An industrial field instrument data acquisition analysis visual monitoring system

Abstract

The present application relates to the technical field of data analysis, in particular to an industrial field instrument data acquisition and analysis visual monitoring system, in the present application, the time series data of multiple instruments in the industrial field is acquired through an instrument data acquisition module, and the working condition segment is divided according to the correlation change of multiple instruments by a working condition segment identification module; a working condition segment fingerprint construction module extracts the working condition segment fingerprint with the coordinated change relationship of multiple instruments; a coordinated processing module determines the response boundary of each instrument under the same process event and generates working condition semantic alignment time series data by a segment-level alignment correction unit relying on the working condition segment fingerprint, and then the attribution determination of the current fluctuation is performed by a transient discrimination suppression unit; finally, the working condition segment, the working condition semantic alignment time series data and the attribution determination result of the current fluctuation are associated and mapped and displayed in linkage by a final visual monitoring module, so that the consistency organization and semantic presentation of multiple instrument data under the same process event are realized.

Description

Technical Field

[0001] This invention relates to the field of data analysis technology, specifically to an industrial field instrument data acquisition, analysis, visualization, and monitoring system. Background Technology

[0002] In the field of industrial automation monitoring, data collected by field instruments often exhibits characteristics such as multi-source heterogeneity, asynchronous timing, and strong dynamic coupling. When multiple instruments participate in the response to the same process event, due to differences in equipment response characteristics, transmission link delays, and control interventions, data changes often manifest as a state of asynchronous acquisition, superposition of local disturbances, and intertwining of operating condition switching. Traditional monitoring methods typically rely on single-variable threshold alarms or simple synchronization processing based on a global time axis, making it difficult to accurately capture the collaborative evolution patterns among multiple instruments under specific process semantics. Especially when operating conditions frequently switch or short-term dynamic transition phases exist, the system is prone to misjudging normal process transition fluctuations as abnormal events, or the abnormal signals may be masked and missed due to blurred boundaries.

[0003] Furthermore, existing technologies often employ a fragmented design for data organization, anomaly detection, and visualization, lacking a unified representation benchmark between modules. This results in weak interpretability of analysis results and necessitates manual cross-interface comparisons by operators, making it difficult to support the high-reliability and high-real-time intelligent operation and maintenance requirements of industrial processes. Although some studies have attempted to introduce time series segmentation or correlation analysis, a structured process condition expression mechanism capable of simultaneously guiding data alignment and fluctuation attribution has not yet been established, nor has it effectively solved the technical problem of distinguishing between process transition behaviors and actual faults in complex dynamic scenarios. Summary of the Invention

[0004] The purpose of this invention is to provide an industrial field instrument data acquisition, analysis, visualization, and monitoring system to solve the problems mentioned in the background art. Specific technical problems include how to construct a unified operating condition representation mechanism that can synchronously constrain the segment-level alignment boundary positioning and fluctuation attribution judgment logic of multiple instrument data. This enables consistent processing and accurate classification of multiple instrument data under the same process event in complex industrial scenarios where operating condition switching and real anomalies coexist. This solves the problems of high misjudgment rate, poor interpretability, and low human-machine collaboration efficiency caused by the lack of semantic consistency in existing monitoring systems.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an industrial field instrument data acquisition, analysis, visualization, and monitoring system, comprising an instrument data acquisition module, a working condition segment identification module, a working condition segment fingerprint construction module, a collaborative processing module, and a visualization monitoring module, wherein:

[0006] The instrument data acquisition module is used to collect time-series data from multiple instruments in the industrial field, specifically including:

[0007] Each instrument is equipped with an analog signal acquisition interface, a bus communication interface, and a wireless receiving interface to receive instrument data from multiple instruments in the industrial field.

[0008] After receiving various types of instrument data, a unified timestamp is added to each data stream based on a unified clock source to achieve unified timestamp synchronization.

[0009] According to the preset data quality rules, missing segments, duplicate records, communication error records, and out-of-bounds records are identified, and abnormal records are removed from the original data stream to form standardized time-series data.

[0010] The instrument data acquisition module serves to: access raw data from multiple instruments via heterogeneous interfaces, synchronize timestamps using a unified clock source, and eliminate invalid records using preset data quality rules, thereby forming standardized time-series data. This process provides the basic input for subsequent identification of changes in the correlation between multiple instruments, helps alleviate the apparent asynchrony problem caused by differences in the acquisition links, and enables subsequent segmentation of operating conditions based on collaborative evolution feature extraction to have a more reliable time-series consistency premise.

[0011] The operating condition segment identification module is used to divide operating condition segments based on the correlated changes of multiple instruments in time-series data, specifically including:

[0012] The normalized time-series data of multiple instruments are scanned window by window using a continuous sliding time window. The dynamic evolution trend of the correlation coefficient between each instrument within each sliding time window is extracted, and the inflection point of synchronous change of multiple instruments in the time dimension is identified.

[0013] After detecting the inflection point of synchronous change, the inflection point of synchronous change is used as the candidate boundary;

[0014] By combining the dynamic evolution trend of the correlation coefficient within the sliding time window, candidate boundaries are merged and confirmed, and working condition segments with clear start and end times are formed.

[0015] The detection of synchronous change inflection points in the working condition segment recognition module includes the following steps:

[0016] The first-order difference sequence is calculated for the time-series data of multiple instruments within a sliding time window; the first-order difference sequence of each instrument is absolute valued to obtain the change intensity sequence of each instrument; the change intensity sequence is normalized to a preset interval; the normalized change intensity sequence is aligned by time, and its mean at each moment is calculated to form a joint change intensity curve of multiple instruments; local extremum detection is performed on the joint change intensity curve to screen out the peak points whose amplitude is greater than the preset joint change intensity threshold and whose slopes in the left and right neighboring regions have opposite signs; the timestamp corresponding to the peak point is taken as the synchronous change inflection point.

[0017] The effect of this working condition segment identification module is as follows: by using a sliding time window scanning method, combined with the detection of local extreme values ​​of the joint change intensity curve of multiple instruments and the dynamic evolution trend analysis of the correlation coefficient, it identifies synchronous change inflection points and divides working condition segments with clear start and end times accordingly; this method can reduce the impact of single-point mutations or occasional noise on boundary division while preserving the stage characteristics of the process, and provides time granular support for constructing a structured representation that reflects the evolution of the same process event.

[0018] The operating condition segment fingerprint construction module is used to construct corresponding operating condition segment fingerprints based on the change direction, change sequence, persistence characteristics, and correlation coupling relationships of multiple instruments within each operating condition segment. Specifically, it includes:

[0019] For each operating condition segment, the unidirectional monotonic change, bidirectional oscillatory change, and step jump in the direction of change are determined based on the first-order difference symbol sequence and local extreme value distribution of each instrument within the operating condition segment.

[0020] By extracting the moment when each instrument first reaches a changed state within the operating condition segment, the response time offset of each instrument is obtained, and topological sorting is performed based on the response time offset of each instrument to form the change order;

[0021] By statistically analyzing the slope stability range, plateau holding range, and oscillation decay trajectory of each instrument within the operating condition segment, the stability of the rate of change, the length of the plateau period, and the decay characteristics of the fluctuation amplitude are extracted to form a continuous feature.

[0022] The information dependency, time-driven relationship and state transit relationship between multiple instruments are modeled by mutual information, Granger causality and dynamic Bayesian network to form an associated coupling relationship;

[0023] The direction of change, the order of change, the continuous characteristics, and the associated coupling relationship are fused and encoded to generate a working condition segment fingerprint with multi-instrument collaborative change relationship.

[0024] The function of the operating condition segment fingerprint construction module is as follows: for each operating condition segment, it extracts the distinctive features of the coordinated changes of multiple instruments from four dimensions: change direction, change sequence, continuous characteristics, and correlation coupling relationship, and integrates and encodes them into an operating condition segment fingerprint. This operating condition segment fingerprint not only depicts the morphological rules and topological relationships of the responses of each instrument under a specific operating condition, but also serves as a structural benchmark for unified constraint alignment, boundary positioning, and fluctuation attribution determination in subsequent stages, which helps to enhance the logical coherence and semantic consistency of the system between different processing stages.

[0025] The fragment-level alignment correction unit in the collaborative processing module is used to determine the response boundaries of multiple instruments for the same process event based on the operating condition fragment fingerprint, specifically including:

[0026] A collaborative response template for multiple instruments is constructed based on the fingerprint of operating condition segments;

[0027] In the original time series data, a sliding window scan is performed on each instrument to calculate its Euclidean distance to the corresponding segment of the collaborative response template. When the Euclidean distance is less than the preset matching threshold for the first time, it is recorded as the response start boundary of the instrument. When the Euclidean distance exceeds the preset matching threshold again and remains stable, it is recorded as the response termination boundary of the instrument.

[0028] The minimum and maximum values ​​of the response start boundary and response termination boundary of multiple instruments are taken respectively to generate a unified response start boundary and a unified response termination boundary.

[0029] The fragment-level alignment correction unit performs fragment-level alignment correction on the time-series data of multiple instruments according to the response boundaries, generating operating condition semantically aligned time-series data, specifically including:

[0030] Using the unified response start boundary as the time zero point, the time series data of each instrument are shifted according to the time coordinate.

[0031] Calculate the number of target points that need to be resampled for each instrument based on the time difference between the unified response termination boundary and the unified response start boundary;

[0032] The time series data of each instrument is resampled within the corresponding response interval. After resampling, the response interval data of each instrument are spliced ​​together along the time axis to generate operating condition semantically aligned time series data.

[0033] The effect of the segment-level alignment correction unit is as follows: based on the collaborative response pattern represented by the operating condition segment fingerprint, a multi-instrument response template is constructed, and the start and end boundaries of each instrument response are located in the original time series data through Euclidean distance matching; then, coordinate translation and resampling are performed with the unified response start boundary as the time zero point to generate operating condition semantically aligned time series data; this process realizes dynamic process alignment for the same process semantic event, so that the originally scattered and asynchronous multi-instrument responses can be compared on a unified time axis, providing a semantically consistent data base for subsequent fluctuation attribution determination.

[0034] The transient discrimination and suppression unit in the collaborative processing module receives the semantically aligned time-series data of operating conditions and determines the attribution of the current fluctuation based on the fingerprint of the operating condition segment. If the current fluctuation is located within the corresponding operating condition segment and satisfies the multi-instrument collaborative change relationship represented by the operating condition segment fingerprint, the current fluctuation is determined to be a transient response; otherwise, the current fluctuation is determined to be an abnormal event. The process of determining the current fluctuation specifically includes:

[0035] The received operating condition semantically aligned time series data is scanned window by window according to a fixed-length sliding time window; within each sliding time window, the first-order difference sequence, the change intensity sequence, and the local variance within the window of each instrument time series data are calculated respectively.

[0036] The average value of the change intensity of each instrument at the same time is obtained by aligning the time values ​​and forming a joint fluctuation intensity sequence.

[0037] The joint fluctuation intensity sequence is compared with a preset fluctuation threshold. When the sequence exceeds the preset fluctuation threshold for multiple consecutive sampling times, the moment when it first exceeds the preset fluctuation threshold is determined as the fluctuation start time. When the sequence falls below the preset recovery threshold for multiple consecutive sampling times, the moment when it first falls below the preset recovery threshold is determined as the fluctuation end time. Based on the fluctuation start time and fluctuation end time, the time series data of multiple instruments within the time interval are uniformly extracted to form the current fluctuation.

[0038] The attribution determination process in the transient discrimination and suppression unit specifically includes:

[0039] Confirm whether the time interval of the current fluctuation is completely within the start and end time range of the corresponding operating condition segment; verify item by item whether the change direction of each instrument in the current fluctuation is consistent with the unidirectional monotonic change, bidirectional oscillatory change or step jump recorded in the operating condition segment fingerprint;

[0040] Verify whether the order of changes conforms to the topological relationship obtained by sorting based on response time offset;

[0041] Verify whether the sustained characteristics match the specific numerical ranges of change rate stability, plateau period length, and fluctuation amplitude decay characteristics.

[0042] Verify whether the associated coupling relationship satisfies the information dependency, time-driven relationship and state transit relationship between instruments as specified by mutual information, Granger causality and dynamic Bayesian network modeling.

[0043] The current fluctuation is determined to be a transient response only if all of the above conditions are met; if any condition is not met, the current fluctuation is determined to be an abnormal event.

[0044] The transient discrimination and suppression unit works as follows: after receiving the semantically aligned time-series data of the operating conditions, it first defines the current fluctuation range that meets the fluctuation characteristics, and then compares each item based on the change direction, change order, continuity characteristics and correlation coupling relationship recorded in the operating condition segment fingerprint; only when the time interval of the current fluctuation is within the corresponding operating condition segment and all features match, is it judged as a transient response; otherwise, it is regarded as an abnormal event. This dual constraint mechanism can reduce the possibility of misjudging normal transitions caused by operating condition switching as faults and improve the pertinence and interpretability of anomaly identification.

[0045] The visualization monitoring module is used to correlate and display the working condition segments, semantically aligned time-series data of working conditions, and the attribution determination results of current fluctuations. Specifically, it includes:

[0046] The process of association mapping and linked display specifically includes:

[0047] A working condition segment display index is established based on the start and end times, segment number, and corresponding working condition segment fingerprint;

[0048] A time-series curve display index is established based on the synchronous change relationship of multiple instruments in the time-series data aligned with the semantics of the operating conditions.

[0049] An event identifier index is established based on the current fluctuation time interval and attribution determination results;

[0050] Following a unified interface coding rule, the working condition segment display index, time series curve display index, and event identifier index are associated and mapped. The working condition segment fingerprint corresponding to the current working condition segment, the synchronous change curves of multiple instruments, and the attribution determination result corresponding to the current fluctuation are displayed synchronously on the same monitoring screen. This enables the working condition segment, working condition semantics, time series data, and the attribution determination result of the current fluctuation to be displayed in a corresponding and linked manner in the time dimension.

[0051] The visualization monitoring module achieves the following effects: it establishes a display index for operating condition segments based on the start and end times and fingerprints of the operating condition segments; it establishes a display index for time-series curves based on the semantic alignment of operating conditions with time-series data; and it establishes an event identifier index based on the time interval of the current fluctuation and its attribution determination result. Through a unified interface coding rule, it associates and maps these three elements, enabling the operating condition segments, semantic alignment of operating conditions with time-series data, and the attribution determination result of the current fluctuation to be displayed in a coordinated manner across the same monitoring screen. This design helps operators understand the evolution of process events in a single view, improving the efficiency of identifying and coordinating the handling of operating condition changes and abnormal events.

[0052] Compared with the prior art, the beneficial effects of the present invention are:

[0053] This invention constructs a unified operating condition representation mechanism centered on operating condition segment fingerprints, organically connecting the operating condition segment identification module, operating condition segment fingerprint construction module, collaborative processing module, and visualization monitoring module. This allows operating condition segment fingerprints to simultaneously apply to the segment-level alignment correction of multi-instrument data and the attribution determination process of current fluctuations. This enables consistent processing of time-series data from multiple instruments under the same process event in complex industrial scenarios where operating condition switching and real anomalies coexist. Utilizing the generated operating condition semantic-aligned time-series data, the response processes of different instruments around the same process semantic event can be organized and compared under a unified time reference. Combined with a transient discrimination and suppression unit, it constrains and determines whether the current fluctuation belongs to a transient response or an abnormal event, helping to reduce misjudgments caused by fragmented criteria in processing stages and improving the coherence and interpretability of attribution analysis. Furthermore, by associating and mapping operating condition segments, operating condition semantic-aligned time-series data, and the attribution determination results of current fluctuations, it enhances the ability of the monitoring screen to express complex process events, thereby improving the system's applicability and collaborative effectiveness in industrial field data acquisition, analysis, visualization, and monitoring. Attached Figure Description

[0054] Figure 1 This is a schematic diagram of the core process for generating working condition segment fingerprints according to the present invention;

[0055] Figure 2 This is a schematic diagram of the core process of the segment-level alignment correction unit of the present invention;

[0056] Figure 3 This is a schematic diagram of the core process of the transient discrimination and suppression unit of the present invention;

[0057] Figure 4 This is a schematic diagram of the overall module unit of the present invention.

[0058] In the diagram: 100, Instrument data acquisition module; 200, Operating condition segment recognition module; 300, Operating condition segment fingerprint construction module; 400, Collaborative processing module; 401, Segment-level alignment correction unit; 402, Transient discrimination and suppression unit; 500, Visual monitoring module. Detailed Implementation

[0059] The technical solutions in 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, and 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.

[0060] Next, the present invention provides a technical solution: an industrial field instrument data acquisition, analysis and visualization monitoring system, including an instrument data acquisition module 100, a working condition segment recognition module 200, a working condition segment fingerprint construction module 300, a collaborative processing module 400 and a visualization monitoring module 500.

[0061] The instrument data acquisition module 100 supports heterogeneous protocol access and is compatible with various transmission methods such as analog signals, digital communication buses, and wireless sensor networks. It also performs timestamp synchronization and invalid data removal preprocessing during the data access phase. The specific process is as follows:

[0062] The system is equipped with analog signal acquisition interface, bus communication interface and wireless receiving interface to receive instrument data from multiple instruments in the industrial field; these instruments are different types of field instruments in the industrial process, including but not limited to temperature instruments, pressure instruments, flow instruments, level instruments, vibration instruments, electrical parameter instruments and actuator feedback instruments.

[0063] After receiving various types of instrument data, a unified timestamp is added to each data stream based on a unified clock source to achieve unified timestamp synchronization. Subsequently, missing segments, duplicate records, communication error records, and out-of-bounds records are identified according to preset data quality rules, and abnormal records are removed from the original data stream to form standardized time-series data. This standardized time-series data is further transmitted to the operating condition segment identification module 200 for subsequent processing. Among these steps:

[0064] Preset data quality rules refer to a set of logical criteria set according to typical abnormal patterns of industrial instrument data. These include missing segment identification rules based on continuous timestamp breakpoints, duplicate record identification rules based on redundant matching of the same timestamp and numerical combination, communication error record identification rules based on abnormal protocol verification fields or frame structure mismatch, and out-of-bounds record identification rules based on joint threshold judgment of the range of each instrument and historical rate of change constraints.

[0065] The technical purpose of unified timestamp synchronization is to eliminate the apparent time difference caused by transmission delays and buffer differences in different acquisition links, and to avoid misjudging the actual response under the same process event as multiple independent events; the technical purpose of invalid data removal is to reduce the interference of communication anomalies, distorted sampling and out-of-bounds drift on fragment identification and anomaly judgment.

[0066] The threshold for determining out-of-bounds records in the preset data quality rules can be determined based on the upper and lower limits of the instrument's range. It is usually set to 0 of the instrument's calibrated range or extended to -5 to consider a safety margin. The identification of missing segments is based on the number of consecutive missing sampling points. When the number of consecutive missing sampling points is greater than 2 to 5 sampling cycles, it is determined to be a missing segment. The above thresholds are determined based on the instrument type, sampling frequency, and field communication stability.

[0067] Please see Figure 1 The operating condition segment identification module 200 employs a multivariate time-series correlation abrupt change detection method. By detecting the inflection points of synchronous changes in multiple instruments along the time dimension, and combining this with the dynamic evolution trend of correlation coefficients within a sliding time window, it divides the operating condition segments into segments with clearly defined start and end times. The specific process is as follows:

[0068] First, the time-series data from multiple instruments uploaded by the instrument data acquisition module 100 are synchronously preprocessed to form a unified sampling sequence. The formation of the unified sampling sequence is used to eliminate the comparison deviation caused by the inconsistency of the sampling periods of different instruments, and to provide a unified time reference for subsequent joint change intensity calculation and correlation coefficient calculation.

[0069] Subsequently, the time-series data of multiple instruments are scanned window by window using a continuous sliding time window. The dynamic evolution trend of the correlation coefficient between each instrument within each sliding time window is extracted, and the inflection point of synchronous change of multiple instruments in the time dimension is identified. This process allows for dynamic observation of the changes in the coupling relationship between multiple instruments through window-by-window scanning. When the operating conditions are stable, the correlation structure usually remains relatively stable. When operating condition switching, control actions, or fault propagation occur, the correlation structure will change significantly within a local time period. Therefore, the inflection point of synchronous change is used to help determine the segment boundary.

[0070] After detecting the synchronous change inflection point, the synchronous change inflection point is used as a candidate boundary. Then, combined with the dynamic evolution trend of the correlation coefficient within the sliding time window, the candidate boundaries are merged and confirmed, and finally the working condition segments with clear start and end times are formed. The working condition segments are then output to the working condition segment fingerprint construction module 300. This process uses the synchronous change inflection point and the dynamic evolution trend of the correlation coefficient in combination. The purpose is to avoid completing the segment segmentation based on a single extreme point or a single instrument mutation, and to reduce the risk of missegmentation caused by occasional noise, local spikes and single-point distortion.

[0071] The detection of synchronous change inflection points in the working condition segment recognition module 200 is accomplished through the following steps:

[0072] First, the first-order difference sequence of time-series data from multiple instruments is calculated within a sliding time window. Then, the first-order difference sequence of each instrument is absolute valued to obtain the change intensity sequence of each instrument. Next, the change intensity sequences of multiple instruments are normalized to the interval [0, 1]. Subsequently, the normalized change intensity sequences are aligned by time, and their mean at each moment is calculated to form a joint change intensity curve of multiple instruments. Then, local extremum detection is performed on the joint change intensity curve to filter out peak points with amplitudes greater than a preset threshold and opposite signs of slopes in their left and right neighboring regions. Finally, the timestamp corresponding to the peak point is taken as the synchronous change inflection point. The preferred range of the joint change intensity threshold is 0.3 to 0.6, which is set based on the mean and standard deviation of the joint change intensity sequence under historical steady-state data.

[0073] The extraction of the dynamic evolution trend of the correlation coefficient within the sliding time window in the working condition segment recognition module 200 is completed through the following steps:

[0074] A fixed-length time window is set and slides forward in unit time steps. Within each time window, subsequences of time-series data from multiple instruments are extracted. The Pearson correlation coefficient is calculated for each pair of instrument subsequences. A correlation coefficient matrix is ​​constructed from the pairwise correlation coefficients of all instruments. The average value of the elements outside the main diagonal of the correlation coefficient matrix is ​​calculated to obtain the average correlation coefficient within the time window. The average correlation coefficients corresponding to all time windows are arranged in chronological order to form a dynamic evolution trend sequence of correlation coefficients. This sequence represents the dynamic evolution trend of correlation coefficients within the sliding time window. The average correlation coefficient is used to characterize the changing trend of the overall coupling strength of multiple instruments. When the industrial process is under relatively stable operating conditions, multiple instruments often maintain a relatively stable correlation structure. When operating condition switching, control mode changes, or anomaly propagation occur, the correlation structure often undergoes abrupt changes. Therefore, this trend can be used to verify candidate boundaries. If there are obvious nonlinear coupling scenarios, the Pearson correlation coefficient can be used in conjunction with the rank correlation coefficient or mutual information. However, in this embodiment, the Pearson correlation coefficient is preferentially used for rapid online calculation to meet the real-time requirements of industrial sites.

[0075] The operating condition segment fingerprint construction module 300 is used to construct corresponding operating condition segment fingerprints based on the change direction, change sequence, persistence characteristics, and correlation coupling relationships of multiple instruments within each operating condition segment. Specifically: the change direction includes unidirectional monotonic change, bidirectional oscillatory change, and step jump; the change sequence is based on topological sorting of the response time offsets of each instrument; persistence characteristics cover the stability of the change rate, the length of the plateau period, and the attenuation characteristics of the fluctuation amplitude; the correlation coupling relationships are characterized through mutual information, Granger causality, and dynamic Bayesian network modeling. The specific process is as follows:

[0076] For each operating condition segment, the unidirectional monotonic change, bidirectional oscillatory change, and step jump in the direction of change are determined based on the first-order difference symbol sequence and local extreme value distribution of each instrument in the operating condition segment. This direction of change is used to distinguish different types of process response modes. Among them, the unidirectional monotonic change often corresponds to load increase or decrease or setpoint switching, the bidirectional oscillatory change often corresponds to the control loop adjustment process, and the step jump often corresponds to the switching action or actuator state switching.

[0077] By extracting the moment when each instrument first reaches a changing state within the operating condition segment, the response time offset of each instrument is obtained, and topological sorting is performed based on the response time offset of each instrument to form the change sequence. This change sequence is used to characterize the propagation path and sequential relationship of the same process event among different instruments. For example, the actuator acts first, the flow rate changes later, and the pressure responds later. This sequential relationship has a clear process physical meaning.

[0078] By statistically analyzing the slope stability range, plateau holding range, and oscillation decay trajectory of each instrument within the operating condition segment, the stability of the rate of change, the length of the plateau period, and the decay characteristics of the fluctuation amplitude are extracted to form a continuous feature. This continuous feature is used to reflect the complete dynamic process of the process from excitation to stability. Among them, the stability of the rate of change reflects the inertia of the equipment and the smoothness of the control, the length of the plateau period reflects the maintenance characteristics of entering the steady state, and the decay characteristics of the fluctuation amplitude reflect the damping or regulation capability of the system.

[0079] This study models the information dependencies, time-series driving relationships, and state transit relationships among multiple instruments using mutual information, Granger causality, and dynamic Bayesian networks, forming an associated coupling relationship. Mutual information is used to describe nonlinear information dependencies, suitable for complex coupled operating conditions; Granger causality is used to examine the direction of time-series driving, suitable for distinguishing between leading and response variables; and dynamic Bayesian networks are used to describe multivariate state transition relationships, suitable for expressing conditional dependencies and propagation structures in the process. The combined use of these three technologies avoids the limitations of single correlation indicators in non-stationary and nonlinear scenarios.

[0080] Subsequently, the direction of change, the order of change, the persistent features, and the associated coupling relationships are fused and encoded to generate a working condition segment fingerprint with multi-instrument collaborative change relationships. The fusion encoding encodes the direction of change as a discrete pattern identifier, the order of change as a response topology sequence, the persistent features as a normalized continuous feature vector, and the associated coupling relationships as a relation matrix or probabilistic graph structure summary, ultimately forming a unified working condition segment fingerprint representation. The technical purpose of this working condition segment fingerprint is to compress the collaborative dynamic behavior of multiple instruments into a unified and comparable structured representation to support subsequent segment alignment and fluctuation attribution determination.

[0081] The collaborative processing module 400 includes a segment-level alignment correction unit 401 and a transient discrimination and suppression unit 402. It uses the working condition segment fingerprint as a unified working condition characterization benchmark to perform collaborative processing on multi-instrument data under the same process event.

[0082] Please see Figure 2The fragment-level alignment correction unit 401 in the collaborative processing module 400 determines the response boundaries of multiple instruments for the same process event based on the operating condition fragment fingerprint, and performs fragment-level alignment correction on the time series data of multiple instruments according to the response boundaries to generate operating condition semantically aligned time series data; the specific process is as follows:

[0083] First, based on the change direction, change order, continuity characteristics and correlation coupling relationship in the operating condition segment fingerprint, the collaborative response pattern of multiple instruments for the same process event is identified. The collaborative response pattern is an abstraction of the typical response template of the same process event on multiple instruments. The purpose is to find the corresponding segment belonging to the same event in the original asynchronous response data.

[0084] Subsequently, the response start boundary and response end boundary that match the collaborative response pattern are located in the time series data of multiple instruments, forming the response boundaries of multiple instruments for the same process event; the technical purpose of boundary localization is to uniformly map the response processes that appear scattered on multiple instruments and have lag differences to the start and end range of the same process event.

[0085] Then, based on the response boundary, the time series data of multiple instruments are uniformly extracted, time base shifted, and fragment aligned and recombined to complete the fragment-level alignment correction; finally, the operating condition semantic aligned time series data is generated; the fragment-level alignment correction is not a simple time synchronization, but a response process alignment for the same process semantic event. Its essence is to allow different instruments to compare the dynamic process of the same event on a unified time axis.

[0086] The positioning of the response boundary in the fragment-level alignment correction unit 401 is accomplished through the following steps:

[0087] Based on the change direction, change order, persistence characteristics and correlation coupling relationship recorded in the working condition segment fingerprint, a collaborative response template for multiple instruments under this working condition segment is constructed;

[0088] In the original time series data before the operating condition semantic alignment, a sliding window scan is performed on the time series data of each instrument; within each sliding window, the Euclidean distance between the time series data of the instrument and the corresponding segment of the collaborative response template is calculated; the Euclidean distance is used to measure the morphological difference between the current window and the target response template. The smaller the distance, the more likely the window is to correspond to the actual response segment of the instrument to the same process event.

[0089] When the Euclidean distance is less than the preset matching threshold for the first time, the moment is recorded as the response start boundary of the instrument; continue scanning forward, and when the Euclidean distance exceeds the preset matching threshold again and remains stable, the moment is recorded as the response termination boundary of the instrument; the preset matching threshold is determined based on the 95th percentile of the matching distance distribution in historical samples of similar working conditions;

[0090] Finally, the minimum and maximum values ​​of the response start boundary and response termination boundary of multiple instruments are taken respectively to form a unified response start boundary and unified response termination boundary of multiple instruments for the same process event, which are then merged to form the response boundary. The purpose of taking the minimum start boundary and the maximum termination boundary is to cover the complete propagation and attenuation process of the same process event on each instrument, so as to avoid omitting subsequent variables with response lag.

[0091] The fragment-level alignment correction operation in the fragment-level alignment correction unit 401 is completed through the following steps:

[0092] Obtain the response boundaries of multiple instruments for the same process event, including a unified response start boundary and a unified response end boundary;

[0093] Using the unified response start boundary as the time zero point, the time series data of each instrument are shifted in time coordinates to align the response start boundaries of all instruments to time zero. The significance of unifying the time zero point is to eliminate the lag difference in the response start time of different instruments and provide a common reference system for cross-instrument dynamic process comparison.

[0094] Based on the duration difference between the unified response termination boundary and the unified response start boundary, the number of target points that need to be resampled for each instrument is calculated; a linear interpolation method is used to resample the time series data of each instrument within the corresponding response interval, so that all instruments have the same number of data points within the response interval; the purpose of resampling is to compare segments of different response durations at a uniform length, thereby supporting subsequent multivariate collaborative determination of transients and anomalies.

[0095] After resampling, the response interval data of each instrument are spliced ​​together along the time axis to form semantically aligned time-series data. In this data, the response processes of multiple instruments under the same process event are completely consistent in the time dimension. Complete consistency means that after unified boundary and translation resampling, the response intervals of each instrument to the same event can be aligned and displayed and analyzed in a unified time coordinate, which does not mean that the physical quantity values ​​of each instrument are the same.

[0096] Please see Figure 3 The transient discrimination and suppression unit 402 in the collaborative processing module 400 is used to receive the working condition semantic alignment time series data and determine the current fluctuation in the working condition semantic alignment time series data, specifically including:

[0097] First, in the received working condition semantic alignment time series data, the time series data of multiple instruments are scanned window by window according to a fixed length sliding time window; the fixed length sliding time window can be the same as the time window length in the working condition segment recognition module 200.

[0098] Subsequently, within each sliding time window, the first-order difference sequence, the change intensity sequence, and the local variance within the window of the time series data of each instrument are calculated. The change intensity of each instrument at the same moment is time-aligned and the mean is calculated to form a joint fluctuation intensity sequence. This joint fluctuation intensity sequence is used to characterize the overall strength of the common fluctuation of multiple instruments.

[0099] Next, the joint fluctuation intensity sequence is compared with the preset fluctuation threshold. When the joint fluctuation intensity sequence exceeds the preset fluctuation threshold for multiple consecutive sampling times, the moment when it first exceeds the preset fluctuation threshold is determined as the fluctuation start time. When the joint fluctuation intensity sequence is less than the preset recovery threshold for multiple consecutive sampling times, the moment when it first falls below the preset recovery threshold is determined as the fluctuation end time. The preset fluctuation threshold is determined based on the mean and standard deviation of the joint fluctuation intensity under historical healthy operating conditions.

[0100] Finally, based on the start and end times of the fluctuation, the time series data of multiple instruments within the time interval are uniformly extracted to generate the current fluctuation; this current fluctuation is the basic analysis object for subsequent attribution determination, and its essence is a candidate event fragment.

[0101] The transient discrimination and suppression unit 402 determines the attribution of current fluctuations based on the operating condition segment fingerprint, specifically including:

[0102] Extract the time interval and multi-instrument behavior features of the current fluctuation that have been determined, and strictly compare them with the working condition segment fingerprint of the corresponding working condition segment;

[0103] First, confirm whether the time interval of the current fluctuation is completely within the start and end time range of the operating condition segment; second, verify item by item whether the change direction of each instrument in the current fluctuation is consistent with the unidirectional monotonic change, bidirectional oscillatory change, or step jump recorded in the operating condition segment fingerprint, whether the change sequence conforms to the topological relationship obtained by sorting based on the response time offset, whether the persistence characteristics match the specific numerical range of change rate stability, plateau period length, and fluctuation amplitude decay characteristics, and whether the correlation and coupling relationship satisfies the information dependency relationship, time-series driving relationship, and state transmission relationship between instruments as specified by mutual information, Granger causality, and dynamic Bayesian network modeling; only when all the above conditions are met, the current fluctuation is judged as a transient response; if any condition is not met, the current fluctuation is judged as an abnormal event.

[0104] The visualization monitoring module 500 receives operating condition segments, operating condition semantic alignment time-series data, and the attribution determination result of the current fluctuation, and generates corresponding monitoring screens according to a unified interface coding rule. The monitoring screen includes at least an operating condition segment display area, an operating condition semantic alignment time-series data display area, and an attribution determination result display area. The operating condition segment display area shows the start and end times, segment number, and corresponding operating condition segment fingerprint of the current operating condition segment. The operating condition semantic alignment time-series data display area shows the synchronous change curves of multiple instruments in the operating condition semantic alignment time-series data and the time interval corresponding to the current fluctuation. The attribution determination result display area shows the determination result of whether the current fluctuation is judged as a transient response or an abnormal event, thereby outputting a monitoring screen corresponding to the operating condition segment, operating condition semantic alignment time-series data, and the attribution determination result of the current fluctuation. This process, by displaying the segment boundaries, aligned multi-instrument curves, and determination results in a unified interface, allows operators to observe the difference between the operating condition switching process and abnormal events in the same time dimension.

[0105] The visualization monitoring module 500 is used not only to display operating condition segments, semantically aligned time-series data of operating conditions, and the attribution determination results of current fluctuations, but also to uniformly organize, correlate, map, and link data results from the above different sources, thereby establishing a corresponding relationship between operating condition segment information under the same process event, the synchronous change process of multiple instruments, and the attribution determination results of current fluctuations in the same time dimension; specifically including:

[0106] The visualization monitoring module 500 receives the working condition segments output by the working condition segment recognition module 200, the working condition semantic alignment timing data output by the segment-level alignment correction unit 401, and the attribution determination result of the current fluctuation output by the transient discrimination and suppression unit 402.

[0107] Subsequently, following the unified interface coding rules, the start and end times, segment numbers, and corresponding operating condition segment fingerprints of the operating condition segments are extracted to establish an operating condition segment display index; the time axis positions and synchronous change relationships of multiple instruments in the operating condition semantic alignment time series data are extracted to establish a time series curve display index; and the time interval of the current fluctuation and its determination result as a transient response or an abnormal event are extracted to establish an event identifier index. The unified interface coding rules refer to aligning the operating condition segments, operating condition semantic alignment time series data, and the attribution determination result of the current fluctuation along the time axis, and constructing a two-dimensional visualization mapping table with the operating condition segment number as the top-level index, the timestamp as the horizontal coordinate, and the instrument identifier as the vertical dimension. The operating condition segment display area, the operating condition semantic alignment time series data display area, and the attribution determination result display area share the same time base, and the association, positioning, and linkage refresh of the three types of information are achieved through color identification, border highlighting, and layer overlay.

[0108] After the above indexes are established, the operating condition segment display index, time series curve display index, and event identifier index are associated and mapped according to a unified time base, so that the operating condition segments, operating condition semantic alignment time series data, and the attribution determination results of the current fluctuation form a one-to-one linkage relationship within the same monitoring screen. Among them, the operating condition segment display area is used to display the start and end time, segment number, and corresponding operating condition segment fingerprint of the current operating condition segment; the operating condition semantic alignment time series data display area is used to display the synchronous change curves of multiple instruments in the operating condition semantic alignment time series data, and to locate and identify the time interval corresponding to the current fluctuation; the attribution determination result display area is used to display the determination result of whether the current fluctuation is determined to be a transient response or an abnormal event.

[0109] Furthermore, when the monitored object switches to any operating condition segment or the current fluctuation, based on the established association mapping relationship, the corresponding operating condition semantic alignment time series data and the attribution determination result of the current fluctuation are synchronously called for linked display. This allows operators to simultaneously observe the operating condition segment boundary, the synchronous change curve of multiple instruments, and the attribution determination result of the current fluctuation from the same time dimension, thereby improving the efficiency of distinguishing between the operating condition switching process and abnormal events. The technical purpose of this linked display is to reduce the cognitive burden of operators switching and comparing between multiple interfaces and improve the efficiency of identifying and handling complex process events.

[0110] As can be seen from the above description, the industrial field instrument data acquisition, analysis, visualization, and monitoring system provided in this embodiment has the following technical effects:

[0111] Please see Figure 4 The system acquires multi-instrument time-series data through the instrument data acquisition module 100, and divides the operating condition into segments by the operating condition segment identification module 200. The operating condition segment fingerprint construction module 300 extracts the change direction, change order, continuity characteristics, and correlation coupling relationships of multiple instruments within each operating condition segment, generating a unified operating condition segment fingerprint. The collaborative processing module 400, relying on this operating condition segment fingerprint, enables the segment-level alignment correction unit 401 to locate the multi-instrument response boundaries and generate operating condition semantic alignment time-series data. Simultaneously, it guides the transient discrimination and suppression unit 402 to determine the attribution of the current fluctuation. Only when the fluctuation is located within the corresponding operating condition segment and satisfies the multi-instrument collaborative change relationship represented by the operating condition segment fingerprint is it determined to be a transient response; otherwise, it is determined to be an abnormal event. Finally, the visualization monitoring module 500 correlates and displays the operating condition segments, operating condition semantic alignment time-series data, and the attribution determination results of the current fluctuation, achieving semantically consistent organization and collaborative presentation of multi-instrument data in the time dimension under the same process event. This helps improve the consistency and interpretability of system analysis in complex operating condition switching and abnormal coexistence scenarios.

[0112] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. An industrial field instrument data acquisition, analysis, visualization, and monitoring system, characterized in that, It includes an instrument data acquisition module, a working condition segment recognition module, a working condition segment fingerprint construction module, a collaborative processing module, and a visualization monitoring module, among which: The instrument data acquisition module is used to collect time-series data from multiple instruments in the industrial field; The operating condition segment identification module is used to divide the operating condition segments according to the correlation changes of multiple instruments in the time series data; The operating condition segment fingerprint construction module is used to construct the corresponding operating condition segment fingerprint based on the change direction, change order, continuous characteristics and associated coupling relationship of multiple instruments in each operating condition segment; The collaborative processing module includes a fragment-level alignment correction unit and a transient discrimination and suppression unit, wherein: The segment-level alignment correction unit is used to determine the response boundaries of multiple instruments for the same process event based on the operating condition segment fingerprint, and to perform segment-level alignment correction on the time series data of multiple instruments according to the response boundaries to generate operating condition semantic alignment time series data. The transient discrimination and suppression unit is used to receive the working condition semantic alignment time series data, and to determine the attribution of the current fluctuation in the working condition segment based on the working condition segment fingerprint. When the current fluctuation is located in the corresponding working condition segment and satisfies the multi-instrument collaborative change relationship represented by the working condition segment fingerprint, the current fluctuation is determined to be a transient response; otherwise, the current fluctuation is determined to be an abnormal event. The visualization monitoring module is used to perform correlation mapping and linkage display of the working condition segment, the working condition semantic alignment time series data, and the attribution determination result of the current fluctuation.

2. The industrial field instrument data acquisition, analysis, visualization, and monitoring system according to claim 1, characterized in that, The time-series data acquisition process in the instrument data acquisition module includes: Each instrument is equipped with an analog signal acquisition interface, a bus communication interface, and a wireless receiving interface to receive instrument data from multiple instruments in the industrial field. After receiving various types of instrument data, a unified timestamp is added to each data stream based on a unified clock source to achieve unified timestamp synchronization. According to the preset data quality rules, missing segments, duplicate records, communication error records, and out-of-bounds records are identified, and abnormal records are removed from the original data stream to form standardized time-series data.

3. The industrial field instrument data acquisition, analysis, visualization, and monitoring system according to claim 2, characterized in that, The process of dividing the operating condition segments specifically includes: The normalized time-series data of multiple instruments are scanned window by window using a continuous sliding time window. The dynamic evolution trend of the correlation coefficient between each instrument within each sliding time window is extracted, and the inflection point of synchronous change of multiple instruments in the time dimension is identified. After detecting the synchronous change inflection point, the synchronous change inflection point is used as a candidate boundary; By combining the dynamic evolution trend of the correlation coefficient within the sliding time window, the candidate boundaries are merged and confirmed, and the working condition segments with clear start and end times are formed.

4. The industrial field instrument data acquisition, analysis, visualization, and monitoring system according to claim 3, characterized in that, The detection process for the synchronous change inflection point specifically includes: The first-order difference sequence is calculated for the time-series data of multiple instruments within a sliding time window; the first-order difference sequence of each instrument is processed by absolute value to obtain the change intensity sequence of each instrument; the change intensity sequence is normalized to a preset interval; the normalized change intensity sequence is aligned by time, and its mean at each moment is calculated to form a joint change intensity curve of multiple instruments; local extremum detection is performed on the joint change intensity curve to screen out the peak points whose amplitude is greater than the preset joint change intensity threshold and whose slopes in the left and right neighboring regions have opposite signs; the timestamp corresponding to the peak point is taken as the synchronous change inflection point.

5. The industrial field instrument data acquisition, analysis, visualization, and monitoring system according to claim 1, characterized in that, The process of constructing the working condition segment fingerprint specifically includes: For each of the aforementioned operating conditions, the unidirectional monotonic change, bidirectional oscillatory change, and step jump in the direction of change are determined based on the first-order difference symbol sequence and local extreme value distribution of each instrument within the operating condition segment. By extracting the moment when each instrument first reaches a changed state within the operating condition segment, the response time offset of each instrument is obtained, and a topological sort is performed based on the response time offset of each instrument to form the change order; By statistically analyzing the slope stability range, plateau holding range, and oscillation decay trajectory of each instrument within the operating condition segment, the stability of the rate of change, the length of the plateau period, and the decay characteristics of the fluctuation amplitude are extracted to form a continuous feature. The information dependency, time-driven relationship and state transit relationship between multiple instruments are modeled by mutual information, Granger causality and dynamic Bayesian network to form an associated coupling relationship; The change direction, change order, continuous features, and associated coupling relationship are fused and encoded to generate a working condition segment fingerprint with multi-instrument collaborative change relationship.

6. The industrial field instrument data acquisition, analysis, visualization, and monitoring system according to claim 1, characterized in that, The process of locating the response boundary in the fragment-level alignment correction unit specifically includes: A collaborative response template for multiple instruments is constructed based on the fingerprint of operating condition segments; In the original time series data, a sliding window scan is performed on each instrument to calculate its Euclidean distance to the corresponding segment of the collaborative response template. When the Euclidean distance is less than the preset matching threshold for the first time, it is recorded as the response start boundary of the instrument. When the Euclidean distance exceeds the preset matching threshold again and remains stable, it is recorded as the response termination boundary of the instrument. The minimum and maximum values ​​of the response start boundary and response termination boundary of multiple instruments are taken respectively to generate a unified response start boundary and a unified response termination boundary.

7. The industrial field instrument data acquisition, analysis, visualization, and monitoring system according to claim 6, characterized in that, The process of generating the working condition semantic alignment time series data specifically includes: Using the unified response start boundary as the time zero point, the time series data of each instrument are shifted according to the time coordinate. Based on the time difference between the unified response termination boundary and the unified response start boundary, calculate the number of target points that each instrument needs to resample; The time series data of each instrument is resampled within the corresponding response interval. After resampling, the response interval data of each instrument are spliced ​​together along the time axis to generate operating condition semantically aligned time series data.

8. The industrial field instrument data acquisition, analysis, visualization, and monitoring system according to claim 1, characterized in that, The process of determining the current fluctuation specifically includes: The received operating condition semantically aligned time series data is scanned window by window according to a fixed-length sliding time window; within each sliding time window, the first-order difference sequence, the change intensity sequence, and the local variance within the window of each instrument time series data are calculated respectively. The average value of the change intensity of each instrument at the same time is obtained by aligning the time values ​​and forming a joint fluctuation intensity sequence. The joint fluctuation intensity sequence is compared with a preset fluctuation threshold. When the sequence exceeds the preset fluctuation threshold for multiple consecutive sampling times, the moment when it first exceeds the preset fluctuation threshold is determined as the fluctuation start time. When the sequence falls below the preset recovery threshold for multiple consecutive sampling times, the moment when it first falls below the preset recovery threshold is determined as the fluctuation end time. Based on the fluctuation start time and fluctuation end time, the time series data of multiple instruments within the time interval are uniformly extracted to form the current fluctuation.

9. The industrial field instrument data acquisition, analysis, visualization, and monitoring system according to claim 1, characterized in that, The attribution determination process specifically includes: Confirm whether the time interval of the current fluctuation is completely within the start and end time range of the corresponding operating condition segment; verify item by item whether the change direction of each instrument in the current fluctuation is consistent with the unidirectional monotonic change, bidirectional oscillatory change or step jump recorded in the operating condition segment fingerprint; Verify whether the order of changes conforms to the topological relationship obtained by sorting based on response time offset; Verify whether the sustained characteristics match the specific numerical ranges of change rate stability, plateau period length, and fluctuation amplitude decay characteristics. Verify whether the associated coupling relationship satisfies the information dependency, time-driven relationship and state transit relationship between instruments as specified by mutual information, Granger causality and dynamic Bayesian network modeling. The current fluctuation is determined to be a transient response only if all of the above conditions are met; if any condition is not met, the current fluctuation is determined to be an abnormal event.

10. The industrial field instrument data acquisition, analysis, visualization, and monitoring system according to claim 1, characterized in that, The process of association mapping and linked display specifically includes: A working condition segment display index is established based on the start and end times, segment number, and corresponding working condition segment fingerprint; A time-series curve display index is established based on the synchronous change relationship of multiple instruments in the time-series data aligned with the semantics of the operating conditions. An event identifier index is established based on the current fluctuation time interval and attribution determination results; According to the unified interface coding rules, the working condition segment display index, the time series curve display index, and the event identifier index are associated and mapped. The working condition segment fingerprint corresponding to the current working condition segment, the synchronous change curves of multiple instruments, and the attribution determination result corresponding to the current fluctuation are displayed synchronously on the same monitoring screen, so that the working condition segment, working condition semantic alignment time series data, and the attribution determination result of the current fluctuation are displayed in a corresponding linkage in the time dimension.

Images