A method and system for integrating alarms in a device process control system

Abstract

The application discloses a kind of method and system for integrating alarm in equipment process control system, belong to equipment process control technical field, by obtaining the input and output data of multiple equipment and adjusting response information, extract conservation correlation feature sequence, identify local correlation relationship deviation under the condition that overall numerical value remains stable, form implicit breakage feature set;Further construct forward evolution process and reverse backtracking process, form bidirectional consistency feature sequence, and filter dominant abnormal path according to this;On this basis, the cumulative degree of deviation is progressively analyzed, the critical change interval is identified and the instability approaching index is generated;By comparing the instability approaching index with the safety boundary, determine the warning intensity and divide the alarm level, output alarm information and generate process regulation instruction when the trigger condition is met;The application can identify implicit abnormalities and give early warning under the condition of system apparent stable state, improve the safety and reliability of process control.

Description

Technical Field

[0001] This invention relates to the field of equipment process control technology, and more specifically to a method and system for integrating alarms in equipment process control systems. Background Technology

[0002] In complex process scenarios involving multi-level feedback regulation and cross-device coupling, system operation typically exhibits a dynamic equilibrium process characterized by apparent stability but continuous internal state reconfiguration. Existing alarm methods largely rely on explicit parameter shifts or simple correlation judgments, failing to identify early anomalous states where "the overall equilibrium remains intact, but local conservation relationships have implicitly deviated." This is particularly problematic when feedback regulation continuously offsets deviations, potentially leaving the system on the verge of instability for extended periods without triggering effective alarms. Furthermore, existing methods lack pre- and post-verification mechanisms for the anomalous evolution process, making it difficult to distinguish between genuine instability trends and short-term disturbances, resulting in both false alarms and missed alarms. Therefore, it is necessary to propose an integrated alarm method capable of characterizing implicit equilibrium breaches and possessing bidirectional verification capabilities. Summary of the Invention

[0003] The purpose of this invention is to provide a method and system for integrating alarms in equipment process control systems to address the shortcomings of the prior art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a method for integrating alarms in a equipment process control system, comprising: Acquire input and output data and regulation response information of multiple devices during operation, and extract the conservation correlation feature sequence reflecting the system equilibrium state; Based on the conserved correlation feature sequence, abnormal segments in which local correlations shift under the condition that the overall values ​​remain stable are identified, forming a set of hidden broken features; For the set of implicit broken features, a forward evolution process is constructed to obtain the changing trend of the anomaly over time, and a reverse backtracking process is constructed simultaneously to verify the continuity of the anomaly in historical stages, forming a bidirectional consistent feature sequence. Based on the bidirectional consistency feature sequence, abnormal paths that simultaneously satisfy the conditions of positive enhancement and reverse traceability are selected as the dominant abnormal paths; A progressive analysis is performed on the degree of offset accumulation in the dominant abnormal path to identify the critical change interval from a stable state to an unstable state and generate the corresponding instability approximation index. The instability approach index is compared with the safety boundary to determine the warning intensity value, and the alarm level is classified according to the warning intensity value; When the alarm level reaches the trigger condition, an alarm message is output, and a process control command matching the critical change range is generated.

[0005] Preferably, all time windows that are determined to be in a stable state or a gradual change state are spliced ​​together in chronological order, and the relationship between the input and output quantities that are consistent or change regularly within the corresponding time period is extracted. This relationship is recorded in the form of a time series, thereby forming a conserved correlation feature sequence.

[0006] Preferably, the process of identifying anomalous segments where local correlations shift while overall numerical values ​​remain stable includes: The conserved correlation feature sequence is segmented according to time order, and the total change of input and output in each time period is calculated. Stable intervals with total change less than a preset threshold are selected. Within the stable interval, the degree of change in the correlation between each measuring point is further calculated. By comparing the difference in the corresponding correlation in adjacent time periods, the correlation offset data sequence is obtained. Based on the associated offset data sequence, segments with offset amplitudes greater than a preset offset threshold are identified at multiple consecutive time points to form candidate abnormal segments; The candidate abnormal segments are subjected to consistency screening, and the abnormal segments that persist within the stable interval and have the same offset direction are retained as the set of hidden defect features.

[0007] Preferably, the step of constructing the forward evolution process includes: Arrange the anomalous segments in the set of hidden broken features in chronological order, calculate the difference in the change value of the correlation between adjacent time points, form a positive change data sequence describing the speed of anomalous development, and construct a positive evolution process based on this.

[0008] Preferably, forming a bidirectional consistent feature sequence includes: Based on the positive change data sequence, segments in continuous time points where the change amplitude continuously increases or decreases are identified, and the corresponding abnormal development trends are extracted to form a change trend sequence of abnormality over time. Starting from the end time of the abnormal segment, the changes in the correlation value are traced back point by point in reverse time. It is determined whether the changes between adjacent time points are continuously connected during the tracing process, and a reverse tracing process is constructed. The sequence of changes in the anomaly over time is matched with the continuity determination results in the reverse backtracking process to filter out the anomaly segments that simultaneously satisfy both positive change consistency and reverse continuity, thus forming the bidirectional consistency feature sequence.

[0009] Preferably, the step of matching the sequence of changes in the anomaly over time with the continuity determination result in the reverse backtracking process includes: Assign trend direction and magnitude values ​​to each time point in the sequence of anomalies over time, and form a positive feature label sequence in chronological order. Based on the continuity determination result in the reverse backtracking process, a continuity identifier is generated for the corresponding time point, and time alignment processing is performed with the forward feature marker sequence to form dual sequence aligned data. In the dual-sequence aligned data, the correspondence between the trend direction indicator and the continuity indicator is compared point by point. When the two remain consistent for no less than 4 consecutive time points of a preset length, they are determined to be a matching segment. The matching segments are subjected to integrity screening, and segments that simultaneously satisfy the consistency of trend direction and the continuity coverage ratio of more than 80% are retained as valid matching results for constructing a bidirectional consistency feature sequence.

[0010] Preferably, the step of filtering out abnormal paths that simultaneously meet the conditions of positive enhancement and reverse traceability includes: The time segments in the bidirectional consistency feature sequence are divided into paths, and multiple candidate anomaly paths are constructed according to the connection order of the measurement points corresponding to the correlation relationship. Based on the candidate abnormal paths, the length of time during which the trend direction identifier in each path is continuously in the same direction is calculated, and the degree of positive enhancement is determined by combining the cumulative result of the change amplitude value. Perform reverse backtracking verification on the candidate abnormal paths, count the number of time points in the path that meet the continuity identifier, and calculate the proportion of the time points to the total number of time points in the path to characterize the degree of reverse traceability. The positive enhancement level and the reverse traceability level are jointly compared to select candidate abnormal paths that simultaneously satisfy the condition that the positive enhancement level is greater than the preset enhancement threshold and the reverse traceability level is greater than 80%, which are then selected as the dominant abnormal paths.

[0011] Preferably, the steps of performing asymptotic analysis on the degree of offset accumulation in the dominant abnormal path and generating an instability approximation index include: The correlation change values ​​at each time point are extracted along the dominant anomaly path in chronological order, and then accumulated point by point to form a cumulative offset data sequence; Based on the cumulative offset data sequence, the changes in the cumulative increment between adjacent time points are calculated, and the segments in the continuous time points where the cumulative increment changes from stable fluctuations to continuous growth are identified as candidate critical change intervals. The growth rate of the cumulative increment within the candidate critical change interval is analyzed. When the cumulative increment at consecutive time points shows an increasing trend and the duration is not less than 3 time points, the interval is determined to be the critical change interval. The instability approximation index is calculated based on the combination relationship between the cumulative offset value and the corresponding growth rate within the critical change interval.

[0012] Preferably, the step of comparing the instability approximation index with the safety boundary and determining the warning intensity value includes: Based on the statistical results of the instability approximation index during the historical stable operation phase, its maximum value and average value are extracted to construct a safety boundary interval that is dynamically updated over time. The instability approach index at the current moment is compared with the safety boundary interval, the degree of deviation from the upper limit of the safety boundary is calculated, and the degree of deviation is mapped to the warning intensity value according to the ratio. The warning intensity values ​​are classified according to their numerical range. When the warning intensity values ​​are within a preset range, they are classified into different alarm levels. The alarm level is continuously verified. If the alarm level remains at or above the same level within a continuous time period, the alarm level is confirmed as a valid output.

[0013] The present invention also provides a system for integrating alarms in a process control system for equipment, comprising: Data acquisition module: acquires input and output data and adjustment response information of multiple devices during operation, and extracts conservation correlation feature sequences that reflect the equilibrium state of the system; Implicit Flaw Identification Module: Based on the conserved correlation feature sequence, it identifies abnormal segments where local correlations shift under the condition that the overall values ​​remain stable, forming a set of implicit flaw features; Bidirectional consistency analysis module: For the set of implicit broken features, a forward evolution process is constructed to obtain the changing trend of the anomaly over time, and a reverse backtracking process is constructed simultaneously to verify the continuity of the anomaly in historical stages, forming a bidirectional consistency feature sequence. Dominant path filtering module: Based on the bidirectional consistency feature sequence, filter out the abnormal paths that simultaneously meet the conditions of positive enhancement and reverse traceability, and use them as the dominant abnormal paths; Instability Analysis and Index Generation Module: Performs asymptotic analysis on the degree of offset accumulation in the dominant abnormal path, identifies the critical change interval from stable state to unstable state, and generates the corresponding instability approximation index; Early warning assessment module: compares the instability approach index with the safety boundary, determines the early warning intensity value, and classifies the alarm level according to the early warning intensity value; Adjustment execution module: When the alarm level reaches the trigger condition, it outputs alarm information and generates process adjustment instructions that match the critical change range.

[0014] The technical effects and advantages provided by the present invention in the above technical solution are as follows: 1. This invention introduces a joint analysis mechanism combining conserved correlation feature sequences and implicitly broken feature sets to characterize the dynamic balance changes between input and output quantities of multiple devices at the physical mechanism level. Unlike existing alarm methods that rely solely on single-point thresholds or simple trend judgments, this invention continuously tracks the input-output conservation relationship and identifies shifts in local correlations without disrupting the overall apparent balance of materials or energy, thereby capturing implicit imbalances that are difficult to detect using traditional methods. This technique directly acts on the internal physical transmission relationships of the system, making anomaly identification no longer dependent on explicit numerical out-of-bounds events, but based on the more fundamental physical change process of broken conservation relationships. This significantly improves the ability to perceive early anomalies and effectively solves the problem of the inability to provide early warnings for hidden anomalies in existing technologies.

[0015] 2. This invention reconstructs the physical evolution path of anomaly propagation by constructing a bidirectional consistency judgment mechanism for both forward evolution and reverse backtracking processes, combined with dominant anomaly path screening and progressive analysis of the instability approximation index. Unlike existing technologies that only make isolated judgments on the current state, this invention, from the perspective of temporal continuity and propagation path, ensures that the identified anomalies possess both continuous evolutionary characteristics and historical traceability, thus avoiding interference from short-term disturbances on the judgment results. Based on this, an instability approximation index is generated by combining the offset accumulation process with the growth rate, and compared with the safety boundary to quantitatively characterize the transition process from stability to instability in the system. This technique directly reflects the physical process of offset energy or material accumulation and release within the system, providing a clear physical basis for alarm triggering, thereby improving the accuracy and targeted nature of early warnings, significantly reducing false alarm rates, and enhancing the overall safety of the process control system. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0017] Figure 1 This is a flowchart of the method of the present invention.

[0018] Figure 2 This is a flowchart of the system modules of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Example 1, please refer to Figure 1 As shown in this embodiment, a method for integrating alarms in a equipment process control system includes: The system acquires input and output data and regulation response information of multiple devices during operation, and extracts conservation correlation feature sequences that reflect the system's equilibrium state.

[0021] In this embodiment, to extract the conservation correlation feature sequence reflecting the system's equilibrium state, the input / output data and adjustment response information are first processed for time synchronization. Specifically, using the system's unified clock as a reference, a fixed sampling interval of 1 second is set, and the input / output data and adjustment response information collected by each device are mapped to the corresponding sampling time according to the timestamp. When there is data missing between adjacent sampling times, the numerical difference between two adjacent valid sampling points is used to fill in the missing data according to the time ratio. For example, when the values ​​corresponding to time points 1 and 3 are 80 and 100 respectively, the data value of time point 2 is determined to be 90 according to the linear change between 80 and 100, thereby forming a continuous and missing original process data sequence under a unified time reference.

[0022] After obtaining the original process data sequence, the differences between the input and output quantities of each device are calculated point by point to form a balance offset data sequence. Specifically, at each sampling time, the input value of the corresponding device is subtracted from the output value to obtain the balance offset value at that time. The balance offset values ​​of consecutive sampling times are arranged in chronological order to reflect the trajectory of the change in the degree of system input-output imbalance over a short period of time.

[0023] Based on this, response compensation processing is performed on the balance offset data sequence by incorporating adjustment response information. Specifically, the adjustment direction and amplitude at each sampling moment are extracted from the adjustment response information, and the balance offset value is corrected according to the adjustment direction: when the adjustment direction is to increase output or decrease input, the corresponding adjustment amplitude is subtracted from the balance offset value at that moment; when the adjustment direction is to increase input or decrease output, the corresponding adjustment amplitude is added to the balance offset value. For example, if the balance offset value is 5 at a certain sampling moment, and the adjustment response information indicates an increase in output adjustment with an amplitude of 2, then the corrected balance characteristic data is 3; if the adjustment direction is opposite, then the corrected value is 7. Through the above point-by-point correction, the obtained data can eliminate the direct influence of human adjustment factors on the system balance judgment, resulting in corrected balance characteristic data that is closer to the actual operating state.

[0024] Furthermore, continuity analysis is performed on the corrected equilibrium feature data to extract stable or gradually changing correlations. Specifically, a time window length of 10 consecutive sampling points is set, and the window slides in steps of 1 sampling point. Within each time window, the maximum and minimum values ​​of the equilibrium offset are calculated, and the difference between them is taken as the fluctuation amplitude. Simultaneously, the direction of numerical change between adjacent sampling points is compared in chronological order. When the number of sampling points with the same continuous direction of change is no less than 7, it is determined to be a unidirectional trend. When the fluctuation amplitude is less than a preset threshold of 2, the equilibrium state corresponding to that time window is determined to be stable; when the fluctuation amplitude is greater than or equal to 2 and satisfies a unidirectional trend, it is determined to be a gradually changing state. Time windows that do not meet the above two conditions are not included in subsequent feature extraction.

[0025] Finally, all time windows determined to be in a stable or gradually changing state are concatenated in chronological order, and the relationships between input and output quantities that remain consistent or exhibit regular changes within the corresponding time periods are extracted. These relationships are recorded in time series form, thus forming a conserved correlation feature sequence. This conserved correlation feature sequence can be used in subsequent steps to identify hidden broken feature sets and analyze anomaly propagation paths, providing fundamental data support for achieving high-precision integrated alarms.

[0026] Based on the conserved correlation feature sequence, abnormal segments in which local correlations shift under the condition that the overall values ​​remain stable are identified, forming a set of hidden broken features.

[0027] In this embodiment, in order to identify abnormal segments where local correlations shift under the condition that the overall numerical values ​​remain stable, based on the conservation correlation feature sequence, the conservation correlation feature sequence is first segmented in chronological order, and the total change of input and output in each time period is calculated to screen out stable intervals.

[0028] Specifically, the conservation correlation feature sequence is divided into fixed-length time periods, each consisting of 20 consecutive sampling points. A sliding update method is used, with a sliding step size of 5 sampling points between adjacent time periods, resulting in a 15-point overlap between adjacent time periods to ensure the stability of the temporal continuity analysis. Within each time period, the input quantities of all devices participating in the conservation correlation are accumulated point-by-point to form the total input for that time period. Simultaneously, the corresponding output quantities are accumulated point-by-point to form the total output. The total change in the total input is obtained by subtracting the total output from the total input.

[0029] Furthermore, the change in total amount is compared with a preset threshold, which is set according to the proportion of the total input amount, with a proportion coefficient of 1%. That is, when the total input amount in a certain time period is 2000, the corresponding preset threshold is 20. When the change in total amount is less than the preset threshold, it is determined that the time period meets the condition of maintaining overall value stability, and the time period is marked as a stable interval. By performing the above filtering on all time periods, multiple stable interval sets are formed and arranged in chronological order.

[0030] After obtaining the stable interval, the degree of change in the correlation between each measuring point is further calculated within the stable interval. Specifically, based on the correspondence between the input and output quantities recorded in the conserved correlation feature sequence, each pair of correlated measuring points is compared one by one between adjacent time periods. For each measuring point, the difference between the input and output quantities is calculated in the current time period, and the corresponding difference is calculated in the previous time period. The absolute value of the difference between the two is taken as the change in the correlation between the measuring point and the two time periods.

[0031] For example, if the input at a certain measuring point was 120 and the output was 115 in the previous time period, the corresponding difference was 5. However, in the current time period, the difference is 9, resulting in a change in the correlation value of 4. To avoid misjudgments caused by single-point fluctuations, the differences between multiple sampling points for each measuring point in the current time period are averaged. This involves averaging the differences of all sampling points in the current time period and comparing it with the average difference of the previous time period to obtain a more stable correlation change value. By performing this process on all measuring points, a correlation offset data sequence arranged chronologically is formed. This correlation offset data sequence is used to characterize the degree of deviation of local relationships under the overall conservation state.

[0032] After forming the correlated offset data sequence, candidate anomalous segments are identified based on this sequence. Specifically, firstly, statistical analysis is performed on all correlation changes within the stable interval, and their average value and average deviation are calculated. The average deviation is obtained by averaging the absolute values ​​of the differences between each correlation change value and the average value.

[0033] Subsequently, the offset threshold is set to the average value plus twice the average deviation, so that the offset threshold can reflect the upper limit of the normal fluctuation range. When the change value of the correlation corresponding to a certain time point is greater than the offset threshold, it is determined that there is an abnormal offset at that time point; further, continuous time points are detected, and when the number of continuous time points is not less than 3 and all of them meet the condition that the change value of the correlation is greater than the offset threshold, the continuous segment is marked as a candidate abnormal segment.

[0034] For example, if the above conditions are met at four consecutive time points between time point 15 and time point 18, then the segment is identified as a candidate anomalous segment. Meanwhile, to avoid unclear boundaries, the starting point of a candidate anomalous segment is defined as the first time point exceeding the offset threshold, and the ending point is defined as the last consecutive time point exceeding the offset threshold, thus forming a clearly defined time interval.

[0035] After obtaining candidate anomalous segments, consistency screening is performed to form a set of hidden breach features. Specifically, the correlation offset direction within each candidate anomalous segment is analyzed. The correlation offset direction is determined by comparing the change in the correlation relationship between the current time point and the previous time point: when the change in the correlation relationship shows a continuous increasing trend, it is determined to be a positive offset; when the change in the correlation relationship shows a continuous decreasing trend, it is determined to be a negative offset; when the change directions are inconsistent, it is recorded as directional instability. Within the candidate anomalous segment, the offset directions of all time points are statistically analyzed. When the proportion of time points with the same offset direction to the total number of time points in the segment is greater than or equal to 80%, the candidate anomalous segment is determined to have directional consistency; otherwise, the candidate anomalous segment is considered to lack stable evolutionary characteristics and is eliminated.

[0036] Furthermore, a stability interval constraint check is performed on the candidate anomaly segments that pass the directional consistency screening. This involves determining whether all time points of the candidate anomaly segment fall within the aforementioned stability interval. If any time point exceeds the stability interval, it indicates that the segment is affected by overall numerical fluctuations and should be removed. Only candidate anomaly segments that simultaneously satisfy both directional consistency and stability interval constraints are retained and arranged in chronological order.

[0037] Finally, the selected candidate anomaly fragments are integrated in chronological order. For each anomaly fragment, its start time, end time, and corresponding correlation change value sequence are recorded. These anomaly fragments are then combined to form a set of hidden malfunction features. This set of hidden malfunction features can accurately reflect the process of continuous shift in local correlations while maintaining a stable overall input-output ratio, thereby revealing the potential anomaly evolution trend within the system.

[0038] For the set of implicit broken features, a forward evolution process is constructed to obtain the changing trend of the anomaly over time, and a reverse backtracking process is constructed simultaneously to verify the continuity of the anomaly in historical stages, forming a bidirectional consistent feature sequence.

[0039] In this embodiment, to construct a forward evolution process based on the set of hidden broken features and form a bidirectional consistent feature sequence by combining it with a reverse backtracking process, the anomalous segments in the set of hidden broken features are first arranged in chronological order. Specifically, each anomalous segment is sorted in ascending order according to its starting time point, and the correlation change values ​​are arranged in the sampling time order within each anomalous segment, thus forming a continuous time series structure. Based on this, the difference between the correlation change values ​​between adjacent time points is calculated. That is, the correlation change value at the current time point is subtracted from the correlation change value at the previous time point, based on the correlation change value at the previous time point, to obtain the change difference value at that time point. This change difference value is used to characterize the anomaly development speed.

[0040] After obtaining the positive change data sequence, abnormal development trends are further identified based on this sequence. Specifically, the direction of the change difference between adjacent time points in the positive change data sequence is determined. When the change difference between consecutive time points is positive and shows an increasing relationship, the segment is determined to be an accelerating upward trend; when the change difference is negative and the absolute value gradually increases, it is determined to be an accelerating downward trend; when the change difference maintains a consistent sign but the numerical change is small and there is no obvious increasing or decreasing relationship, it is determined to be a stable change trend. The trend determination length is further set. When the number of time points that consecutively meet the same trend type is not less than four, the segment is extracted as a valid trend segment, and its start time, end time, and corresponding change difference range are recorded, thus forming a sequence of abnormal trend development over time.

[0041] After establishing a sequence of anomalies and their changing trends over time, a reverse backtracking process is constructed, starting from the end time of the anomaly segment. Specifically, starting from this end time, the changes in correlation values ​​are traced back point by point in reverse chronological order, and the difference in changes between adjacent time points is calculated during the backtracking process. For each backtracking time point, the change in correlation value at the current time point is compared with the change in correlation value at the previous backtracking time point. When the absolute value of the difference between the two is less than a preset continuity threshold, the time point is determined to meet the continuity condition. The continuity threshold is determined based on the fluctuation range of correlation value changes within a historical stable interval; for example, the maximum value of this fluctuation range is taken as the continuity threshold. When the number of time points that continuously meet the continuity condition is not less than three, the backtracking segment is determined to be continuous, and the start and end positions of the segment are recorded, thus forming the continuity determination result in the reverse backtracking process.

[0042] After obtaining the anomaly's trend sequence over time and the continuity determination results from the reverse backtracking process, the two are further matched. Specifically, firstly, each time point in the anomaly's trend sequence is assigned a trend direction identifier and a change magnitude value. The trend direction identifier is determined by the sign of the change difference; a positive change difference indicates an upward trend, and a negative change difference indicates a downward trend. Simultaneously, the absolute value of the corresponding change difference is used as the change magnitude value, and a positive feature label sequence is formed in chronological order. Subsequently, based on the continuity determination results from the reverse backtracking process, a continuity identifier is generated for the corresponding time point. A value of 1 is assigned when a time point belongs to a continuous backtracking segment, and a value of 0 otherwise. These continuity identifiers are then arranged in chronological order.

[0043] Based on this, the positive feature marker sequence and the continuity marker are time-aligned. Specifically, using a unified time axis as a benchmark, the two types of sequences are matched at the same time points. If data is missing from one sequence at a certain time point, it is filled in by interpolation between adjacent time points, thus forming double-sequence aligned data. In the double-sequence aligned data, the correspondence between the trend direction marker and the continuity marker is compared point by point for each time point. When the change direction corresponding to the trend direction marker is consistent with the segment where the continuity marker is 1, the time point is determined to meet the matching condition. A continuous matching length is further set; when the number of time points that continuously meet the matching condition is not less than 4, the continuous segment is determined to be a matched segment.

[0044] After obtaining the matching segments, they are subjected to integrity screening. Specifically, for each matching segment, the number of time points within it that satisfy the trend direction consistency requirement is counted, and the proportion of this number to the total number of time points in the matching segment is calculated. When this proportion is greater than or equal to 80%, the matching segment is determined to meet the trend direction consistency requirement. At the same time, the number of time points within the matching segment with a continuity flag of 1 is counted, and their proportion is calculated. When this proportion is also greater than or equal to 80%, the matching segment is determined to meet the continuity coverage requirement. Only matching segments that simultaneously meet the above two conditions are retained and are considered valid matching results.

[0045] Finally, all valid matching results are integrated in chronological order, and trend direction indicators, change magnitude values, and continuity indicators for the corresponding time periods are extracted to form a bidirectional consistency feature sequence. This bidirectional consistency feature sequence reflects both the development trend of anomalies in the forward evolution process and the continuity characteristics in the reverse backtracking process, thus effectively distinguishing between true persistent anomalies and short-term disturbances, providing a reliable basis for subsequent screening of dominant anomaly paths and calculation of early warning intensity.

[0046] Based on the bidirectional consistency feature sequence, abnormal paths that simultaneously satisfy the conditions of positive enhancement and reverse traceability are selected as the dominant abnormal paths.

[0047] In this embodiment, to screen dominant abnormal paths based on bidirectional consistency feature sequences, the time segments within the bidirectional consistency feature sequences are first divided into paths. Specifically, the bidirectional consistency feature sequences are segmented according to temporal continuity. When the trend direction indicator or continuity indicator changes between adjacent time points, it serves as the segment boundary, thus forming multiple time segments that are temporally continuous and feature-consistent. Within each time segment, based on the correlation between measurement points recorded in the conserved correlation feature sequence, the corresponding measurement points are connected sequentially according to the transmission order between input and output quantities to form a path structure. For example, if there is an input-output transmission relationship between measurement points A, B, and C within a certain time segment, candidate abnormal paths are constructed in the order from A to B and then to C. By performing the above operations on all time segments, multiple candidate abnormal paths are obtained. Each candidate abnormal path includes a corresponding time segment, a trend direction indicator sequence, a change magnitude value sequence, and a continuity indicator sequence.

[0048] After obtaining candidate abnormal paths, the positive enhancement degree of each candidate abnormal path is further calculated. Specifically, in each candidate abnormal path, the number of time points where the trend direction indicator is consecutively in the same direction is counted, and the continuous length of the corresponding time period is calculated. For example, if the trend direction indicator of a certain path is upward for 6 consecutive time points, then the continuous length of that path in that segment is 6. At the same time, the change amplitude values ​​within that segment are accumulated and summed, that is, the change amplitude values ​​of each time point within that segment are added together to obtain the cumulative change amplitude value. Further, the continuous length and the cumulative change amplitude value are combined, and the continuous length multiplied by the cumulative change amplitude value is used as the result of calculating the positive enhancement degree. For example, when the continuous length is 6 and the cumulative change amplitude value is 15, then the positive enhancement degree is 90. To ensure the comparability between different paths, the positive enhancement degree is normalized, specifically by dividing the positive enhancement degree of each path by the maximum positive enhancement degree among all candidate abnormal paths, thus obtaining a positive enhancement degree value ranging from 0 to 1.

[0049] After calculating the positive enhancement level, a reverse backtracking check is performed on each candidate abnormal path to characterize the reverse traceability level. Specifically, in each candidate abnormal path, the number of time points with a continuity marker of 1 is counted, and the proportion of this number to the total number of time points in the path is calculated. For example, if a path contains 10 time points, and 8 of them have a continuity marker of 1, then the reverse traceability level of the path is 0.8. Furthermore, to improve the stability of the reverse traceability determination, continuous segment analysis is performed on the time points with a continuity marker of 1. When the length of a continuous segment is less than 2 time points, the segment is considered an isolated point and is removed. Only time points with a continuous segment length of at least 2 are retained for the proportion calculation, thereby avoiding interference from occasional continuity markers.

[0050] After obtaining the positive enhancement level and the reverse traceability level, the two are jointly compared to filter the dominant abnormal paths. Specifically, an enhancement threshold is first set, which is 0.6 times the average positive enhancement level of all candidate abnormal paths. For example, when the average positive enhancement level of all paths is 0.5, the enhancement threshold is 0.3. When the positive enhancement level of a candidate abnormal path is greater than this enhancement threshold, it is determined that it meets the positive enhancement condition. At the same time, the reverse traceability level is compared with a preset ratio threshold, which is set to 0.8. When the reverse traceability level is greater than or equal to 0.8, the path is determined to meet the reverse traceability condition. Only candidate abnormal paths that simultaneously meet both the positive enhancement condition and the reverse traceability condition are retained.

[0051] Furthermore, when multiple candidate anomaly paths simultaneously meet the above conditions, they are prioritized and ranked. Specifically, the product of the positive enhancement degree and the reverse traceability degree is used as the ranking criterion, and the candidate anomaly path with the largest product value is determined as the highest priority path; when multiple paths have the same product value, their time coverage lengths are further compared, and the path with the longer time coverage length is selected as the dominant anomaly path. Finally, the selected dominant anomaly paths are output in chronological order, and their corresponding measurement point connection relationships, trend direction identification sequences, change magnitude numerical sequences, and continuity identification sequences are recorded.

[0052] In this embodiment, the above processing steps enable effective screening and optimization of abnormal paths based on bidirectional consistency feature sequences. By simultaneously introducing two constraints—positive enhancement degree and reverse traceability degree—this method not only identifies paths with significant abnormal development trends but also ensures good continuity of these paths across historical periods, thus avoiding misjudging short-term fluctuations or local disturbances as dominant abnormal paths. The resulting dominant abnormal path accurately reflects the main propagation direction and evolution process of anomalies within the system, providing a reliable basis for subsequent early warning intensity calculations and process adjustment command generation.

[0053] A progressive analysis is performed on the degree of offset accumulation in the dominant abnormal path to identify the critical change interval from a stable state to an unstable state, and the corresponding instability approximation index is generated.

[0054] In this embodiment, to progressively analyze the cumulative degree of offset in the dominant anomaly path and identify the critical change intervals transitioning from a stable to an unstable state, the correlation change values ​​at each time point along the dominant anomaly path are first extracted in chronological order, and a cumulative offset data sequence is constructed. Specifically, the correlation change values ​​at each time point in the dominant anomaly path are arranged in chronological order, and the correlation change value at the first time point is used as the initial cumulative value. Subsequently, an accumulation operation is performed point by point for subsequent time points, that is, the correlation change value at the current time point is added to the cumulative result at the previous time point to form a new cumulative value.

[0055] After obtaining the cumulative offset data sequence, the cumulative increment between adjacent time points is calculated to identify the trend of change. Specifically, the cumulative offset value of the current time point is subtracted from the cumulative offset value of the previous time point to obtain the cumulative increment of that time point.

[0056] Furthermore, a stability assessment is performed on the cumulative increment. In the initial stage, a sliding interval of 5 time points is selected. Within this interval, the average value of the cumulative increment and the difference between the maximum and minimum values ​​are calculated. When this difference is less than a preset stability threshold, the interval is considered to be in a stable fluctuation state. The stability threshold is determined based on the cumulative increment fluctuation range during historical normal operation phases; for example, the average level of the difference between the maximum and minimum cumulative increments in historical phases is taken as the stability threshold. When the cumulative increment at subsequent time points continuously exceeds this stability threshold, the system is considered to have entered the offset enhancement phase.

[0057] After identifying the offset enhancement phase, the changes in cumulative increments across consecutive time points are further analyzed to determine candidate critical change intervals. Specifically, starting from the time point where the cumulative increment first exceeds the stability threshold, the changes in cumulative increments are detected point by point. When the cumulative increments across consecutive time points are all greater than the previous time point and maintain an increasing relationship, this consecutive segment is marked as a candidate critical change interval. For example, if the cumulative increment sequence in a certain phase is 4, 6, 9, 13, then this segment meets the increasing condition and is identified as a candidate critical change interval. To avoid misjudgment due to short-term fluctuations, the length of this consecutive segment must be at least three time points; if this condition is not met, it is not considered a candidate critical change interval.

[0058] After obtaining the candidate critical change interval, a growth rate analysis is performed on the cumulative increment within this interval to further confirm the critical change interval. Specifically, within the candidate critical change interval, the difference in cumulative increments between adjacent time points is calculated, that is, the cumulative increment at the current time point is subtracted from the cumulative increment at the previous time point to obtain the growth rate value.

[0059] Furthermore, the growth rate is trend-judged. When the growth rate maintains an increasing relationship within a continuous time period and the number of continuous time points is not less than 3, the candidate interval is determined to be a critical change interval. If the growth rate decreases or fluctuates significantly, the interval is considered to lack stable unstable transition characteristics and should be eliminated.

[0060] After determining the critical change interval, the instability approximation index is calculated based on the combination of the cumulative offset value and the corresponding growth rate within that interval. Specifically, first, the average cumulative offset value at all time points within the critical change interval is calculated, along with the average growth rate. Then, the average cumulative offset value is multiplied by the average growth rate to obtain the initial instability index value. For example, if the average cumulative offset value within a critical change interval is 20 and the average growth rate is 3, the initial instability index value is 60. Further, to facilitate comparison between different paths, the initial instability index value is normalized by dividing it by the maximum initial instability index value among all dominant abnormal paths, thus obtaining an instability approximation index ranging from 0 to 1.

[0061] Furthermore, to enhance the sensitivity of the instability approximation index to the length of the critical change interval, an interval length correction factor is introduced based on the normalized index. This correction factor is the ratio of the number of critical change interval time points to a preset baseline length. For example, if the baseline length is set to 5, the correction factor is 0.8 when the interval length is 4. Finally, the normalized instability approximation index is multiplied by the correction factor to obtain the final instability approximation index. This index comprehensively reflects the degree of cumulative deviation and its growth trend, thus characterizing the extent to which the dominant anomalous path transitions from a stable state to an unstable state.

[0062] Through the above processing, not only is a quantitative description of the offset accumulation process achieved, but also key critical change intervals are identified through multi-level judgment rules and quantitatively expressed in the form of an instability approximation exponent. This method can effectively distinguish between slow changes and rapid instability processes, providing a forward-looking criterion for subsequent early warning intensity calculations, thereby improving the accuracy and timeliness of integrated alarms in equipment process control systems.

[0063] The instability approach index is compared with the safety boundary to determine the warning intensity value, and the alarm level is classified according to the warning intensity value.

[0064] In this embodiment, to compare the instability approach index with the safety boundary and determine the warning intensity value, a safety boundary interval is first constructed based on the statistical results of the instability approach index during historical stable operation phases. Specifically, within the historical time period when the system is in a stable operating state, the instability approach index at the corresponding time point is extracted and a historical sequence is formed in chronological order. In this historical sequence, the average and maximum values ​​of all instability approach indices are calculated. The average value is obtained by summing all values ​​and dividing by the number of data points, and the maximum value is obtained by comparing each value and selecting the largest value. Further, the average value is used as the lower limit of the safety boundary, and the maximum value is used as the upper limit of the safety boundary, thereby forming the initial safety boundary interval.

[0065] To ensure the safety boundary interval adapts to changes in operating conditions, it is dynamically updated during subsequent operation. Specifically, a sliding time window method is used to update the instability approximation index that has been determined to be in a stable state within a recent period. The time window length is set to 50 consecutive sampling points, and the update step is 10 sampling points. When new stable data enters the window, the average and maximum values ​​within that window are recalculated, and the safety boundary interval is updated, thus ensuring that the safety boundary interval reflects the latest stable operating characteristics.

[0066] After obtaining the safe boundary interval, the instability approximation index at the current moment is compared with this safe boundary interval to calculate the degree of deviation. Specifically, when the current instability approximation index is less than or equal to the upper limit of the safe boundary, it is determined to be within the safe range, and the corresponding deviation degree is recorded as 0; when the current instability approximation index is greater than the upper limit of the safe boundary, the upper limit of the safe boundary is subtracted from the current instability approximation index to obtain the value of the excess portion, and this value is divided by the upper limit of the safe boundary to obtain the normalized deviation ratio. For example, when the upper limit of the safe boundary is 0.2, and the current instability approximation index is 0.3, then the excess portion is 0.1, and the normalized deviation ratio is 0.1 divided by 0.2, which is 0.5.

[0067] After obtaining the normalized deviation ratio, it is mapped to a warning intensity value. Specifically, the normalized deviation ratio is limited to the range of 0 to 1, taking a value of 1 when it is greater than 1 and a value of 0 when it is less than 0, thus obtaining a standardized warning intensity value. This warning intensity value directly reflects the degree of deviation of the current state from the safety boundary; the larger the value, the higher the risk of instability.

[0068] After determining the warning intensity value, the alarm level is classified according to its numerical range. Specifically, the warning intensity value is divided into multiple intervals. For example, a warning intensity value less than 0.2 is classified as a Level 1 alarm; a warning intensity value greater than or equal to 0.2 and less than 0.5 is classified as a Level 2 alarm; a warning intensity value greater than or equal to 0.5 and less than 0.8 is classified as a Level 3 alarm; and a warning intensity value greater than or equal to 0.8 is classified as a Level 4 alarm. The above interval classification can be adjusted according to the system's risk tolerance, but it remains fixed within the same operating cycle to ensure consistency in judgment.

[0069] To avoid frequent changes in alarm levels due to momentary fluctuations, a continuity check is performed on alarm levels. Specifically, alarm levels at consecutive sampling points are statistically analyzed over time. An alarm level is considered valid if there are at least three consecutive time points with alarm levels all at or above the same level. For example, if three consecutive time points have alarm levels of level three or higher, the level three alarm is considered valid. If any time point in between has a lower level, the continuity check is repeated. This method filters out false alarms caused by short-term disturbances, improving the stability of alarm output.

[0070] In addition, to further enhance stability during the continuity verification process, a buffer mechanism can be set for changes in alarm levels. That is, when the current alarm level increases compared to the previous time point, the continuity condition must be met before it can be upgraded; when the alarm level decreases, the downgrade can be executed immediately when the condition is met at a single time point, thereby achieving cautious confirmation of risk increases and rapid response to risk decreases.

[0071] Through the above steps, a complete mapping process is achieved from the instability approach index to the warning intensity value and then to the alarm level. This method, by introducing dynamic safety boundaries, normalized deviation ratios, and continuity verification rules, ensures that the warning results reflect the system's current true risk level while avoiding misjudgments due to instantaneous fluctuations, thereby improving the accuracy and reliability of integrated alarms in the equipment process control system. Simultaneously, the warning intensity value and alarm level provide a clear basis for subsequent process adjustment commands, enabling timely intervention measures when the system approaches an instability state.

[0072] When the alarm level reaches the trigger condition, an alarm message is output, and a process control command matching the critical change range is generated.

[0073] In this embodiment, when the alarm level meets the triggering conditions, alarm output and process adjustment instructions are generated to achieve timely response and control of abnormal states. Specifically, the alarm level, confirmed by continuity verification, is first assessed for triggering conditions. These triggering conditions include the alarm level reaching a preset threshold and the duration meeting the minimum trigger duration requirement. The preset threshold can be set to alarm level three or higher, and the minimum trigger duration is set to three consecutive sampling points. When an alarm level is not lower than alarm level three for three consecutive sampling points, the alarm level is determined to meet the triggering conditions.

[0074] Upon fulfillment of the triggering conditions, a corresponding alarm message is generated. Specifically, the alarm message includes the alarm level, the dominant anomaly path identifier, the start and end times of the critical change interval, and the corresponding instability approximation index value. The dominant anomaly path identifier is represented by recording the order of measurement points involved in the anomaly propagation; the critical change interval is determined through the aforementioned asymptotic analysis steps; and the instability approximation index uses the calculation results within the corresponding time interval. The above information is combined according to a predetermined format, for example, outputting it in the structure of "alarm level - path identifier - time interval - index value," thereby achieving a complete description of the abnormal state.

[0075] While outputting alarm information, process control instructions matching the critical change range are generated. Specifically, firstly, based on the cumulative offset data sequence and its growth rate within the critical change range, the main direction of offset growth is determined, i.e., whether the overall trend of the correlation change value is upward or downward. When there is an upward trend, it indicates that the system offset is continuously expanding, requiring suppression by reducing the input or increasing the output; when there is a downward trend but it is still at a high level, the current control direction needs to be maintained and the rate of change controlled. Further, based on the average level of the cumulative offset value within the critical change range, the control amplitude is divided into multiple levels. For example, when the average cumulative offset value is less than 10, the control amplitude is set to 0.5 times the baseline value; when the average cumulative offset value is between 10 and 20, the control amplitude is set to the baseline value; when the average cumulative offset value is greater than 20, the control amplitude is set to 1.5 times the baseline value.

[0076] Subsequently, the adjustment direction and adjustment range are combined to form specific process adjustment instructions. For example, when it is determined that the input quantity is to be reduced and the adjustment range is 1.5 times the reference value, an instruction is generated to lower the corresponding device input control parameter by 1.5 times the reference adjustment amount; when it is determined that the output quantity is to be increased and the adjustment range is the reference value, an instruction is generated to increase the output control parameter by the reference adjustment amount. The reference adjustment amount is determined based on the average adjustment range during the normal operation of the device, for example, the average value of the adjustment range in historical adjustment response information can be taken as the reference.

[0077] After generating the process adjustment command, the command is associated with alarm information and output to the control execution unit of the corresponding device. To ensure the stability of the adjustment process, a feedback detection mechanism is set up during command execution. That is, the instability approach index is recalculated within 5 consecutive sampling points after command execution. When the index shows a downward trend and falls back to the safe boundary range, the adjustment command is deemed valid. If no downward trend is observed, the adjustment range is gradually adjusted according to predetermined rules, such as increasing the base adjustment amount by 0.2 times each time, until the instability approach index falls back or reaches the maximum adjustment limit.

[0078] In this embodiment, the above steps realize the closed-loop linkage of alarm triggering, information output and process adjustment, enabling the system to quickly generate targeted adjustment commands and verify their effects after identifying the critical change range, thereby effectively suppressing the abnormal development trend and improving the safety and stability of the equipment process control system.

[0079] Example 2, please refer to Figure 2 As shown in this embodiment, a system for integrating alarms in a process control system for equipment includes: Data acquisition module: acquires input and output data and adjustment response information of multiple devices during operation, and extracts conservation correlation feature sequences that reflect the equilibrium state of the system; Implicit Flaw Identification Module: Based on the conserved correlation feature sequence, it identifies abnormal segments where local correlations shift under the condition that the overall values ​​remain stable, forming a set of implicit flaw features; Bidirectional consistency analysis module: For the set of implicit broken features, a forward evolution process is constructed to obtain the changing trend of the anomaly over time, and a reverse backtracking process is constructed simultaneously to verify the continuity of the anomaly in historical stages, forming a bidirectional consistency feature sequence. Dominant path filtering module: Based on the bidirectional consistency feature sequence, filter out the abnormal paths that simultaneously meet the conditions of positive enhancement and reverse traceability, and use them as the dominant abnormal paths; Instability Analysis and Index Generation Module: Performs asymptotic analysis on the degree of offset accumulation in the dominant abnormal path, identifies the critical change interval from stable state to unstable state, and generates the corresponding instability approximation index; Early warning assessment module: compares the instability approach index with the safety boundary, determines the early warning intensity value, and classifies the alarm level according to the early warning intensity value; Adjustment execution module: When the alarm level reaches the trigger condition, it outputs alarm information and generates process adjustment instructions that match the critical change range.

[0080] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for integrating alarms in a process control system for equipment, characterized in that: include: Acquire input and output data and regulation response information of multiple devices during operation, and extract the conservation correlation feature sequence reflecting the system equilibrium state; Based on the conserved correlation feature sequence, abnormal segments in which local correlations shift under the condition that the overall values ​​remain stable are identified, forming a set of hidden broken features; For the set of implicit broken features, a forward evolution process is constructed to obtain the changing trend of the anomaly over time, and a reverse backtracking process is constructed simultaneously to verify the continuity of the anomaly in historical stages, forming a bidirectional consistent feature sequence. Based on the bidirectional consistency feature sequence, abnormal paths that simultaneously satisfy the conditions of positive enhancement and reverse traceability are selected as the dominant abnormal paths; A progressive analysis is performed on the degree of offset accumulation in the dominant abnormal path to identify the critical change interval from a stable state to an unstable state and generate the corresponding instability approximation index. The instability approach index is compared with the safety boundary to determine the warning intensity value, and the alarm level is classified according to the warning intensity value; When the alarm level reaches the trigger condition, an alarm message is output, and a process control command matching the critical change range is generated.

2. The method for integrating alarms in a process control system for equipment according to claim 1, characterized in that: All time windows that are determined to be in a stable or gradually changing state are spliced ​​together in chronological order, and the relationship between the input and output quantities that are consistent or change regularly within the corresponding time period is extracted. This relationship is recorded in the form of a time series, thus forming a conserved correlation feature sequence.

3. The method for integrating alarms in a process control system of equipment according to claim 1, characterized in that: The process of identifying anomalous segments where local correlations shift while overall numerical values ​​remain stable includes: The conserved correlation feature sequence is segmented according to time order, and the total change of input and output in each time period is calculated. Stable intervals with total change less than a preset threshold are selected. Within the stable interval, the degree of change in the correlation between each measuring point is further calculated. By comparing the difference in the corresponding correlation in adjacent time periods, the correlation offset data sequence is obtained. Based on the associated offset data sequence, segments with offset amplitudes greater than a preset offset threshold are identified at multiple consecutive time points to form candidate abnormal segments; The candidate abnormal segments are subjected to consistency screening, and the abnormal segments that persist within the stable interval and have the same offset direction are retained as the set of hidden defect features.

4. A method for integrating alarms in a process control system of equipment according to claim 1, characterized in that: The steps for constructing the positive evolution process include: Arrange the anomalous segments in the set of hidden broken features in chronological order, calculate the difference in the change value of the correlation between adjacent time points, form a positive change data sequence describing the speed of anomalous development, and construct a positive evolution process based on this.

5. A method for integrating alarms in a process control system of equipment according to claim 4, characterized in that: The formation of the bidirectional consistent feature sequence includes: Based on the positive change data sequence, segments in continuous time points where the change amplitude continuously increases or decreases are identified, and the corresponding abnormal development trends are extracted to form a change trend sequence of abnormality over time. Starting from the end time of the abnormal segment, the changes in the correlation value are traced back point by point in reverse time. It is determined whether the changes between adjacent time points are continuously connected during the tracing process, and a reverse tracing process is constructed. The sequence of changes in the anomaly over time is matched with the continuity determination results in the reverse backtracking process to filter out the anomaly segments that simultaneously satisfy both positive change consistency and reverse continuity, thus forming the bidirectional consistency feature sequence.

6. A method for integrating alarms in a process control system of equipment according to claim 5, characterized in that: The step of matching the sequence of changes in the anomaly over time with the continuity determination result in the reverse backtracking process includes: Assign trend direction and magnitude values ​​to each time point in the sequence of anomalies over time, and form a positive feature label sequence in chronological order. Based on the continuity determination result in the reverse backtracking process, a continuity identifier is generated for the corresponding time point, and time alignment processing is performed with the forward feature marker sequence to form dual sequence aligned data. In the dual-sequence aligned data, the correspondence between the trend direction indicator and the continuity indicator is compared point by point. When the two remain consistent for no less than 4 consecutive time points of a preset length, they are determined to be a matching segment. The matching segments are subjected to integrity screening, and segments that simultaneously satisfy the consistency of trend direction and the continuity coverage ratio of more than 80% are retained as valid matching results for constructing a bidirectional consistency feature sequence.

7. A method for integrating alarms in a process control system of equipment according to claim 1, characterized in that: The steps for filtering out abnormal paths that simultaneously meet the conditions of positive enhancement and reverse traceability include: The time segments in the bidirectional consistency feature sequence are divided into paths, and multiple candidate anomaly paths are constructed according to the connection order of the measurement points corresponding to the correlation relationship. Based on the candidate abnormal paths, the length of time during which the trend direction identifier in each path is continuously in the same direction is calculated, and the degree of positive enhancement is determined by combining the cumulative result of the change amplitude value. Perform reverse backtracking verification on the candidate abnormal paths, count the number of time points in the path that meet the continuity identifier, and calculate the proportion of the time points to the total number of time points in the path to characterize the degree of reverse traceability. The positive enhancement level and the reverse traceability level are jointly compared to select candidate abnormal paths that simultaneously satisfy the condition that the positive enhancement level is greater than the preset enhancement threshold and the reverse traceability level is greater than 80%, which are then selected as the dominant abnormal paths.

8. A method for integrating alarms in a process control system for equipment according to claim 1, characterized in that: The steps for asymptotically analyzing the degree of offset accumulation in the dominant anomalous path and generating an instability approximation index include: The correlation change values ​​at each time point are extracted along the dominant anomaly path in chronological order, and then accumulated point by point to form a cumulative offset data sequence; Based on the cumulative offset data sequence, the changes in the cumulative increment between adjacent time points are calculated, and the segments in the continuous time points where the cumulative increment changes from stable fluctuations to continuous growth are identified as candidate critical change intervals. The growth rate of the cumulative increment within the candidate critical change interval is analyzed. When the cumulative increment at consecutive time points shows an increasing trend and the duration is not less than 3 time points, the interval is determined to be the critical change interval. The instability approximation index is calculated based on the combination relationship between the cumulative offset value and the corresponding growth rate within the critical change interval.

9. A method for integrating alarms in a process control system for equipment according to claim 1, characterized in that: The steps for comparing the instability approach index with the safety boundary and determining the warning intensity value include: Based on the statistical results of the instability approximation index during the historical stable operation phase, its maximum value and average value are extracted to construct a safety boundary interval that is dynamically updated over time. The instability approach index at the current moment is compared with the safety boundary interval, the degree of deviation from the upper limit of the safety boundary is calculated, and the degree of deviation is mapped to the warning intensity value according to the ratio. The warning intensity values ​​are classified according to their numerical range. When the warning intensity values ​​are within a preset range, they are classified into different alarm levels. The alarm level is continuously verified. If the alarm level remains at or above the same level within a continuous time period, the alarm level is confirmed as a valid output.

10. A system for integrating alarms in a process control system for equipment, for implementing the method for integrating alarms in a process control system for equipment as described in any one of claims 1-9, characterized in that: include: Data acquisition module: acquires input and output data and adjustment response information of multiple devices during operation, and extracts conservation correlation feature sequences that reflect the equilibrium state of the system; Implicit Flaw Identification Module: Based on the conserved correlation feature sequence, it identifies abnormal segments where local correlations shift under the condition that the overall values ​​remain stable, forming a set of implicit flaw features; Bidirectional consistency analysis module: For the set of implicit broken features, a forward evolution process is constructed to obtain the changing trend of the anomaly over time, and a reverse backtracking process is constructed simultaneously to verify the continuity of the anomaly in historical stages, forming a bidirectional consistency feature sequence. Dominant path filtering module: Based on the bidirectional consistency feature sequence, filter out the abnormal paths that simultaneously meet the conditions of positive enhancement and reverse traceability, and use them as the dominant abnormal paths; Instability Analysis and Index Generation Module: Performs asymptotic analysis on the degree of offset accumulation in the dominant abnormal path, identifies the critical change interval from stable state to unstable state, and generates the corresponding instability approximation index; Early warning assessment module: compares the instability approach index with the safety boundary, determines the early warning intensity value, and classifies the alarm level according to the early warning intensity value; Adjustment execution module: When the alarm level reaches the trigger condition, it outputs alarm information and generates process adjustment instructions that match the critical change range.

Images