into a preset fault tree model for logical reasoning specifically includes: The operating stage of the target dry gas sealing system when the fault diagnosis process is triggered is determined based on the second type of parameters. Match the fault tree model corresponding to the operation phase from among the multiple fault tree models as the target fault tree model; The first type of parameters, the second type of parameters, and the third type of parameters are input into the target fault tree model for logical reasoning.

6. The method according to claim 1, characterized in that, The operator library contains multiple operators, each of which is used to characterize a specific correlation between multiple time-series data. The failure modes in the fault tree model are pre-established with the operators in the operator library. The mapping relationship is used to characterize the expected correlation that should be satisfied between the time series data of the first type of parameter and the time series data of the second type of parameter and / or the time series data of the third type of parameter when the failure mode is established. The step of calling a target operator from a preset operator library based on the candidate failure modes, and verifying the correlation between the time-series data of the first type of parameters and the time-series data of the second type of parameters and / or the time-series data of the third type of parameters using the target operator to obtain a verification result, specifically includes: Based on the mapping relationship, the target operator corresponding to the candidate failure mode is called from the operator library; The actual correlation between the time series data of the first type of parameters and the time series data of the second type of parameters and / or the time series data of the third type of parameters is calculated using the target operator. The actual relationship is compared with the corresponding expected relationship to obtain the verification result.

7. The method according to claim 6, characterized in that, Before calculating the actual correlation between the time-series data of the first type of parameters and the time-series data of the second type of parameters and / or the time-series data of the third type of parameters using the target operator, the method further includes: Determine the driving timing data and response timing data in the timing data involved in the target operator; the driving timing data is the timing data of the second type of parameters and / or the timing data of the third type of parameters, and the response timing data is the timing data of the first type of parameters; In the driving timing data, the moment when the numerical change exceeds a preset change threshold is identified as the driving change moment; Within a preset search window after the driving change moment, the moment when the numerical change in the response time series data exceeds the corresponding preset change threshold is identified as the response change moment. The time difference between the response change time and the drive change time is determined as the propagation delay time between the drive timing data and the response timing data; The response timing data is time-shifted according to the propagation delay time to obtain aligned response timing data.

8. A condition monitoring system, characterized in that, The status monitoring system includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code including computer instructions, and the one or more processors call the computer instructions to cause the status monitoring system to perform the method as described in any one of claims 1-7.

9. A computer-readable storage medium storing computer instructions, characterized in that, When the computer instructions are executed on the condition monitoring system, the condition monitoring system performs the method as described in any one of claims 1-7.

10. A computer program product, comprising a computer program or instructions, characterized in that, When the computer program or instructions are run on the condition monitoring system, the condition monitoring system performs the method as described in any one of claims 1-7.

Images