An abnormal condition intelligent identification system based on chemical production process

Abstract

The present application relates to the technical field of abnormal condition identification, in particular to an abnormal condition intelligent identification system based on chemical production process, which comprises a signal synchronization module, a turning point identification module, a phase construction module, an offset integration module and an abnormality determination module. The present application constructs a multi-measurement-point data time sequence structure around a unified time reference, establishes a variable response sequence expression form in combination with a continuous sampling change turning point position, concatenates response relationships between different process sections into a continuous link structure, sorts and divides the direction continuation in the link, reconstructs a discrete change process into a correlation structure with process sequence characteristics, and then completes abnormal position definition based on the section direction connection relationship, so that the transmission relationship between variables forms a clear evolution path in the continuous change process, enhances the completeness of the multi-variable linkage relationship expression, and improves the stability of the sequence determination in the working condition identification process.

Description

Technical Field

[0001] This invention relates to the field of abnormal operating condition identification technology, and in particular to an intelligent identification system for abnormal operating conditions based on chemical production processes. Background Technology

[0002] Abnormal operating condition identification primarily involves the monitoring and analysis of the operational status of industrial processes. Through the collection and processing of multi-source data during production, it identifies and classifies operating conditions that deviate from normal operating conditions, focusing on process safety assurance and operational status awareness. It is widely used in chemical, petroleum, power, and other process industries. Among these, the intelligent abnormal operating condition identification system for traditional chemical production processes refers to a system that analyzes and judges changes in the state of various operating parameters during chemical production. This system is used to identify abnormal operating states, typically based on data such as temperature, pressure, and flow rate collected by the process control system, combined with data analysis methods and state discrimination rules, to continuously monitor and identify the operating conditions of the production process.

[0003] Existing abnormal operating condition identification revolves around single-variable parameter changes and fixed discrimination rules. However, deviations in time stamping during data acquisition at various measurement points make it difficult to form a unified time series reference for changes in different parameters. The sequential relationship between variables lacks a continuous description method, and the transmission paths between process sections are discrete. The coupling relationship is difficult to form a complete link expression, resulting in the unstable presentation of the correlation between changes in various variables during operation. In the process of multi-variable linkage, the order of judgment may be misjudged, which in turn makes it difficult to have an effective basis for identifying some abnormal states in the evolution stage, affecting the consistency of the overall operating status judgment. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of existing technologies and propose an intelligent identification system for abnormal operating conditions in chemical production processes.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: an intelligent identification system for abnormal operating conditions in a chemical production process, the system comprising: The signal synchronization module is based on chemical production equipment. It calibrates the correspondence between the measurement point wiring port and the signal channel, judges the current, voltage and pulse sampling status, unifies the controller clock and the acquisition time stamp sequence, determines the channel differences, screens off-target measurement points, and obtains a unified time stamp data sequence. The transition recognition module, based on the unified time-stamped data sequence, checks the changes in sampling direction before and after the same measurement point, identifies the locations of rising and falling transitions, counts the intervals between adjacent transitions within the sampling window, filters out continuous transition segments, and obtains the transition response position sequence. Based on the transition response position sequence, the phase construction module compares the sequence of transitions between the preceding and following process segments, distinguishes the order of the leading and lagging states, organizes the marking order of the process connection segments, determines the link continuation direction, and obtains the process phase association chain. The offset integration module, based on the process phase correlation chain, checks the direction continuity within a continuous time window, compares whether adjacent markers are in the same direction, counts the corresponding time sequence of the start and end points of continuous segments, screens the process position to which the segment belongs, and obtains the phase offset connection record. Based on the phase offset segment record, the anomaly determination module analyzes the connection relationship between the process positions before and after, compares whether the directional changes of adjacent segments are continuous, distinguishes the corresponding positions of the same-direction segments and the turning segments, and obtains the anomaly identification result.

[0006] The present invention is improved in that the unified time-stamped data sequence includes a time alignment sequence, a sampling order sequence, and a signal consistency sequence; the transition response position sequence includes a transition position sequence, a transition direction sequence, and a transition interval sequence; the process phase correlation chain includes a phase order sequence, a phase correlation sequence, and a direction marking sequence; the phase offset segment record includes a segment interval sequence, a direction continuity sequence, and a segment affiliation sequence; and the operating condition anomaly identification result includes an abnormal segment identifier, a direction change classification, and a process node identifier.

[0007] The present invention is improved in that the signal synchronization module includes: The channel-corresponding submodule is based on chemical production equipment. It analyzes the measurement point wiring port identifier and signal channel identifier, compares the corresponding order of the same batch of acquisition records, determines whether the access position and input path are misaligned, adjusts the registration order of misaligned records, filters corresponding consistent records, and obtains the port channel mapping sequence. Based on the port channel mapping sequence, the state conversion submodule obtains the single-channel signal record and wiring type, determines the compatibility relationship of current signal, voltage signal and pulse signal, adjusts the conversion order, calculates the engineering quantity conversion state, compares the continuity relationship of sampling records before and after the same source measurement point, and obtains the engineering quantity state sequence. The timing compilation submodule obtains the controller clock and acquisition time stamp based on the engineering quantity state sequence, compares the sequential relationship of the channel records in the same batch, determines the position of the time stamp offset measurement point, adjusts the timing arrangement of the offset records, filters and aligns the records, and obtains a unified time stamp data sequence.

[0008] The present invention is improved in that the turning point recognition module includes: The direction discrimination submodule analyzes the direction of continuous sampling records at the same measuring point based on the unified time-stamped data sequence, compares the relationship between the change direction of the sampling records before and after the turning position, determines the sampling point to which the rising or falling turning position belongs, filters the direction switching records, and obtains the turning mark sequence. The turning interval submodule analyzes the temporal relationship between adjacent turning positions based on the turning mark sequence, compares the interval between the previous turning position and the next turning position within the continuous sampling window, calculates the turning connection order, filters out continuously occurring turning records, and obtains the turning time interval sequence. The response sorting submodule analyzes the timing of each measurement point's turning point based on the turning point time interval sequence, compares the turning point sequence of each measurement point within the same sampling window, determines the corresponding position of the process section response order, adjusts the order of the turning point records, and obtains the turning point response position sequence.

[0009] The present invention is improved in that the phase construction module includes: Based on the transition response position sequence, the transition association submodule checks the transition location one by one according to the corresponding measurement points of the preceding and following process segments, compares the front and rear fit relationship of the transitions of adjacent process segments in the same sampling window, determines whether the paired records are continuously corresponding, and obtains the process segment pairing sequence. The sequence determination submodule compares the timing relationship between the transition times of adjacent process segments based on the process segment pairing sequence, determines the order of the leading state and the lagging state, adjusts the record arrangement order, and obtains the leading-lagging sequence. The link direction determination submodule checks the upstream and downstream connection order according to the leading and lagging states based on the preceding and lagging order sequence, compares whether the marking directions of adjacent process segments are consistent, determines the link connection direction, and obtains the process phase association chain.

[0010] The present invention is improved in that the offset integration module includes: The continuation check submodule analyzes the arrangement order of adjacent markers within a continuous time window based on the process phase correlation chain, compares the correspondence between the direction of the previous marker and the direction of the next marker, determines whether the direction switching position falls into the same continuous segment, filters records with unchanged direction, and obtains the direction continuation sequence. The segment delimitation submodule calculates the corresponding time sequence of the first and last records of a continuous segment based on the direction continuation sequence, compares the correspondence between the interruption position and the continuation position of adjacent records, adjusts the arrangement order of the segment boundaries, determines the start and end range of the segment, and obtains the segment interval sequence. The attribution determination submodule analyzes the process position of continuous segments based on the segment interval sequence, compares the directional consistency and position connection relationship of adjacent segments, determines the segment attribution status, adjusts the position order and the correspondence with the attribution mark, and obtains the phase offset segment record.

[0011] The present invention is improved in that the anomaly determination module includes: The connection verification submodule, based on the phase offset connection record, checks the sequential order of the process position, compares the registration of directional changes in adjacent sections, determines whether the connection of continuous sections is broken, filters out abnormal connection records, and obtains the connection relationship sequence. Based on the connection sequence, the segment positioning submodule calculates the corresponding positions of the same-direction segments and turning segments, determines the process link to which the abnormal position belongs, adjusts the segment positioning order, and obtains the segment positioning quantity. The boundary alignment submodule, based on the segment positioning data, verifies the start and end positions of the same record with the corresponding content of the process node, compares the ownership and continuation status of adjacent records, determines whether the boundary cutting position is misplaced, aligns the ownership boundary, and obtains the abnormal working condition identification result.

[0012] The present invention is improved in that the signal channel refers to the input path corresponding to each sensor signal, the channel sequence refers to the time order of acquisition or recording of different signals within the same sampling period, and the offset measurement point refers to the time offset measurement point signal where the sampling time is not aligned with the system clock.

[0013] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, a multi-measurement point data time sequence structure is constructed around a unified time reference. A variable response sequence expression is established by combining the turning points of continuous sampling changes. The response relationship between different process sections is connected into a continuous link structure. The directional continuity within the link is categorized and divided. The discrete change process is reconstructed into a correlation structure with process sequence characteristics. Then, the abnormal position is defined based on the directional connection relationship of the sections. This enables the transmission relationship between variables to form a clear evolution path in the continuous change process, enhances the completeness of the expression of multi-variable linkage relationship, and improves the stability of the sequence determination in the process of working condition identification. Attached Figure Description

[0014] Figure 1 This is a system flowchart of the present invention; Figure 2 This is a flowchart of the signal synchronization module in this invention; Figure 3 This is a flowchart of the transition recognition module in this invention; Figure 4 This is a flowchart of the phase construction module in this invention; Figure 5 This is a flowchart of the offset integration module in this invention; Figure 6 This is a flowchart of the anomaly detection module in this invention. Detailed Implementation

[0015] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0016] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0017] All user-related information involved in this invention (including but not limited to biometric information, identity verification information, behavioral data, device information, and other data that can be used for identity verification and personalized services) is collected and processed with the user's full knowledge and voluntary consent. The use of data is limited to purposes necessary for providing the technical services of this invention, and reasonable technical and management measures will be taken to ensure the security and confidentiality of users' personal information in terms of information protection and privacy.

[0018] Example: Please refer to Figure 1 This invention provides a technical solution: an intelligent identification system for abnormal operating conditions in a chemical production process, comprising: The signal synchronization module is based on chemical production equipment. It analyzes the correspondence between the measurement point wiring port and the signal channel, calculates the current and voltage pulse conversion status within the sampling period, adjusts the corresponding order of the controller clock and the acquisition time scale, judges whether the channels are consistent, filters the time scale offset measurement points, and obtains a unified time scale data sequence. The transition identification module is based on a unified time-scaled data sequence. It analyzes the changes in sampling direction before and after the same measurement point, compares the temporal relationship between the rising transition position and the falling transition position, calculates the interval between adjacent transitions within the sampling window, filters continuous transition segments, determines the response order of the process section, and obtains the transition response position sequence. The phase construction module analyzes the correspondence between transitions in the preceding and following process segments based on the transition response position sequence, compares the order of transitions in adjacent process segments, calculates the order of leading and lagging states, adjusts the marking order of process connection segments, determines the link continuation direction, and obtains the process phase association chain. The offset integration module analyzes the directional continuity relationship within a continuous time window based on the process phase association chain, compares whether adjacent markers maintain the same direction, calculates the time sequence corresponding to the start and end points of continuous segments, filters the process positions to which segments with the same direction belong, determines the segment ownership status, and obtains the phase offset connection record. The anomaly detection module analyzes the connection relationship between process positions based on phase offset segment records, compares whether the directional changes of adjacent sections are continuous, calculates the corresponding positions of the same-direction sections and turning sections, filters the process links to which the abnormal position belongs, adjusts the boundary of the same record, and obtains the results of the abnormal condition identification.

[0019] The unified time-stamped data sequence includes time-aligned sequence, sampling order sequence, and signal consistency sequence; the turning response position sequence includes turning position sequence, turning direction sequence, and turning interval sequence; the process phase association chain includes phase order sequence, phase association sequence, and direction mark sequence; the phase offset segment record includes segment interval sequence, direction continuity sequence, and segment affiliation sequence; and the operating condition anomaly identification results include abnormal segment identification, direction change classification, and process node identification.

[0020] In the signal synchronization module, chemical production equipment refers to process units such as reactors, heat exchangers, pipelines, regulating valves, circulating pumps, and storage tanks; measuring point wiring ports refer to the wiring positions corresponding to the sensor signals connected to the control system; signal channels refer to the input paths corresponding to each sensor signal in the control system; current, voltage, and pulse conversion status refers to the status data after converting current signals, voltage signals, and pulse signals into engineering quantities such as temperature, pressure, flow rate, and liquid level; controller clock refers to the system clock used by the PLC or DCS system to unify the time reference; acquisition time stamp refers to the time tag corresponding to each sampled data; channel sequence refers to the time order of acquisition or recording of different signals within the same sampling period; time stamp offset measuring points refer to measuring point signals whose sampling time is not aligned with the system clock and have a time offset.

[0021] In the transition recognition module, the change in sampling direction refers to the change in value from rising to falling or from falling to rising in continuous sampled data; the rising transition position refers to the time point when the signal changes from a falling trend to an rising trend; the falling transition position refers to the time point when the signal changes from an rising trend to a falling trend; the adjacent transition interval refers to the time interval or sampling step size difference between two consecutive transition points; the continuous transition segment refers to the interval of transition behavior signals that continuously appear within multiple continuous sampling windows; the process segment response order refers to the order in which different equipment or variables change under the same disturbance.

[0022] In the phase construction module, the correspondence relationship refers to the temporal relationship between variables in two process segments; the order of turning points refers to the time sequence of the turning points of different variables; the leading state refers to the state in which the change of one variable occurs before the occurrence of another variable; the lagging state refers to the state in which the change of one variable occurs after the occurrence of another variable; the process connection segment refers to the relationship between connected equipment or variables in the process flow; and the link connection direction refers to the relationship of variable response sequence formed along the process flow direction.

[0023] In the offset integration module, directional continuity refers to the maintenance of the leading or lagging relationship between variables within a continuous time window; adjacent markers refer to the corresponding leading or lagging identifiers of adjacent variable pairs in the phase chain; continuous segment start and end points refer to the start and end positions where a certain directional relationship continues to appear; directional consistent segments refer to recorded segments in which the leading or lagging relationship has not changed within a continuous time; process position refers to the position of the segment corresponding to a specific equipment or process link; and attribution status refers to the status of the segment being classified as a leading or lagging type.

[0024] In the anomaly determination module, the preceding and following connections refer to the connection relationship and sequence between different process segments in the process flow; whether the directional change is continuous refers to whether the leading or lagging relationship between adjacent process segments remains continuous and consistent; the corresponding position refers to the specific equipment position or process node corresponding to the same or turning relationship; the abnormal position refers to the process link where the directional change occurs or is inconsistent with the normal response sequence; the attribution boundary refers to the starting and ending position of an anomaly record in time or process flow.

[0025] Please see Figure 2 The signal synchronization module includes: The channel-corresponding submodule is based on chemical production equipment. It analyzes the measurement point wiring port identifier and signal channel identifier, compares the corresponding order of the same batch of acquisition records, determines whether the access position and input path are misaligned, adjusts the registration order of misaligned records, filters corresponding consistent records, and obtains the port channel mapping sequence. When dealing with on-site objects such as reactor temperature measuring points, heat exchanger pressure measuring points, and pipeline flow measuring points, retrieve the port number table and control cabinet input channel registration table from the equipment ledger. Create a one-to-one correspondence list for port 1, port 2, and port 3 with channel 2, channel 3, and channel 1, categorized by batch. Then, expand each record in the order of its registration. Assuming the writing times of the three records are 10.000 seconds, 10.001 seconds, and 10.003 seconds respectively, first arrange them from earliest to latest time. Then, check the port number and channel number corresponding to each record. First, check if the port number and channel number are in the same sequence. Then, check if there is a jump between adjacent records. If port 1 falls into channel 2, port 2 falls into channel 3, and port 3 falls into channel 1, then... The first two offsets are recorded as adjacent offsets, and the second offset is recorded as a two-bit offset. Then, the allowable offset value is set to 1 bit, because if there is an adjacent bit in the same batch of registration or writing, it can still be rearranged. If it exceeds 1 bit, it is recorded as an obvious misalignment. Then, the misaligned records are rearranged, and the cross-bit records are moved to the last verification area. The adjacent offset records are rearranged according to the natural order of the ports. The correspondence between the ports and channels in the same position after rearrangement is checked again. If the port number and channel number still differ by 2 bits after rearrangement, they are directly removed. If they differ by 0 bits or 1 bit, they are retained and added to the consistent record set. For example, port 1 to channel 2 and port 2 to channel 3 are retained first, and port 3 to channel 1 is moved to the waiting area. Finally, the retained records are arranged in ascending order of port number, and each record is attached with batch number and time number to form a port channel mapping sequence.

[0026] The state conversion submodule obtains single-channel signal records and wiring types based on the port channel mapping sequence, determines the compatibility relationship of current signals, voltage signals and pulse signals, adjusts the conversion order, calculates the engineering quantity conversion state, compares the continuity relationship of sampling records before and after the same source measurement point, and obtains the engineering quantity state sequence. After receiving the port and channel correspondences retained in the previous stage, the system extracts the original signal records and wiring type registration information for each single channel one by one. Each record is then categorized into three types: current, voltage, and pulse, and placed into the corresponding processing queues. For example, if a temperature measuring point is registered as a current connection, with a current sampling value of 12, a lower limit of 4, an upper limit of 20, and an engineering quantity range of 0 to 200, the system first checks whether 12 falls within the range of 4 to 20. If it does, the system continues to calculate, first determining the number of values ​​exceeding the lower limit by 8, then calculating the total span of 16, and finally converting 8 and 16 proportionally to obtain a range percentage of 0.5. This percentage is then mapped to the engineering quantity range of 0 to 200 to obtain the state value of 100. If another pressure measuring point is registered as a voltage connection, with a sampling value of 5, a range of 0 to 10, and an engineering quantity range of 0 to 1.6, the system first confirms that 5... If the flow measurement is in the middle of the range, calculate the state value of 0.8 by converting it to half the range. If a flow measurement point is registered as a pulse connection, and the previous cumulative value is 240 and the current value is 248, then take the difference of 8. Then convert it according to the equipment registration that every 8 pulses correspond to 1 flow unit to get the state value of 1. After the conversion is completed, continue to check whether the records of the same source measurement point are continuous. For example, if the temperature changes from 98 to 100, the change is 2. Set the continuous threshold to 5. Since the normal temperature fluctuation in a single sampling cycle of a chemical plant usually does not exceed 5, if the change is not greater than 5, it is recorded as continuous. If the pressure changes from 0.7 to 1.2, the change is 0.5. Set the pressure continuous threshold to 0.3. Then the record is recorded as a jump to be checked. Finally, only the records that have been converted and the continuity relationship between the two records is established are sorted in chronological order to form the engineering quantity state sequence.

[0027] The timing compilation submodule obtains the controller clock and the acquisition time stamp based on the engineering quantity state sequence, compares the sequential relationship of the records in the same batch of channels, determines the location of the time stamp offset measurement point, adjusts the timing arrangement of the offset records, filters and aligns the records, and obtains a unified time stamp data sequence. List the controller's unified clock and acquisition timestamp for each record, then place records from the same batch of channels into a time-verification table. For example, if the controller's base time is registered as 10.000 seconds, the temperature status record timestamp is 10.002 seconds, the pressure status record timestamp is 9.998 seconds, and the flow status record timestamp is 10.000 seconds, first calculate the offset of each record relative to the controller clock: temperature is 0.002 seconds ahead, pressure is 0.002 seconds behind, and flow has no offset. Then set the timestamp offset judgment value to 0.001 seconds. Because under the same controller, if the batch of acquisitions and writes are in a normal synchronization state, the single-channel error is controlled within one-thousandth of a second. Any record exceeding this value is recorded as an offset measurement point. Finally, sort the records in the same batch according to their order. The original time sequences of 9.998 seconds, 10.000 seconds, and 10.002 seconds are obtained. The slower records are then padded back by 0.002 seconds, and the faster records are pushed forward by 0.002 seconds, so that the three records are all aligned to 10.000 seconds. The adjusted record order is then checked to see if it still conflicts with the channel registration order. If a record is corrected to be the same time but the channel position is two positions ahead, it is rearranged back to its original position according to the original channel number. If the corrected time is the same and the channel order no longer conflicts, it is included in the aligned record set. Then, all aligned records are reorganized according to the unified time and channel number. For example, temperature 100, pressure 0.8, and flow rate 1 all fall at the time of 10.000 seconds. The batch number, channel number, and equipment location number are then added to obtain a unified time-stamped data sequence.

[0028] Please see Figure 3 The transition recognition module includes: The direction discrimination submodule analyzes the direction of continuous sampling records at the same measuring point based on a unified time-scaled data sequence, compares the correspondence between the changing direction of the sampling records before and after the turning position, determines the sampling point to which the rising or falling turning position belongs, filters the direction switching records, and obtains the turning mark sequence. The system receives a sequence of single-point temperature or pressure measurements at a unified time scale. For example, the temperature of the same reactor is recorded as 98, 100, 103, 101, and 99 at five consecutive sampling points. It then reads the difference between adjacent points sequentially, first calculating 100 minus 98 to get 2, then 103 minus 100 to get 3, then 101 minus 103 to get -2, and finally 99 minus 101 to get -2. The resulting difference sequence is recorded as [2, 3, -2, -2]. Subsequently, the sign of each difference is determined: positive values ​​are marked as an upward trend, negative values ​​as a downward trend, and zero values ​​are marked as a stable segment. This forms a direction sequence [upward, upward, downward, downward]. Adjacent directions are then compared pairwise. The first pair of upward trends is determined as [upward, upward, downward]. No turning occurred. The second pair, ascending and descending, was determined to have a direction change, and the sampling point 103 corresponding to this position was marked as a turning candidate point. The third pair, descending and descending, was determined not to have a turning. Then, the candidate points were further distinguished by type. The difference before and after the turning was taken as 3 and -2 respectively. If the previous value was positive and the subsequent value was negative, the point was marked as a descending turning point. If the previous value was negative and the subsequent value was positive, it was marked as an ascending turning point. The effective threshold for direction was set to 1. When the absolute value of the difference was less than 1, it did not participate in the turning determination. For example, if the difference was 0.5, it was classified into the stable region and no turning record was formed. Finally, the sampling points that met the direction change and whose absolute values ​​of the difference were all greater than 1 were extracted. For example, point 103 was retained to form a turning mark sequence arranged by time.

[0029] The turning interval submodule analyzes the temporal relationship between adjacent turning positions based on the turning mark sequence, compares the interval between the previous turning position and the next turning position within the continuous sampling window, calculates the turning connection order, filters out continuously occurring turning records, and obtains the turning time interval sequence. After receiving the turning point sequence, the timestamps corresponding to each turning point are listed one by one. For example, if the three turning points occur at 10.000 seconds, 10.006 seconds, and 10.020 seconds, the time series is first sorted in ascending order, and then the interval between adjacent points is calculated item by item. The first interval is 10.006 minus 10.000, which is 0.006 seconds. The second interval is 10.020 minus 10.006, which is 0.014 seconds. The interval values ​​are recorded as [0.006, 0.014]. Then, each interval is divided into intervals, and the short interval threshold is set to 0.01 seconds. This value is obtained by multiplying the on-site sampling period of 0.005 seconds by 2. When the interval is less than 0.01 seconds, the threshold is set to 0.01 seconds. A transition period of 0.01 seconds or less is considered a dense transition period, while a transition period greater than 0.01 seconds is considered a sparse transition period. Here, 0.006 is classified as a dense transition period, and 0.014 is classified as a sparse transition period. The continuous sampling window is then checked, with a window length of 0.02 seconds. The number of transition points within this window is counted. If the number reaches two or more, it is marked as a continuous transition area. For example, if there are three transition points in the interval from 10.000 to 10.020, the condition is met. The transition points are then rearranged in their original time order, and their interval status is recorded. At the same time, their order is sorted according to the interval size, for example, dense transitions are sorted before sparse transitions, forming a transition time interval sequence with time interval identifiers and order.

[0030] The response sorting submodule analyzes the timing of the turning points at each measurement point based on the turning time interval sequence, compares the turning order of each measurement point within the same sampling window, determines the corresponding position of the response sequence of the process section, adjusts the order of the turning records, and obtains the turning response position sequence. The process is based on the transition time records of multiple measuring points. For example, the transition time for reactor temperature is 10.006 seconds, the transition time for heat exchanger pressure is 10.008 seconds, and the transition time for pipeline flow is 10.004 seconds. First, all transition times within the same sampling window are extracted and arranged as [10.004, 10.006, 10.008]. Then, the difference between each transition time and the window start time of 10.000 seconds is calculated, yielding values ​​of 0.004, 0.006, and 0.008. These differences are used as the sorting criteria; a smaller difference indicates an earlier response. Subsequently, the three measuring points are sorted according to their differences. The flow rate, temperature, and pressure are sorted from smallest to largest to obtain their order. Then, the corresponding process location is labeled for each item in the sorted results. For example, flow rate corresponds to the pipe inlet, temperature corresponds to the reactor body, and pressure corresponds to the heat exchanger outlet. Next, the original transition record sequence is rearranged to make its order consistent with the above sorting. At the same time, it is checked whether there are any parallel cases with a time difference of less than 0.001 seconds. If so, a second sorting is performed according to the physical process order of the equipment. For example, the pipe comes first, the reactor is in the middle, and the heat exchanger comes last. Finally, the sorted transition records are output in both time and process location order to form a transition response location sequence.

[0031] Please see Figure 4 The phase construction module includes: The transition correlation submodule, based on the transition response position sequence, checks the transition location one by one according to the corresponding measurement points of the preceding and following process segments, compares the front and rear fit relationship of the transitions of adjacent process segments within the same sampling window, determines whether the paired records correspond continuously, and obtains the process segment pairing sequence. After receiving the transition response location records, the measuring points are arranged into continuous process segments according to the process flow sequence. For example, the pipeline flow rate, reactor temperature, and heat exchanger pressure are numbered 1, 2, and 3 respectively. The transition time and sampling point number corresponding to each measuring point within the same sampling window are extracted. For example, the times are 10.004, 10.006, and 10.008, and the sampling point numbers are 801, 802, and 804. Then, adjacent process segments are checked one by one. First, the records corresponding to number 1 and number 2 are taken, and the time difference 10.006 minus 10.004 is calculated to get 0.002. Then, the sampling point difference 802 minus 801 is calculated to get 1. The time difference is then compared with the set threshold 0. The time difference 0.002 and the sampling point difference 2 are compared. If neither exceeds the threshold, they are marked as consecutively corresponding. The same operation is then performed on number 2 and number 3 to obtain the time difference 0.002 and the sampling point difference 2. If they also meet the condition, they are marked as consecutively corresponding. If either exceeds the threshold, they are marked as disconnected records and removed. All consecutively corresponding records are then rearranged according to the process segment order, keeping the numbering increment relationship unchanged. At the same time, the offset of each pair of records relative to the window start point 10.000 is calculated and the matching interval is recorded. Records with an interval in the range of 0 to 0.003 are retained, forming a process segment pairing sequence that corresponds to adjacent process segments one by one.

[0032] The sequence determination submodule is based on the process segment pairing sequence. It compares the timing relationship between the transition times of adjacent process segments, determines the order of the leading and lagging states, adjusts the record arrangement order, and obtains the leading and lagging sequence. Each pair of records is split into the time of the previous measurement point and the time of the next measurement point, and the time difference is calculated for each pair. For example, the first pair is 10.004 and 10.006, with a difference of 0.002; the second pair is 10.006 and 10.008, with a difference of 0.002. The time difference is then divided into intervals: 0 to 0.001 is designated as the near-synchronous zone, 0.001 to 0.004 as the clearly defined sequence zone, and greater than 0.004 as the long interval zone. Since the differences all fall within the clearly defined sequence zone, the record corresponding to the previous time is marked as the leading record, and the record corresponding to the next time is marked as the lagging record. If the difference is negative, it is marked as reversed and moved to the abnormal area. Then, all records determined to be leading and lagging are sorted in ascending order of leading time, for example, 10.004 first, 10.006 in the middle, and 10.008 last. It is then checked whether there is a time difference less than 0.001. If so, the process position number is introduced for secondary sorting. For example, number 1 is prioritized over number 2. The sorted leading and lagging records are then expanded according to the process order, so that the previous pair of lagging items and the next pair of leading items are sequentially connected. Finally, a leading and lagging order sequence is obtained by arranging the records in both time and position order.

[0033] The link direction determination submodule is based on the pre-lag sequence. It checks the upstream and downstream connection order according to the relationship between the leading state and the lagging state, compares whether the marked directions of adjacent process sections are consistent, determines the link continuation direction, and obtains the process phase association chain. The preceding and following measurement points and their process position numbers are extracted item by item from each record. Adjacent records are then checked for connection. For example, if the first record is from position 1 to position 2 and the second record is from position 2 to position 3, the following checks are performed to see if the following measurement point in the first record is consistent with the preceding measurement point in the second record. If they are consistent, the record is considered connectable. The difference between the position numbers is then calculated. The difference between position 1 and 2 is 1, and the difference between position 2 and 3 is 1. Records with a difference of 1 are marked as positive relationships, and records with a difference of -1 are marked as negative relationships. If the absolute value of the difference is greater than 1, it is marked as a segment crossing and removed. The consistency of the directions of adjacent records is then compared. If both segments are positive, they are retained. If one is positive and the other is negative, the link is broken at that point. Subsequently, all records with consistent directions and continuous connection are connected in chronological order. For example, 10.004 corresponds to 1 to 2, and 10.006 corresponds to 2 to 3. These are then spliced ​​together to form a continuous link. The measurement point and time of each node are written to the corresponding position. Finally, a process phase association chain containing sequential relationships and direction markings is obtained.

[0034] Please see Figure 5 The offset integration module includes: The continuation check submodule is based on the process phase correlation chain. It analyzes the arrangement order of adjacent markers within a continuous time window, compares the correspondence between the direction of the previous marker and the direction of the next marker, determines whether the direction switching position falls into the same continuous segment, filters records with unchanged direction, and obtains the direction continuation sequence. Expand the link markers within the same continuous time window chronologically, and list the process position number, direction marker, and recording time for each marker. For example, within a certain window, three records are formed: position 1 to position 2 (forward), position 2 to position 3 (forward), and position 3 to position 4 (reverse), with times of 10.004, 10.006, and 10.009 respectively. Then, check the direction of each pair of adjacent records. First, compare the first forward record with the second forward record; if the directions match, register it as a continuation relationship. Then, compare the second forward record with the third reverse record; if the directions do not match, register it as a switching relationship. Simultaneously, record the switching time 10.009 and the corresponding positions 3 to 4. Finally, match the direction switching position with the current continuous segment range. For comparison, the provisional rule for continuous segments is that the time interval between adjacent records does not exceed 0.003 and the position number increases or decreases by 1 consecutive position. In this example, the interval between 10.006 and 10.009 is 0.003, which meets the time requirement, but the direction has changed. Therefore, the first two records are grouped into the same continuation segment, and the third record is assigned to the starting point of a new segment. Then, all records are filtered, and only records with consistent direction and no time interval exceeding the limit are retained. For example, the first and second records are retained, and the third record is saved separately. If the time interval of a record reaches 0.005 or the position jumps by 2 positions, it is directly recorded as an interruption and is not included in the direction continuation set. Finally, the retained records are sorted in ascending order of time and order of position to form a direction continuation sequence.

[0035] The segment delimitation submodule calculates the time sequence corresponding to the first and last records of a continuous segment based on the direction continuation sequence, compares the correspondence between the interruption position and the continuation position of adjacent records, adjusts the arrangement order of the segment boundaries, determines the start and end range of the segment, and obtains the segment interval sequence. Extract the first record time, last record time, first position, and last position from each continuous segment, and arrange them from earliest to latest. For example, in one continuous segment, the first record time is 10:004 and the last record time is 10:006, corresponding to positions 1 to 3; in another continuous segment, the first record time is 10:009 and the last record time is 10:012, corresponding to positions 3 to 5. Then, first check the beginning and end of each segment, writing the time difference between the first and last records as the interval length. The first segment has a length of 0.002, the second segment has a length of 0.003, and so on. Next, compare the breakpoints and continuation points between adjacent segments, starting with the last record time of the previous segment, 10:00. The interval between 06 and the first time of the next segment, 10.009, is 0.003. Then, the end position 3 of the previous segment is compared with the beginning position 3 of the next segment. If the positions are the same, it is registered as a continuation of the edge. If the time interval does not exceed 0.003 and the position difference does not exceed 1, the two segments are included in the rearrangement area. If the time exceeds 0.003 or the position difference reaches 2, the original boundary is kept unchanged. In this example, the time is exactly equal to 0.003 and the positions are the same, so the two segments are put into the correction table. Then, the boundaries are rearranged in the order of the start point first and the end point to ensure that the start time of each segment is less than the end time and the position numbers are not reversed. Finally, the start and end times and corresponding position ranges of each segment are given to form a sequence of connected segments.

[0036] The attribution determination submodule analyzes the process position of continuous segments based on the segment interval sequence, compares the directional consistency and position connection relationship of adjacent segments, determines the segment attribution status, adjusts the position order and the correspondence with the attribution mark, and obtains the phase offset segment record. For each continuous segment, the start position, end position, segment direction, and corresponding time range are extracted. Then, the segments are unfolded sequentially according to the process flow. For example, the first segment is forward-facing, positions 1 to 3, time 10.004 to 10.006; the second segment is forward-facing, positions 3 to 5, time 10.009 to 10.012; and the third segment is reverse-facing, positions 5 to 4, time 10.015 to 10.017. Next, the directional relationship between adjacent segments is compared. If the first and second segments are both forward-facing, they are considered to have the same direction. If the second and third segments are reverse-facing, they are considered to have different directions. Then, the positional connection is compared. If the end point 3 of the first segment is the same as the start point 3 of the second segment, it is recorded as a direct connection. The endpoint 5 of the second segment is the same as the starting point 5 of the third segment, and is also registered as a direct connection. However, due to the different directions, a boundary is set between the second and third segments. Then, the segments with the same direction and the established position connection are merged and sorted. The first and second segments are listed in the same group according to the starting time 10.004 and 10.009, respectively. The third segment is listed in another group. If the direction is the same but the position difference is more than 2 positions, it is transferred to the separation group and is not merged into the same state. Then, the position order of each segment is adjusted to correspond with the position mark so that the segments in the same group maintain the position increment or decrement consistency. Finally, the phase offset connection record with the segment interval, direction continuity, and process position position mark is output.

[0037] Please see Figure 6 The anomaly detection module includes: The connection verification submodule, based on the phase offset connection record, checks the sequential order of the process position, compares the registered contents of the directional change of adjacent sections, determines whether the connection of continuous sections is broken, filters out abnormal connection records, and obtains the connection relationship sequence. The starting and ending position numbers, direction markers, and corresponding time intervals of each segment are expanded one by one and sorted from earliest to latest. For example, the first segment has positions 1 to 3, a forward direction, and times from 10:004 to 10:006; the second segment has positions 3 to 5, a forward direction, and times from 10:009 to 10:012; and the third segment has positions 5 to 4, a reverse direction, and times from 10:015 to 10:017. Then, adjacent segments are checked for continuity. First, the position difference between the ending position 3 of the first segment and the starting position 3 of the second segment is calculated to be 0. Then, the position difference between the ending position 5 of the second segment and the starting position 5 of the third segment is also calculated to be 0. Records with a position difference of 0 or 1 are marked as continuous, and records with a position difference greater than 1 are marked as break candidates. Finally, the time intervals are calculated: the end time of the first segment (10:006) and the start time of the second segment (10:006) are compared. The interval of 0.009 is 0.003. The interval between the end time of the second segment (10.012) and the start time of the third segment (10.015) is 0.003. Records with a time interval of no more than 0.003 are marked as time continuous, and those with a time interval of more than 0.003 are marked as time disconnected. Then, the direction markings are compared pair by pair. If the first and second are both positive, they are marked as consistent direction. If the second and third are opposite, they are marked as direction switching. Records with consistent direction, continuous position, and continuous time are included in the normal continuation set. Records that do not meet any of the conditions are included in the abnormal set. For example, if the third segment is continuous in position and time but the direction changes, it is registered as a direction abnormality. Then, the records in the abnormal set are removed one by one. Only records that simultaneously meet the conditions of position difference no more than 1, time interval no more than 0.003, and consistent direction are retained. Finally, the retained records are reorganized in chronological order to form a sequence of connection relationships.

[0038] The segment positioning submodule calculates the corresponding positions of same-direction segments and turning segments based on the connection relationship sequence, using the following formula: ; Determine the process step to which the abnormal location belongs, adjust the segment location order, and obtain the segment location quantity. This indicates the segment positioning quantity, used to characterize the degree of positional difference between unidirectional segments and turning segments within the process. Indicates the total number of segments. Indicates the first The position of each segment in the same direction reflects its sorting position within the process node. Indicates the first The position of each turning segment reflects its corresponding location within the process node. Indicates the first The directional consistency weight of each segment is used to reflect the degree to which changes in the direction of that segment affect the overall positioning. Indicates the first The interval between segments in the process is used to characterize the distance or span between segments. This represents a smoothing factor to prevent the denominator from being zero, and is set based on the minimum sampling interval or the minimum process step size. The segment positioning quantity represents the comprehensive measurement result of the positional offset and directional difference between different segments under the same process structure. It depicts the relative distribution relationship between unidirectional and turning segments in the process link, and is used to support the subsequent identification and boundary adjustment of the process links to which abnormal positions belong.

[0039] Extracting the position sequence of the same direction segment With the turning section position sequence Extracting weighted sequences with consistent direction With process interval sequence ,right , , , The original location data was processed using min-max normalization. The minimum value is 2 and the maximum value is 8. After normalization, they are as follows: , , ,get ; Raw location data The minimum value is 3 and the maximum value is 7. After normalization, they are respectively: , , ,get ; Original weight data The minimum value is 0.4 and the maximum value is 1.0. After normalization, they are as follows: , , ,get ; raw interval data The minimum value is 1 and the maximum value is 2. After normalization, they are as follows: , , ,get ; Smoothing factor Take the normalized interval reference value After substituting the normalized data: when hour, ; when hour, ; when hour, ; Summation of the numerators is ; The summation of the denominators is: ,but: ; The determination of the segment location quantity is divided into the following intervals: when At this time, the corresponding slight offset interval refers to the positional difference between the same-direction section and the turning section, but the positional difference is concentrated within the range of adjacent process nodes. The direction change record and the process connection record remain consistent, which is a local positional offset. when At this time, corresponding to the general offset interval, this interval refers to the position difference between the same direction segment and the turning segment has expanded to multiple consecutive nodes, and some direction change records and process connection records are misaligned, which is a continuous segment offset; when At this time, the corresponding obvious offset interval refers to the interval where the position difference between the same direction section and the turning section has covered a longer process section. The direction change record and the process connection record are continuously misaligned, which is a cross-segment position offset. when At this time, the corresponding mismatch offset interval refers to the position difference between the same-direction section and the turning section that runs through the main process section. The direction change record and the process connection record are difficult to keep in correspondence, which is a position mismatch of the whole section.

[0040] result ,satisfy If the value falls within the slight offset range, it indicates that there is a deviation between the position of the same-direction section and the position of the turning section. However, the deviation is concentrated within the range of adjacent process nodes, indicating that there is an offset record in a local section of the process.

[0041] The boundary straightening submodule, based on the segment positioning quantity, checks the start and end positions of the same record against the corresponding content of the process node, compares the ownership and continuation status of adjacent records, determines whether the boundary cutting position is misplaced, straightens the ownership boundary, and obtains the working condition anomaly identification result. List the start and end position numbers, start and end times, and corresponding process node numbers for each record, and group them according to the same batch of records. For example, in a certain batch, the first record starts at position 1 and ends at position 3, corresponding to node numbers 101 to 103, with times from 10:004 to 10:006; the second record starts at position 3 and ends at position 5, corresponding to node numbers 103 to 105, with times from 10:009 to 10:012. Then, check the start and end positions of each record against the process nodes. First, compare whether the difference between the position number and the node number is in a consistent increasing relationship. For example, if the difference between positions 1 to 3 and nodes 101 to 103 is 2, it is considered a match. If the position span is 2 and the node span is 3, it is marked as a misalignment candidate. Then, check the adjacent records. The system compares the succession status between records, calculates the difference between the ending position 3 of the previous record and the starting position 3 of the next record until it reaches 0, and determines that records with a difference of 0 or 1 are consecutive. Records with a difference greater than 1 are considered to be broken. Then, it compares the direction markings and time intervals of the two records. If the directions are consistent and the time interval does not exceed 0.003, the original boundary is maintained. If the directions are consistent but the time interval exceeds 0.003, it is marked as time misalignment. If the directions are inconsistent, it is marked as a boundary split point. Then, all boundaries marked as misaligned or misplaced are corrected by adjusting their starting point to the ending point of the previous record or adjusting the ending point to the starting point of the next record, so that the position number and the node number maintain a synchronous increasing or decreasing relationship. Finally, the system outputs the set of records after the three corrections of position, time and direction, forming the result of the abnormal working condition identification.

[0042] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. An intelligent identification system for abnormal operating conditions in a chemical production process, characterized in that, The system includes: The signal synchronization module is based on chemical production equipment. It calibrates the correspondence between the measurement point wiring port and the signal channel, judges the current, voltage and pulse sampling status, unifies the controller clock and the acquisition time stamp sequence, determines the channel differences, screens off-target measurement points, and obtains a unified time stamp data sequence. The transition recognition module, based on the unified time-stamped data sequence, checks the changes in sampling direction before and after the same measurement point, identifies the locations of rising and falling transitions, counts the intervals between adjacent transitions within the sampling window, filters out continuous transition segments, and obtains the transition response position sequence. Based on the transition response position sequence, the phase construction module compares the sequence of transitions between the preceding and following process segments, distinguishes the order of the leading and lagging states, organizes the marking order of the process connection segments, determines the link continuation direction, and obtains the process phase association chain. The offset integration module, based on the process phase correlation chain, checks the direction continuity within a continuous time window, compares whether adjacent markers are in the same direction, counts the corresponding time sequence of the start and end points of continuous segments, screens the process position to which the segment belongs, and obtains the phase offset connection record. Based on the phase offset segment record, the anomaly determination module analyzes the connection relationship between the process positions before and after, compares whether the directional changes of adjacent segments are continuous, distinguishes the corresponding positions of the same-direction segments and the turning segments, and obtains the anomaly identification result.

2. The intelligent identification system for abnormal operating conditions based on chemical production processes according to claim 1, characterized in that, The unified time-stamped data sequence includes a time-aligned sequence, a sampling order sequence, and a signal consistency sequence; the turning response position sequence includes a turning position sequence, a turning direction sequence, and a turning interval sequence; the process phase association chain includes a phase order sequence, a phase association sequence, and a direction marker sequence; the phase offset segment record includes a segment interval sequence, a direction continuity sequence, and a segment affiliation sequence; and the operating condition anomaly identification result includes an abnormal segment identifier, a direction change classification, and a process node identifier.

3. The intelligent identification system for abnormal operating conditions in chemical production processes according to claim 1, characterized in that, The signal synchronization module includes: The channel-corresponding submodule is based on chemical production equipment. It analyzes the measurement point wiring port identifier and signal channel identifier, compares the corresponding order of the same batch of acquisition records, determines whether the access position and input path are misaligned, adjusts the registration order of misaligned records, filters corresponding consistent records, and obtains the port channel mapping sequence. Based on the port channel mapping sequence, the state conversion submodule obtains the single-channel signal record and wiring type, determines the compatibility relationship of current signal, voltage signal and pulse signal, adjusts the conversion order, calculates the engineering quantity conversion state, compares the continuity relationship of sampling records before and after the same source measurement point, and obtains the engineering quantity state sequence. The timing compilation submodule obtains the controller clock and acquisition time stamp based on the engineering quantity state sequence, compares the sequential relationship of the channel records in the same batch, determines the position of the time stamp offset measurement point, adjusts the timing arrangement of the offset records, filters and aligns the records, and obtains a unified time stamp data sequence.

4. The intelligent identification system for abnormal operating conditions based on chemical production processes according to claim 1, characterized in that, The transition recognition module includes: The direction discrimination submodule analyzes the direction of continuous sampling records at the same measuring point based on the unified time-stamped data sequence, compares the relationship between the change direction of the sampling records before and after the turning position, determines the sampling point to which the rising or falling turning position belongs, filters the direction switching records, and obtains the turning mark sequence. The turning interval submodule analyzes the temporal relationship between adjacent turning positions based on the turning mark sequence, compares the interval between the previous turning position and the next turning position within the continuous sampling window, calculates the turning connection order, filters out continuously occurring turning records, and obtains the turning time interval sequence. The response sorting submodule analyzes the timing of each measurement point's turning point based on the turning point time interval sequence, compares the turning point sequence of each measurement point within the same sampling window, determines the corresponding position of the process section response order, adjusts the order of the turning point records, and obtains the turning point response position sequence.

5. The intelligent identification system for abnormal operating conditions in chemical production processes according to claim 1, characterized in that, The phase construction module includes: Based on the transition response position sequence, the transition association submodule checks the transition location one by one according to the corresponding measurement points of the preceding and following process segments, compares the front and rear fit relationship of the transitions of adjacent process segments in the same sampling window, determines whether the paired records are continuously corresponding, and obtains the process segment pairing sequence. The sequence determination submodule compares the timing relationship between the transition times of adjacent process segments based on the process segment pairing sequence, determines the order of the leading state and the lagging state, adjusts the record arrangement order, and obtains the leading-lagging sequence. The link direction determination submodule checks the upstream and downstream connection order according to the leading and lagging states based on the preceding and lagging order sequence, compares whether the marking directions of adjacent process segments are consistent, determines the link connection direction, and obtains the process phase association chain.

6. The intelligent identification system for abnormal operating conditions in chemical production processes according to claim 1, characterized in that, The offset integration module includes: The continuation check submodule analyzes the arrangement order of adjacent markers within a continuous time window based on the process phase correlation chain, compares the correspondence between the direction of the previous marker and the direction of the next marker, determines whether the direction switching position falls into the same continuous segment, filters records with unchanged direction, and obtains the direction continuation sequence. The segment delimitation submodule calculates the corresponding time sequence of the first and last records of a continuous segment based on the direction continuation sequence, compares the correspondence between the interruption position and the continuation position of adjacent records, adjusts the arrangement order of the segment boundaries, determines the start and end range of the segment, and obtains the segment interval sequence. The attribution determination submodule analyzes the process position of continuous segments based on the segment interval sequence, compares the directional consistency and position connection relationship of adjacent segments, determines the segment attribution status, adjusts the position order and the correspondence with the attribution mark, and obtains the phase offset segment record.

7. The intelligent identification system for abnormal operating conditions in chemical production processes according to claim 1, characterized in that, The anomaly detection module includes: The connection verification submodule, based on the phase offset connection record, checks the sequential order of the process position, compares the registration of directional changes in adjacent sections, determines whether the connection of continuous sections is broken, filters out abnormal connection records, and obtains the connection relationship sequence. Based on the connection sequence, the segment positioning submodule calculates the corresponding positions of the same-direction segments and turning segments, determines the process link to which the abnormal position belongs, adjusts the segment positioning order, and obtains the segment positioning quantity. The boundary alignment submodule, based on the segment positioning data, verifies the start and end positions of the same record with the corresponding content of the process node, compares the ownership and continuation status of adjacent records, determines whether the boundary cutting position is misplaced, aligns the ownership boundary, and obtains the abnormal working condition identification result.

8. The intelligent identification system for abnormal operating conditions in chemical production processes according to claim 1, characterized in that, The signal channel refers to the input path corresponding to each sensor signal. The channel sequence refers to the time order in which different signals are acquired or recorded within the same sampling period. The offset measurement point refers to the measurement point signal where the sampling time is not aligned with the system clock and there is a time offset.

Images