Petroleum resin production line equipment predictive maintenance system based on digital twinning

Abstract

The present application relates to the field of data processing, and particularly relates to a predictive maintenance system for petroleum resin production line equipment based on digital twinning. The system comprises: a data acquisition module, which is used for acquiring historical data of a to-be-analyzed device in a petroleum resin production line; a data processing module, which is used for clustering the historical data, determining an abnormal period of each monitoring index of the to-be-analyzed device and a preliminary screening abnormal period of the to-be-analyzed device according to a working condition cluster; determining a target abnormal period, acquiring a downstream device, and acquiring the preliminary screening abnormal period; acquiring an abnormal similarity; calculating a real abnormality possibility; determining a target period, calculating a process expansion index, and calculating a deterioration degree index; determining a warning priority index; and an information warning module, which is used for determining warning information at each moment. The present application can improve the abnormality detection accuracy of the device.

Description

Technical Field

[0001] This invention relates to the field of data processing, specifically to a predictive maintenance system for petroleum resin production line equipment based on digital twins. Background Technology

[0002] The production of petroleum resin is usually a continuous process, and it is typically carried out under high temperature and high pressure conditions. When one of the equipment malfunctions, it may cause a loss of output for the entire production line, or even lead to production interruption or abnormal product quality.

[0003] Therefore, it is necessary to monitor the operating status of equipment on the production line in real time during the petroleum resin production process in order to perform predictive maintenance. Existing methods are usually based on data twin technology, which uses static modeling to predict equipment anomalies, or uses LSTM models combined with monitoring data from each piece of equipment to predict equipment anomalies and determine maintenance needs.

[0004] However, in real-world industrial scenarios, there are often numerous random interference factors, making existing methods inaccurate in predicting abnormal equipment states in practical industrial applications, leading to missed or false alarms. For example, existing methods can only obtain the correlation between various monitoring data. When there are temporary interferences (such as load switching, voltage fluctuations, etc., which cause sudden anomalies in monitoring data, but the actual equipment is still in a healthy state), existing methods may misjudge them as equipment anomalies, resulting in frequent false alarms and inaccurate predictions. Summary of the Invention

[0005] This invention provides a predictive maintenance system for petroleum resin production line equipment based on digital twins to address existing problems.

[0006] The predictive maintenance system for petroleum resin production line equipment based on digital twins of the present invention adopts the following technical solution: One embodiment of the present invention provides a predictive maintenance system for petroleum resin production line equipment based on digital twins, the system comprising: The data acquisition module is used to acquire historical data of the equipment to be analyzed in the petroleum resin production line. The data processing module is used to cluster historical data using the DBSCAN algorithm to obtain operating condition clusters, and based on the operating condition clusters, to determine the abnormal periods of each monitoring indicator of the equipment to be analyzed and the initial screening abnormal periods of the equipment to be analyzed. From the initial screening abnormal period of the device to be analyzed, determine the target abnormal period, obtain the downstream devices of the device to be analyzed, and obtain each initial screening abnormal period of each downstream device. The target abnormal period is the initial screening abnormal period in which the target abnormal time is located. Obtain the anomaly similarity between the target anomaly time period and each initial screening anomaly time period of each downstream device; The true probability of an anomaly at the target anomaly time is calculated based on the anomaly similarity between the target anomaly time period and each initial screening anomaly time period of each downstream device. From the initial screening abnormal period of the equipment to be analyzed, the target period is determined. Based on the target period, the process expansion index at the target abnormal time is calculated, and based on the process expansion index at the target abnormal time, the deterioration index at the target abnormal time is calculated. The product of the actual probability of an anomaly at the target anomaly moment and the deterioration index at the target anomaly moment is determined as the early warning priority index for the target anomaly moment. The information early warning module is used to identify each moment in the initial screening abnormal period of the device to be analyzed as the target abnormal moment, obtain the early warning priority index of each moment, and determine the early warning information for each moment based on the early warning priority index of each moment.

[0007] Optionally, in the data processing module, the DBSCAN algorithm is used to cluster historical data to obtain operating condition clusters, specifically including: The raw material input, flow rate difference of the equipment under analysis, liquid level difference of the equipment under analysis, and environmental data at each moment in the historical data are respectively defined as the multidimensional sample space of the DBSCAN algorithm. The mean distance between data samples within a preset time interval at each moment is defined as the neighborhood radius of the DBSCAN algorithm. The preset number of samples is defined as the minimum number of samples for the DBSCAN algorithm. The DBSCAN algorithm is used to cluster the historical data to obtain the working condition cluster. Among them, the flow rate difference of the device under analysis at each moment is the difference between the output flow rate data and the input flow rate data of the device under analysis at each moment, and the liquid level difference of the device under analysis at each moment is the difference between the output liquid level data and the input liquid level data of the device under analysis at each moment.

[0008] Optionally, in the data processing module, based on the operating condition clusters, the abnormal time periods for each monitoring indicator of the equipment to be analyzed and the initial screening abnormal time periods for the equipment to be analyzed are determined, specifically including: Obtain the monitoring value of the monitoring index k of the equipment to be analyzed at time t, and determine the working condition cluster at time t as the working condition cluster to be analyzed; Calculate the monitoring index k of the equipment to be analyzed, and the average difference between the monitoring value of the equipment at each time step in the cluster of working conditions to be analyzed and the monitoring index k of the equipment at time t. Obtain the monitoring index k of the device to be analyzed, the monitoring value of each time within the preset time interval corresponding to time t, and sort them in chronological order to obtain the time series sequence of monitoring index k; The first-order difference of the time series is obtained to obtain the difference series of the monitoring index k; The ratio of the difference value at time t in the difference sequence of monitoring indicator k to the mean of the elements in the difference sequence of monitoring indicator k is determined as the fluctuation intensity of monitoring indicator k. The product of the mean difference of monitoring indicator k and the fluctuation intensity of monitoring indicator k is determined as the degree of abnormality of monitoring indicator k at time t. Obtain the degree of anomaly of monitoring indicator k at each time point; The moment when the degree of abnormality is greater than the preset threshold for abnormality is determined as the abnormal moment of monitoring indicator k. The abnormal moment of each monitoring indicator is obtained and the abnormal moment is determined as the initial screening abnormal moment of the device to be analyzed. The abnormal times of monitoring indicator k are merged to obtain the abnormal time periods of monitoring indicator k, and the abnormal time periods of each monitoring indicator are obtained. The initial screening abnormal times of the equipment to be analyzed are merged to obtain the initial screening abnormal time period of the equipment to be analyzed.

[0009] Optionally, in the data processing module, the anomaly similarity between the target anomaly time period and each initial screening anomaly time period of each downstream device is obtained, specifically including: Obtain the initial screening anomaly period j of downstream device z; The degree of abnormality of the traffic difference at each moment during the target abnormal period is obtained to obtain the first abnormal traffic sequence of the device to be analyzed; The degree of abnormality of the flow difference at each moment in the initial screening abnormal period j of downstream device z is obtained, and the second flow abnormality sequence of the initial screening abnormal period j of downstream device z is obtained. The degree of abnormality of the liquid level difference at each moment during the target abnormal period is obtained to obtain the first liquid level abnormality sequence of the equipment to be analyzed; The degree of abnormality of the liquid level difference at each moment in the initial screening abnormal period j of the downstream equipment z is obtained, and the second liquid level abnormality sequence of the initial screening abnormal period j of the downstream equipment z is obtained. The DTW algorithm is used to calculate the first distance between the first and second traffic anomaly sequences. The second distance between the first and second liquid level anomaly sequences is calculated using the DTW algorithm. The average of the first and second distances is determined as the abnormal similarity between the target abnormal time period and the initial screening abnormal time period j of the downstream device z. Obtain the anomaly similarity between the target anomaly time period and each initial screening anomaly time period of the downstream device z; Obtain the anomaly similarity between the target anomaly time period and each initial screening anomaly time period of each downstream device.

[0010] Optionally, in the data processing module, the true probability of an anomaly at the target anomaly time is calculated based on the anomaly similarity between the target anomaly time period and each initial screening anomaly time period of each downstream device. Specifically, this includes: Based on the anomaly similarity between the target anomaly time period and each initial screening anomaly time period of each downstream device, calculate the probability of anomaly response between the target anomaly time period and each initial screening anomaly time period of each downstream device. The initial screening abnormal period of downstream devices with an abnormal response probability greater than the preset abnormal response probability threshold is determined as the response abnormal period, and the downstream devices corresponding to the response abnormal period are determined as response devices. The sum of the Pearson correlation coefficients of the data sequence of the first monitoring indicator and each of the second monitoring indicators during the target abnormal period is calculated to obtain the data correlation of the first monitoring indicator. The first monitoring indicator is any monitoring indicator during the target abnormal period, and the second monitoring indicator is any monitoring indicator other than the first monitoring indicator during the target abnormal period. During the target abnormal period, each monitoring indicator is determined as the first monitoring indicator. The data correlation of each monitoring indicator is obtained and summed to obtain the data correlation degree of the target abnormal period. Remove the abnormal periods from the historical data of the equipment to be analyzed to obtain the normal periods; The time period within a preset duration interval of the target abnormal time within the normal time period is defined as the normal time period of the target abnormal time. The data correlation of the third monitoring indicator is obtained by summing the Pearson correlation coefficients of the data sequence of the third monitoring indicator and each of the fourth monitoring indicators during the normal period of the target abnormal time. The third monitoring indicator is any monitoring indicator during the normal period of the target abnormal time, and the fourth monitoring indicator is any monitoring indicator other than the third monitoring indicator during the normal period of the target abnormal time. During the normal period of the target abnormal moment, each monitoring indicator is determined as the third monitoring indicator. The data correlation of each monitoring indicator is obtained and summed to obtain the degree of correlation of the normal data at the target abnormal moment. The true probability of an anomaly at the target anomaly time is calculated based on the number of anomaly monitoring indicators at the target anomaly time, the number of response devices at the target anomaly time, the number of downstream devices at the target anomaly time, the correlation of data during the target anomaly period, and the correlation of normal data at the target anomaly time.

[0011] Optionally, in the data processing module, based on the anomaly similarity between the target anomaly time period and each initial screening anomaly time period of each downstream device, the probability of anomaly response between the target anomaly time period and each initial screening anomaly time period of each downstream device is calculated, specifically including: Obtain the start time of the initial screening anomaly period j of downstream device z; Calculate the start time of the initial screening anomaly period j of downstream device z and the target time length between the target anomaly period and the target anomaly time corresponding to the target anomaly period; Obtain the standard material handling time for the device to be analyzed and the downstream device z; The ratio of the standard material transport time to the target time length of the equipment to be analyzed and the downstream equipment z is determined as the standard material transport rate of the initial screening abnormal period j of the downstream equipment z. The reciprocal of the abnormality similarity between the target abnormal time period and the initial screening abnormal time period j of the downstream equipment z is multiplied by the standard material conveying rate of the initial screening abnormal time period j of the downstream equipment z to obtain the abnormal response probability of the target abnormal time period and the initial screening abnormal time period j of the downstream equipment z. Obtain the probability of abnormal response for each initial screening abnormal period of the target abnormal period and the downstream device z; Obtain the probability of abnormal response for the target abnormal period and each initial screening abnormal period for each downstream device.

[0012] Optionally, in the data processing module, the true probability of an anomaly at the target anomaly time is calculated based on the number of anomaly monitoring indicators at the target anomaly time, the number of response devices at the target anomaly time, the number of downstream devices at the target anomaly time, the correlation of data during the target anomaly period, and the correlation of normal data at the target anomaly time. Specifically, this includes: The ratio of the correlation between the data during the target abnormal period and the correlation between the normal data at the target abnormal moment is determined as the internal correlation at the target abnormal moment. The adjustment value is obtained by adding the number of downstream devices at the time of the target anomaly to a preset value divided by zero. The ratio of the number of responding devices to the adjustment value at the time of the target anomaly is determined as the downstream device ratio at the time of the target anomaly. The true probability of an anomaly at the target anomaly moment is determined by multiplying the number of anomaly monitoring indicators, their internal correlation, and the proportion of downstream equipment.

[0013] Optionally, in the data processing module, a target time period is determined from the initial screening of abnormal periods of the devices to be analyzed, specifically including: The cluster of operating conditions at the time of the target anomaly is determined as the target operating condition cluster; Within the target operating condition cluster, other times are identified as historically similar states. By merging the anomalous moments in similar historical states, the target time period is obtained.

[0014] Optionally, in the data processing module, based on the target time period, the process amplification index at the target abnormal moment is calculated, specifically including: Obtain the response devices for the target time period, and obtain the distance of the response device that is farthest from the device to be analyzed during the target time period to obtain the first distance; Obtain the response devices during the target abnormal period, and obtain the distance of the response device that is farthest from the device to be analyzed during the target abnormal period to obtain the second distance; Obtain the abnormal response period of response device a corresponding to the target abnormal period, obtain the abnormal response probability of response device a for each abnormal response period and calculate the average value to obtain the average abnormal response probability of response device a corresponding to the target abnormal period. Obtain the average probability of abnormal response for each responding device during the target abnormal period and calculate the average to obtain the average abnormal response during the target abnormal period; The ratio of the second distance to the first distance is determined as the extent of the target's impact at the moment of an anomaly. The product of the impact breadth at the target anomaly moment and the average anomaly response during the target anomaly period is determined as the process amplification index at the target anomaly moment.

[0015] Optionally, in the data processing module, based on the process amplification index at the target anomaly time, a deterioration index at the target anomaly time is calculated, specifically including: The duration of the target abnormal period, the average duration of the target period, and the growth period in the initial screening abnormal period of the device to be analyzed are obtained. In each growth period, the abnormality of the monitoring indicator at time b is greater than the abnormality of the monitoring indicator at time b-1. The difference between the degree of abnormality of monitoring indicator c at time d and the degree of abnormality of monitoring indicator c at time d-1 is determined as the abnormal growth amount of monitoring indicator c at time d. Obtain the abnormal growth amount of monitoring indicator c at each time point; Obtain the average abnormal growth of each monitoring indicator at each time point to obtain the average abnormal growth. The time duration is determined by the ratio of the duration of the target abnormal period to the average duration of the target period. The deterioration index of the target abnormal moment is determined by multiplying the duration of time, the number of moments in the growth period, the average abnormal growth, and the process expansion index of the target abnormal moment.

[0016] The beneficial effects of the technical solution of the present invention are: In this embodiment of the invention, abnormal moments of each monitoring indicator are initially screened by combining the fluctuations of monitoring data under similar working conditions. Based on digital twin technology, the impact of an abnormality of a single indicator on upstream and downstream processes and the associated abnormal responses of other monitoring indicators are analyzed, thereby distinguishing between temporary interference and real equipment abnormalities. The warning priority is determined by combining the development trend of real abnormalities and the impact of corresponding equipment on upstream and downstream processes, so that staff can take corresponding maintenance measures. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a structural diagram of a predictive maintenance system for petroleum resin production line equipment based on digital twins, provided in one embodiment of the present invention. Detailed Implementation

[0019] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of the predictive maintenance system for petroleum resin production line equipment based on digital twins proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0021] The following description, in conjunction with the accompanying drawings, details the specific solution of the predictive maintenance system for petroleum resin production line equipment based on digital twins provided by this invention.

[0022] This invention provides a predictive maintenance system for petroleum resin production line equipment based on digital twins. Please refer to [link to relevant documentation]. Figure 1 The diagram illustrates the structure of a predictive maintenance system for petroleum resin production line equipment based on digital twins, according to an embodiment of the present invention. The system includes the following modules: The data acquisition module 101 is used to acquire historical data of the equipment to be analyzed in the petroleum resin production line.

[0023] For example, a digital twin is a dynamic, evolvable virtual model. It evolves synchronously with the physical entity by receiving data from the physical equipment in real time (such as raw material input, flow rate, temperature, etc.). Therefore, a digital twin model of a petroleum resin production line can be constructed using digital twin technology: The specific structure of each piece of equipment on the petroleum resin production line is obtained from CAD drawings or simple 3D scans. This structure is then transmitted to edge computing nodes, where the equipment is connected according to the actual production process. This creates a lightweight digital twin model of the production line based on spatial and functional relationships (existing technology). The model can be built not only at edge computing nodes but also at other locations.

[0024] A knowledge graph of petroleum resin production processes is constructed based on a digital twin model: nodes are equipment and sensors, nodes related to the process are connected by edges, and the functional associations of nodes connected by a certain edge are defined, such as the effect of vacuum pump performance on distillation column pressure, etc.

[0025] In this embodiment, the equipment in the digital twin model may include raw material storage tanks, feed pumps, stirring equipment, polymerization reactors, vacuum pumps, distillation columns, and finished product tanks, etc. In this embodiment, the polymerization reactor is used as the device to be analyzed, and historical data from the polymerization reactor is analyzed.

[0026] After the digital twin model is completed, the status of the equipment on the current petroleum resin production line is monitored in real time, including: For polymerization reactors: The status of the stirring equipment in the polymerization reactor is monitored using temperature sensors, current sensors (smart ammeters), vibration sensors, etc. For each piece of equipment: The input and output terminals of the corresponding production processes of all core equipment on the production line are obtained in real time using flow sensors, level sensors, etc.

[0027] Regarding environmental impact factors: the environment of each production workshop is monitored, including temperature and humidity.

[0028] In this embodiment, each type of data collected from the polymerization reactor can be considered a monitoring indicator, such as monitoring indicator 1 and monitoring indicator 2. The timestamps of each sensor are aligned (data is acquired once per minute), and the acquired data is transmitted to a nearby edge computing node. At the edge computing node, the acquired data undergoes preprocessing such as cleaning and alignment, and the processed data is then stored and analyzed. For the polymerization reactor, each type of data acquired by each sensor (temperature, current, vibration, etc.) constitutes a monitoring dimension, or monitoring indicator, for the stirring equipment of the polymerization reactor. Furthermore, the flow monitoring indicator for each device is the difference between the output and input flow rates at a given moment in its production process. Similarly, for liquid level monitoring indicators, the difference between the output and input liquid levels is obtained. .

[0029] Optionally, the length of the historical data obtained in this embodiment can be one week, which can be selected according to actual needs, but it cannot be less than the preset duration interval of each moment in this embodiment. The selected historical data is preferably the most recent reaction data, which can be obtained according to actual needs without specific restrictions.

[0030] The data processing module 102 is used to cluster historical data using the DBSCAN algorithm to obtain working condition clusters, and based on the working condition clusters, to determine the abnormal time periods for each monitoring indicator of the equipment to be analyzed and the initial screening abnormal time periods for the equipment to be analyzed. From the initial screening abnormal period of the device to be analyzed, determine the target abnormal period, obtain the downstream devices of the device to be analyzed, and obtain each initial screening abnormal period of each downstream device. The target abnormal period is the initial screening abnormal period in which the target abnormal time is located. Obtain the anomaly similarity between the target anomaly time period and each initial screening anomaly time period of each downstream device; The true probability of an anomaly at the target anomaly time is calculated based on the anomaly similarity between the target anomaly time period and each initial screening anomaly time period of each downstream device. From the initial screening abnormal period of the equipment to be analyzed, the target period is determined. Based on the target period, the process expansion index at the target abnormal time is calculated, and based on the process expansion index at the target abnormal time, the deterioration index at the target abnormal time is calculated. The product of the actual probability of an anomaly at the target anomaly moment and the deterioration index at the target anomaly moment is determined as the early warning priority index for the target anomaly moment.

[0031] In this embodiment, the data processing module uses the DBSCAN algorithm to cluster historical data to obtain working condition clusters, specifically including: The raw material input, flow rate difference of the equipment under analysis, liquid level difference of the equipment under analysis, and environmental data at each moment in the historical data are respectively defined as the multidimensional sample space of the DBSCAN algorithm. The mean distance between data samples within a preset time interval at each moment is defined as the neighborhood radius of the DBSCAN algorithm. The preset number of samples is defined as the minimum number of samples for the DBSCAN algorithm. The DBSCAN algorithm is used to cluster the historical data to obtain the working condition cluster. Among them, the flow rate difference of the device under analysis at each moment is the difference between the output flow rate data and the input flow rate data of the device under analysis at each moment, and the liquid level difference of the device under analysis at each moment is the difference between the output liquid level data and the input liquid level data of the device under analysis at each moment.

[0032] In the data processing module, based on the operating condition clusters, the abnormal time periods for each monitoring indicator of the equipment to be analyzed, as well as the initial screening abnormal time periods for the equipment to be analyzed, are determined, specifically including: Obtain the monitoring value of the monitoring index k of the equipment to be analyzed at time t, and determine the working condition cluster at time t as the working condition cluster to be analyzed; Calculate the monitoring index k of the equipment to be analyzed, and the average difference between the monitoring value of the equipment at each time step in the cluster of working conditions to be analyzed and the monitoring index k of the equipment at time t. Obtain the monitoring index k of the device to be analyzed, the monitoring value of each time within the preset time interval corresponding to time t, and sort them in chronological order to obtain the time series sequence of monitoring index k; The first-order difference of the time series is obtained to obtain the difference series of the monitoring index k; The ratio of the difference value at time t in the difference sequence of monitoring indicator k to the mean of the elements in the difference sequence of monitoring indicator k is determined as the fluctuation intensity of monitoring indicator k. The product of the mean difference of monitoring indicator k and the fluctuation intensity of monitoring indicator k is determined as the degree of abnormality of monitoring indicator k at time t. Obtain the degree of anomaly of monitoring indicator k at each time point; The moment when the degree of abnormality is greater than the preset threshold for abnormality is determined as the abnormal moment of monitoring indicator k. The abnormal moment of each monitoring indicator is obtained and the abnormal moment is determined as the initial screening abnormal moment of the device to be analyzed. The abnormal times of monitoring indicator k are merged to obtain the abnormal time periods of monitoring indicator k, and the abnormal time periods of each monitoring indicator are obtained. The initial screening abnormal times of the equipment to be analyzed are merged to obtain the initial screening abnormal time period of the equipment to be analyzed.

[0033] In the data processing module, the similarity between the target anomaly time period and each initial screening anomaly time period of each downstream device is obtained, specifically including: Obtain the initial screening anomaly period j of downstream device z; The degree of abnormality of the traffic difference at each moment during the target abnormal period is obtained to obtain the first abnormal traffic sequence of the device to be analyzed; The degree of abnormality of the flow difference at each moment in the initial screening abnormal period j of downstream device z is obtained, and the second flow abnormality sequence of the initial screening abnormal period j of downstream device z is obtained. The degree of abnormality of the liquid level difference at each moment during the target abnormal period is obtained to obtain the first liquid level abnormality sequence of the equipment to be analyzed; The degree of abnormality of the liquid level difference at each moment in the initial screening abnormal period j of the downstream equipment z is obtained, and the second liquid level abnormality sequence of the initial screening abnormal period j of the downstream equipment z is obtained. The DTW algorithm is used to calculate the first distance between the first and second traffic anomaly sequences. The second distance between the first and second liquid level anomaly sequences is calculated using the DTW algorithm. The average of the first and second distances is determined as the abnormal similarity between the target abnormal time period and the initial screening abnormal time period j of the downstream device z. Obtain the anomaly similarity between the target anomaly time period and each initial screening anomaly time period of the downstream device z; Obtain the anomaly similarity between the target anomaly time period and each initial screening anomaly time period of each downstream device.

[0034] In the data processing module, based on the anomaly similarity between the target anomaly time period and each initial screening anomaly time period of each downstream device, the true anomaly probability at the target anomaly time is calculated, specifically including: Based on the anomaly similarity between the target anomaly time period and each initial screening anomaly time period of each downstream device, calculate the probability of anomaly response between the target anomaly time period and each initial screening anomaly time period of each downstream device. The initial screening abnormal period of downstream devices with an abnormal response probability greater than the preset abnormal response probability threshold is determined as the response abnormal period, and the downstream devices corresponding to the response abnormal period are determined as response devices. The sum of the Pearson correlation coefficients of the data sequence of the first monitoring indicator and each of the second monitoring indicators during the target abnormal period is calculated to obtain the data correlation of the first monitoring indicator. The first monitoring indicator is any monitoring indicator during the target abnormal period, and the second monitoring indicator is any monitoring indicator other than the first monitoring indicator during the target abnormal period. During the target abnormal period, each monitoring indicator is determined as the first monitoring indicator. The data correlation of each monitoring indicator is obtained and summed to obtain the data correlation degree of the target abnormal period. Remove the abnormal periods from the historical data of the equipment to be analyzed to obtain the normal periods; The time period within a preset duration interval of the target abnormal time within the normal time period is defined as the normal time period of the target abnormal time. The data correlation of the third monitoring indicator is obtained by summing the Pearson correlation coefficients of the data sequence of the third monitoring indicator and each of the fourth monitoring indicators during the normal period of the target abnormal time. The third monitoring indicator is any monitoring indicator during the normal period of the target abnormal time, and the fourth monitoring indicator is any monitoring indicator other than the third monitoring indicator during the normal period of the target abnormal time. During the normal period of the target abnormal moment, each monitoring indicator is determined as the third monitoring indicator. The data correlation of each monitoring indicator is obtained and summed to obtain the degree of correlation of the normal data at the target abnormal moment. The true probability of an anomaly at the target anomaly time is calculated based on the number of anomaly monitoring indicators at the target anomaly time, the number of response devices at the target anomaly time, the number of downstream devices at the target anomaly time, the correlation of data during the target anomaly period, and the correlation of normal data at the target anomaly time.

[0035] In the data processing module, based on the anomaly similarity between the target anomaly time period and each initial screening anomaly time period of each downstream device, the probability of anomaly response between the target anomaly time period and each initial screening anomaly time period of each downstream device is calculated, specifically including: Obtain the start time of the initial screening anomaly period j of downstream device z; Calculate the start time of the initial screening anomaly period j of downstream device z and the target time length between the target anomaly period and the target anomaly time corresponding to the target anomaly period; Obtain the standard material handling time for the device to be analyzed and the downstream device z; The ratio of the standard material transport time to the target time length of the equipment to be analyzed and the downstream equipment z is determined as the standard material transport rate of the initial screening abnormal period j of the downstream equipment z. The reciprocal of the abnormality similarity between the target abnormal time period and the initial screening abnormal time period j of the downstream equipment z is multiplied by the standard material conveying rate of the initial screening abnormal time period j of the downstream equipment z to obtain the abnormal response probability of the target abnormal time period and the initial screening abnormal time period j of the downstream equipment z. Obtain the probability of abnormal response for each initial screening abnormal period of the target abnormal period and the downstream device z; Obtain the probability of abnormal response for the target abnormal period and each initial screening abnormal period for each downstream device.

[0036] In the data processing module, based on the number of abnormal monitoring indicators at the target abnormal time, the number of response devices at the target abnormal time, the number of downstream devices at the target abnormal time, the correlation of data during the target abnormal period, and the correlation of normal data at the target abnormal time, the true probability of an anomaly at the target abnormal time is calculated, specifically including: The ratio of the correlation between the data during the target abnormal period and the correlation between the normal data at the target abnormal moment is determined as the internal correlation at the target abnormal moment. The adjustment value is obtained by adding the number of downstream devices at the time of the target anomaly to a preset value divided by zero. The ratio of the number of responding devices to the adjustment value at the time of the target anomaly is determined as the downstream device ratio at the time of the target anomaly. The true probability of an anomaly at the target anomaly moment is determined by multiplying the number of anomaly monitoring indicators, their internal correlation, and the proportion of downstream equipment.

[0037] In the data processing module, the target time period is determined from the initial screening of abnormal periods of the devices to be analyzed, specifically including: The cluster of operating conditions at the time of the target anomaly is determined as the target operating condition cluster; Within the target operating condition cluster, other times are identified as historically similar states. By merging the anomalous moments in similar historical states, the target time period is obtained.

[0038] In the data processing module, based on the target time period, the process amplification index at the target abnormal moment is calculated, specifically including: Obtain the response devices for the target time period, and obtain the distance of the response device that is farthest from the device to be analyzed during the target time period to obtain the first distance; Obtain the response devices during the target abnormal period, and obtain the distance of the response device that is farthest from the device to be analyzed during the target abnormal period to obtain the second distance; Obtain the abnormal response period of response device a corresponding to the target abnormal period, obtain the abnormal response probability of response device a for each abnormal response period and calculate the average value to obtain the average abnormal response probability of response device a corresponding to the target abnormal period. Obtain the average probability of abnormal response for each responding device during the target abnormal period and calculate the average to obtain the average abnormal response during the target abnormal period; The ratio of the second distance to the first distance is determined as the extent of the target's impact at the moment of an anomaly. The product of the impact breadth at the target anomaly moment and the average anomaly response during the target anomaly period is determined as the process amplification index at the target anomaly moment.

[0039] In the data processing module, based on the process amplification index at the target anomaly time, the deterioration index at the target anomaly time is calculated, specifically including: The duration of the target abnormal period, the average duration of the target period, and the growth period in the initial screening abnormal period of the device to be analyzed are obtained. In each growth period, the abnormality of the monitoring indicator at time b is greater than the abnormality of the monitoring indicator at time b-1. The difference between the degree of abnormality of monitoring indicator c at time d and the degree of abnormality of monitoring indicator c at time d-1 is determined as the abnormal growth amount of monitoring indicator c at time d. Obtain the abnormal growth amount of monitoring indicator c at each time point; Obtain the average abnormal growth of each monitoring indicator at each time point to obtain the average abnormal growth. The time duration is determined by the ratio of the duration of the target abnormal period to the average duration of the target period. The deterioration index of the target abnormal moment is determined by multiplying the duration of time, the number of moments in the growth period, the average abnormal growth, and the process expansion index of the target abnormal moment.

[0040] For example, the preset time interval in this embodiment can refer to 24 hours, because 24 hours can basically cover all the reaction processes of each reaction device within a day. The specific length of the preset time interval can be selected according to the actual situation, and no specific limitation is made here. Furthermore, in this embodiment, time t is used as the target abnormal time for analysis.

[0041] The preset time interval can be regarded as a sliding monitoring window at each moment. Taking 24 hours as a sliding monitoring window, the possibility of anomalies in the monitoring data of a single dimension of the stirring equipment in the polymerization reactor is screened (taking the monitoring data of monitoring index k at time t as an example): When there exists a moment in the historical data (using all monitoring data within the past week as historical data) where the raw material input is similar to the current moment, the production capacity of the polymerization reactor is similar (i.e., the difference between the output flow and the input flow of the corresponding process of the equipment), and the external environment of the equipment is similar, then that moment can be considered as a historically similar state to the current moment.

[0042] Based on the definition of historical similarity mentioned above, a multidimensional sample space is constructed. The data points of the device in the historical data are arranged in the sample space. The multidimensional sample space is constructed with the following monitoring indicators, and each indicator corresponds to the single-dimensional data in the sample space: the input amount of raw materials of the overall production line at the current moment, the difference between the output and input flow of the corresponding process of the device, the difference between the output and input liquid levels of the device, and the temperature and humidity of the workshop or factory corresponding to the device.

[0043] The mean distance between data samples at each time point within a preset time interval is represented as follows: Taking time t as an example, the mean distance between data samples at time t within a preset time interval is the mean distance between each time point and time t in the sample space within the preset time interval of time t.

[0044] The DBSCAN algorithm is used to cluster the data in the sample space. The neighborhood radius of the DBSCAN algorithm is the average distance between all data points at time t and the corresponding data point at time t within the past hour in the sample space. The minimum number of samples in a cluster using the DBSCAN algorithm is 50. There are 3 operating condition clusters, where each data point in the cluster at time t corresponds to a historical similar state of the equipment at time t.

[0045] Optionally, the preset sample size can be 50 in a preferred embodiment. This is just an example and can be set based on historical experience without any specific limitation.

[0046] If the monitoring data of the device's monitoring indicator k at time t differs significantly from its corresponding historical similar state, and its fluctuation compared to time t-1, and its fluctuation amplitude compared to the previous time within the monitoring window corresponding to time t (i.e., the 24 hours past time t), then the probability of an anomaly in this indicator at time t is high. Larger: Arrange the monitoring values ​​of index k within the monitoring window corresponding to time t in chronological order, and perform a first-order difference on them. The difference between time t and t-1 is denoted as... That is, the difference value corresponding to time t.

[0047] Let the monitoring value of indicator k at time t be . The monitoring values ​​corresponding to all historical similar states are respectively. The mean of the differences is denoted as That is, the mean difference of the monitoring indicator k.

[0048] Therefore, the formula for calculating the degree of anomaly of monitoring indicator k at time t can be:

[0049] in, This indicates the degree of abnormality of monitoring indicator k at time t. This represents the mean difference of the monitoring indicator k. This represents the difference value corresponding to time t. This represents the mean of the elements in the difference sequence of monitoring indicator k. This indicates the fluctuation intensity of the monitoring indicator k.

[0050] Will Normalized to (0, 1), if the normalized result of the abnormality degree at time t is greater than 0.5, it is marked as an abnormal time for the monitoring index k of the equipment. If at any time an index is marked as an abnormal time for the monitoring index k of the equipment, that time is marked as the initial screening abnormal time for the current equipment. When the abnormality degree of any monitoring index is greater than the normality degree at a certain time, this monitoring index can be considered as an abnormal monitoring index at that time.

[0051] Optionally, the preset anomaly threshold can be set to 0.5 in a preferred embodiment. This can be set according to actual needs or historical experience, and no specific numerical limit is imposed here. Correspondingly, the anomaly level can be normalized or not. After normalization, it still needs to meet the preset anomaly threshold. An anomaly level without normalization can be 0.5 or other values, and no specific limit is imposed here.

[0052] Taking the monitoring index k of the polymerization reactor stirring equipment as an example, the abnormal times of the index appearing continuously are merged to obtain the abnormal time period of the index. Similarly, the abnormal time period of the current equipment's initial screening is obtained. Then, in the historical data, except for the abnormal time period of the current equipment's initial screening, the other time periods are normal time periods.

[0053] Within the sliding monitoring window corresponding to time t, the Pearson correlation coefficients between each pair of monitoring indicators during all normal time periods are obtained, which is also the data correlation of the third monitoring indicator.

[0054] When there is a real malfunction in the stirring equipment of the reactor on the petroleum resin production line, it usually affects the production efficiency of other processes. For example, factors such as bearing wear and impeller deformation can cause sudden changes in stirring resistance, resulting in uneven mixing of materials output to the next process, which in turn leads to abnormal separation efficiency in the downstream distillation column, and even abnormal product purity.

[0055] In actual petroleum resin production, numerous temporary random disturbances occur, such as load switching and raw material content fluctuations. These can cause temporary abnormal fluctuations in the monitoring data of the stirring equipment (but the stirring equipment's own production capacity remains unaffected). For example, when the olefin purity is low, process adjustments may lead to a certain increase in the temperature inside the reactor, and there may be interference from upstream equipment outputs during production. Existing methods are prone to misjudging these temporary abnormal disturbances as stirring equipment malfunctions. Therefore, it is necessary to combine the response correlation of each monitoring device and the degree of impact of the current process on downstream production to eliminate temporary abnormal disturbances.

[0056] Temporary anomalies usually only interfere with a few monitoring indicators and do not affect the actual production capacity of the polymerization reactor stirring equipment. However, actual equipment anomalies are usually caused by abnormal production capacity, resulting in multiple monitoring indicators of the equipment being abnormal at the same time.

[0057] Taking a certain initial screening abnormal moment of the current equipment as an example (e.g., time t), if the more indicators are marked as abnormal at that time, and if the correlation between the abnormal indicator and other indicators is poor and there are many downstream processes with similar abnormal fluctuations during the abnormal period of a certain abnormal indicator, then the equipment is more likely to be in a real abnormal state at time t.

[0058] Taking the z-th downstream device as an example (the downstream device is determined based on the data twin model), the initial screening anomaly period of the downstream device z is obtained by the same operation. If the initial screening anomaly period of a certain downstream device z corresponding to the initial screening anomaly period at time t is closer to the duration of time t and the duration of material transfer between the two devices, then the degree of capacity anomaly of the two devices is similar, and the downstream device z responds to the initial screening anomaly period at time t.

[0059] The analysis focuses on the initial screening anomaly period corresponding to time t for the current device and the initial screening anomaly period j for the downstream device z. The analysis then proceeds to obtain the results for each of these two periods. and The two indicators correspond to the degree of abnormality. The DTW distance is used to record the degree of anomaly of the two indicators in the two time periods mentioned above. The mean DTW distance is In other words, the smaller the value of the abnormality similarity between the target abnormal time period and the initial screening abnormal time period j of the downstream equipment z, the more similar the degree of abnormality of the current equipment at time t to the production capacity of equipment z at the initial screening abnormal time period j, that is, the higher the abnormality similarity.

[0060] The probability of an abnormal response of downstream device z at time t during the initial screening anomaly period j. It can be calculated as follows: In the formula, This refers to the standard time for materials to be transported from the current mixing equipment to the downstream equipment z, that is, the standard material transport time. This refers to the time interval between the start time of the initial screening anomaly period j corresponding to device z and time t, which is also the target time length. The smaller the value, the more the initial screening abnormal period j conforms to the standard material conveying rate.

[0061] Optionally, the preset abnormal response probability threshold can be set to 0.5 here. The specific value of the preset abnormal response probability threshold can be modified according to actual needs or set according to historical experience. There are no specific restrictions here.

[0062] Will Normalize to (0, 1), and mark the devices corresponding to time periods with normalization results greater than 0.5 as the response devices at time t.

[0063] Based on the number of abnormal monitoring indicators at the target anomaly time, the number of response devices at the target anomaly time, the number of downstream devices at the target anomaly time, the correlation of data during the target anomaly period, and the correlation of normal data at the target anomaly time, the true anomaly probability at the target anomaly time is calculated. The calculation formula can be:

[0064] in, Indicates the true probability of an anomaly at the moment of target anomaly. This indicates the number of abnormal monitoring indicators at the moment of the target's abnormality. This indicates the number of responding devices at the moment of the target anomaly. This indicates the number of downstream devices at the moment of the target anomaly. This indicates the degree of correlation between data during the target abnormal period. This indicates the degree of correlation between the target abnormal time and the normal data. The default value for division by zero is 0.01, but it can also be other values. There are no specific restrictions here; this is just an example. When a certain device is the last device in the production line, there are subsequent downstream devices. Therefore, a default value for division by zero is set here to avoid the denominator being 0.

[0065] It is obtained by summing the Pearson correlation coefficients between each pair of indicators within the initial anomaly screening period at time t. It is the sum of the Pearson correlation coefficients between each pair of indicators during the normal period of the target abnormal moment. The target is the proportion of downstream equipment at the moment of an anomaly. Real equipment anomalies (especially those affecting capacity or quality) will alter their outputs (such as material flow rate, temperature, and composition), thereby affecting the normal operation of downstream equipment. This chain reaction along a predetermined production path, consistent with material transport time, is extremely strong evidence proving the authenticity of the current anomaly. The higher the proportion, the stronger the authenticity. To understand the internal correlations during a target anomaly, under stable operating conditions, the various parameters within the equipment, controlled by physical laws and processes, form a stable and coordinated relationship (i.e., strong correlation). When a real fault occurs, this inherent stable relationship is broken, the linkage between indicators weakens or even disappears, and the correlation significantly decreases. Therefore, the larger this ratio, the more severe the collapse of the internal order during the anomaly period, and the greater the likelihood of a real anomaly.

[0066] Similarly, taking the device at time t (where time t is the time when abnormal monitoring indicators exist) as an example, we obtain the times when the device is in the same initial screening abnormal state in its historical similar state and the initial screening abnormal period in which it is located, and mark the obtained initial screening abnormal period as the target period.

[0067] Based on the target time period, the formula for calculating the process expansion index at the target abnormal moment can be:

[0068] in, The process expansion index represents the process expansion index at the time of the target anomaly, that is, the process expansion index at time t. It is the distance of the device farthest from the current stirring device among all the response devices in the time period at time t, that is, the distance of the response device farthest from the device to be analyzed among the response devices in the target abnormal time period; This refers to the distance between the current mixing device and the responding device among all responding devices during the target time period; that is, the distance between the current mixing device and the responding device among all responding devices during the target time period. For all response devices in the time period at time t The mean value, that is, the mean value of the abnormal response probability of each response device corresponding to the target abnormal period, is obtained by averaging the mean value of the abnormal response during the target abnormal period.

[0069] In the formula, if the time period t is further away from the range of processes affected by all target time periods, the degree of influence is greater, and it is more in line with the trend of process expansion.

[0070] The formula for calculating the deterioration index at the moment of the target anomaly can be:

[0071] in, Indicates the duration of the period of abnormality in the target. This represents the average duration of the target time period. This indicates the number of moments within the growth period. This represents the average amount of abnormal growth. The process expansion index represents the time of the target anomaly.

[0072] In the formula, during any initial screening anomaly period, all indicators of the equipment correspond to All of them are experiencing growth, among which, Therefore, this period of initial abnormal screening is recorded as the growth period.

[0073] Therefore, the warning priority index at time t of the initial anomaly screening for:

[0074] The information early warning module 103 is used to determine each moment in the initial screening abnormal period of the device to be analyzed as the target abnormal moment, obtain the early warning priority index of each moment, and determine the early warning information of each moment based on the early warning priority index of each moment.

[0075] For example, it can be Normalize to (0, 1) and label the normalization result. Prioritize early warning.

[0076] Predictive maintenance is performed on the production equipment in the petroleum resin production line. Taking the stirring equipment in the polymerization reactor as an example, its operating status is monitored from multiple dimensions. By combining actual working conditions and production line processes, digital twin modeling is carried out. The multi-dimensional monitoring data obtained in the above steps is transmitted to the nearby edge computing node. After preprocessing, the nearby historical data is read to filter the initial abnormal moments.

[0077] By combining data from various dimensions to screen for abnormal moments, the abnormal responses of each dimension's indicators and their impact on subsequent process abnormalities are used to distinguish between temporary abnormal interference and real equipment abnormalities, and the warning priority is determined based on the abnormal development trend.

[0078] The determined warning priorities are transmitted to the output device to warn staff. Specifically: when the priority index is (0, 0.2), no warning is issued; when it is (0.2, 0.4), a low-level warning is issued; when it is (0.4, 0.6), a medium-level warning is issued; when it is (0.6, 0.8), a high-level warning is issued; and when it is (0.8, 1), a severe warning is issued. In other words, the higher the priority, the more likely the current anomaly is a genuine malfunction of the polymerization reactor's agitator and is in a state of rapid deterioration. This allows staff to take timely maintenance measures based on different warning levels.

[0079] In summary, in this embodiment of the invention, abnormal moments of each monitoring indicator are initially screened by combining the fluctuations of monitoring data under similar working conditions. Based on digital twin technology, the impact of abnormality of a single indicator on upstream and downstream processes and the associated abnormal responses of other monitoring indicators are analyzed, thereby distinguishing between temporary interference and real equipment abnormalities. Furthermore, the warning priority is determined by combining the development trend of real abnormalities and the impact of corresponding equipment on upstream and downstream processes, so as to facilitate the staff to take corresponding maintenance measures.

[0080] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0081] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0082] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A predictive maintenance system for petroleum resin production line equipment based on digital twins, characterized in that, The system includes: The data acquisition module is used to acquire historical data of the equipment to be analyzed in the petroleum resin production line. The data processing module is used to cluster historical data using the DBSCAN algorithm to obtain operating condition clusters, and based on the operating condition clusters, to determine the abnormal periods of each monitoring indicator of the equipment to be analyzed and the initial screening abnormal periods of the equipment to be analyzed. From the initial screening abnormal period of the equipment to be analyzed, the target abnormal period is determined. The downstream equipment of the equipment to be analyzed is obtained from the petroleum resin digital twin model. Each initial screening abnormal period of each downstream equipment is obtained. The target abnormal period is the initial screening abnormal period in which the target abnormal time is located. Obtain the anomaly similarity between the target anomaly time period and each initial screening anomaly time period of each downstream device; Based on the anomaly similarity between the target anomaly time period and each initial screening anomaly time period of each downstream device, the true anomaly probability at the target anomaly time is calculated, specifically including: Based on the anomaly similarity between the target anomaly time period and each initial screening anomaly time period of each downstream device, calculate the probability of anomaly response between the target anomaly time period and each initial screening anomaly time period of each downstream device. The initial screening abnormal period of downstream devices with an abnormal response probability greater than the preset abnormal response probability threshold is determined as the response abnormal period, and the downstream devices corresponding to the response abnormal period are determined as response devices. The degree of correlation of data during the target anomaly period is determined by the sum of the Pearson correlation coefficients between each pair of monitoring indicators during the target anomaly period. Remove the abnormal periods from the historical data of the equipment to be analyzed to obtain the normal periods; The time period within a preset duration interval of the target abnormal time within the normal time period is defined as the normal time period of the target abnormal time. The correlation between normal data at the time of the target anomaly is determined by the sum of the Pearson correlation coefficients between each pair of monitoring indicators during the normal period at the time of the target anomaly. The ratio of the correlation between the data during the target abnormal period and the correlation between the normal data at the target abnormal moment is determined as the internal correlation at the target abnormal moment. The adjustment value is obtained by adding the number of downstream devices at the time of the target anomaly to a preset value divided by zero. The ratio of the number of responding devices to the adjustment value at the time of the target anomaly is determined as the downstream device ratio at the time of the target anomaly. The product of the number of abnormal monitoring indicators, internal correlation, and the proportion of downstream equipment at the time of the target abnormality is determined as the true probability of the abnormality at the time of the target abnormality. From the initial screening abnormal period of the equipment to be analyzed, the target period is determined. Based on the target period, the process expansion index at the target abnormal time is calculated, and based on the process expansion index at the target abnormal time, the deterioration index at the target abnormal time is calculated. The product of the actual probability of an anomaly at the target anomaly moment and the deterioration index at the target anomaly moment is determined as the early warning priority index for the target anomaly moment. The information early warning module is used to identify each moment in the initial screening abnormal period of the device to be analyzed as the target abnormal moment, obtain the early warning priority index of each moment, and determine the early warning information for each moment based on the early warning priority index of each moment.

2. The predictive maintenance system for petroleum resin production line equipment based on digital twins according to claim 1, characterized in that, In the data processing module, the DBSCAN algorithm is used to cluster historical data to obtain working condition clusters, specifically including: The raw material input, flow rate difference of the equipment under analysis, liquid level difference of the equipment under analysis, and environmental data at each moment in the historical data are respectively defined as the multidimensional sample space of the DBSCAN algorithm. The mean distance between data samples within a preset time interval at each moment is defined as the neighborhood radius of the DBSCAN algorithm. The preset number of samples is defined as the minimum number of samples for the DBSCAN algorithm. The DBSCAN algorithm is used to cluster the historical data to obtain the working condition cluster. Among them, the flow rate difference of the device under analysis at each moment is the difference between the output flow rate data and the input flow rate data of the device under analysis at each moment, and the liquid level difference of the device under analysis at each moment is the difference between the output liquid level data and the input liquid level data of the device under analysis at each moment.

3. The predictive maintenance system for petroleum resin production line equipment based on digital twins according to claim 1, characterized in that, In the data processing module, based on the operating condition clusters, the abnormal time periods for each monitoring indicator of the equipment to be analyzed and the initial screening abnormal time periods for the equipment to be analyzed are determined, specifically including: Obtain the monitoring value of the monitoring index k of the equipment to be analyzed at time t, and determine the working condition cluster at time t as the working condition cluster to be analyzed; Calculate the monitoring index k of the equipment to be analyzed, and the average difference between the monitoring value of the equipment at each time step in the cluster of working conditions to be analyzed and the monitoring index k of the equipment at time t. Obtain the monitoring index k of the device to be analyzed, the monitoring value of each time within the preset time interval corresponding to time t, and sort them in chronological order to obtain the time series sequence of monitoring index k; The first-order difference of the time series is obtained to obtain the difference series of the monitoring index k; The ratio of the difference value at time t in the difference sequence of monitoring indicator k to the mean of the elements in the difference sequence of monitoring indicator k is determined as the fluctuation intensity of monitoring indicator k. The product of the mean difference of monitoring indicator k and the fluctuation intensity of monitoring indicator k is determined as the degree of abnormality of monitoring indicator k at time t. Obtain the degree of anomaly of monitoring indicator k at each time point; The moment when the degree of abnormality is greater than the preset threshold for abnormality is determined as the abnormal moment of monitoring indicator k. The abnormal moment of each monitoring indicator is obtained and the abnormal moment is determined as the initial screening abnormal moment of the device to be analyzed. The abnormal times of monitoring indicator k are merged to obtain the abnormal time period of monitoring indicator k, and the abnormal time period of each monitoring indicator is obtained. The initial screening abnormal times of the equipment to be analyzed are merged to obtain the initial screening abnormal time period of the equipment to be analyzed.

4. The predictive maintenance system for petroleum resin production line equipment based on digital twins according to claim 1, characterized in that, The data processing module obtains the anomaly similarity between the target anomaly time period and each initial screening anomaly time period of each downstream device, specifically including: Obtain the initial screening anomaly period j of downstream device z; The degree of abnormality of the traffic difference at each moment during the target abnormal period is obtained to obtain the first abnormal traffic sequence of the device to be analyzed; The degree of abnormality of the flow difference at each moment in the initial screening abnormal period j of downstream device z is obtained, and the second flow abnormality sequence of the initial screening abnormal period j of downstream device z is obtained. The degree of abnormality of the liquid level difference at each moment during the target abnormal period is obtained to obtain the first liquid level abnormality sequence of the equipment to be analyzed; The degree of abnormality of the liquid level difference at each moment in the initial screening abnormal period j of the downstream equipment z is obtained, and the second liquid level abnormality sequence of the initial screening abnormal period j of the downstream equipment z is obtained. The DTW algorithm is used to calculate the first distance between the first and second traffic anomaly sequences. The second distance between the first and second liquid level anomaly sequences is calculated using the DTW algorithm. The average of the first and second distances is determined as the abnormal similarity between the target abnormal time period and the initial screening abnormal time period j of the downstream device z. Obtain the anomaly similarity between the target anomaly time period and each initial screening anomaly time period of the downstream device z; Obtain the anomaly similarity between the target anomaly time period and each initial screening anomaly time period of each downstream device.

5. The predictive maintenance system for petroleum resin production line equipment based on digital twins according to claim 1, characterized in that, In the data processing module, the true probability of an anomaly at the target anomaly time is calculated based on the anomaly similarity between the target anomaly time period and each initial screening anomaly time period of each downstream device. Specifically, this includes: Based on the anomaly similarity between the target anomaly time period and each initial screening anomaly time period of each downstream device, calculate the probability of anomaly response between the target anomaly time period and each initial screening anomaly time period of each downstream device. The initial screening abnormal period of downstream devices with an abnormal response probability greater than the preset abnormal response probability threshold is determined as the response abnormal period, and the downstream devices corresponding to the response abnormal period are determined as response devices. The sum of the Pearson correlation coefficients of the data sequence of the first monitoring indicator and each of the second monitoring indicators during the target abnormal period is calculated to obtain the data correlation of the first monitoring indicator. The first monitoring indicator is any monitoring indicator during the target abnormal period, and the second monitoring indicator is any monitoring indicator other than the first monitoring indicator during the target abnormal period. During the target abnormal period, each monitoring indicator is determined as the first monitoring indicator. The data correlation of each monitoring indicator is obtained and summed to obtain the data correlation degree of the target abnormal period. Remove the abnormal periods from the historical data of the equipment to be analyzed to obtain the normal periods; The time period within a preset duration interval of the target abnormal time within the normal time period is defined as the normal time period of the target abnormal time. The data correlation of the third monitoring indicator is obtained by summing the Pearson correlation coefficients of the data sequence of the third monitoring indicator and each of the fourth monitoring indicators during the normal period of the target abnormal time. The third monitoring indicator is any monitoring indicator during the normal period of the target abnormal time, and the fourth monitoring indicator is any monitoring indicator other than the third monitoring indicator during the normal period of the target abnormal time. During the normal period of the target abnormal moment, each monitoring indicator is determined as the third monitoring indicator. The data correlation of each monitoring indicator is obtained and summed to obtain the degree of correlation of the normal data at the target abnormal moment. The true probability of an anomaly at the target anomaly time is calculated based on the number of anomaly monitoring indicators at the target anomaly time, the number of response devices at the target anomaly time, the number of downstream devices at the target anomaly time, the correlation of data during the target anomaly period, and the correlation of normal data at the target anomaly time.

6. The predictive maintenance system for petroleum resin production line equipment based on digital twins according to claim 5, characterized in that, In the data processing module, based on the anomaly similarity between the target anomaly time period and each initial screening anomaly time period of each downstream device, the probability of anomaly response between the target anomaly time period and each initial screening anomaly time period of each downstream device is calculated, specifically including: Obtain the start time of the initial screening anomaly period j of downstream device z; Calculate the start time of the initial screening anomaly period j of downstream device z and the target time length between the target anomaly period and the target anomaly time corresponding to the target anomaly period; Obtain the standard material handling time for the device to be analyzed and the downstream device z; The ratio of the standard material transport time to the target time length of the equipment to be analyzed and the downstream equipment z is determined as the standard material transport rate of the initial screening abnormal period j of the downstream equipment z. The reciprocal of the abnormality similarity between the target abnormal time period and the initial screening abnormal time period j of the downstream equipment z is multiplied by the standard material conveying rate of the initial screening abnormal time period j of the downstream equipment z to obtain the abnormal response probability of the target abnormal time period and the initial screening abnormal time period j of the downstream equipment z. Obtain the probability of abnormal response for each initial screening abnormal period of the target abnormal period and the downstream device z; Obtain the probability of abnormal response for the target abnormal period and each initial screening abnormal period for each downstream device.

7. The predictive maintenance system for petroleum resin production line equipment based on digital twins according to claim 1, characterized in that, In the data processing module, the target time period is determined from the initial screening of abnormal periods of the devices to be analyzed, specifically including: The cluster of operating conditions at the time of the target anomaly is determined as the target operating condition cluster; Within the target operating condition cluster, other times are identified as historically similar states. By merging the anomalous moments in similar historical states, the target time period is obtained.

8. The predictive maintenance system for petroleum resin production line equipment based on digital twins according to claim 1, characterized in that, The data processing module calculates the process amplification index at the target abnormal moment based on the target time period, specifically including: Obtain the response devices for the target time period, and obtain the distance of the response device that is farthest from the device to be analyzed during the target time period to obtain the first distance; Obtain the response devices during the target abnormal period, and obtain the distance of the response device that is farthest from the device to be analyzed during the target abnormal period to obtain the second distance; Obtain the abnormal response period of response device a corresponding to the target abnormal period, obtain the abnormal response probability of response device a for each abnormal response period and calculate the average value to obtain the average abnormal response probability of response device a corresponding to the target abnormal period. Obtain the average probability of abnormal response for each responding device during the target abnormal period and calculate the average to obtain the average abnormal response during the target abnormal period; The ratio of the second distance to the first distance is determined as the extent of the target's impact at the moment of an anomaly. The product of the impact breadth at the target anomaly moment and the average anomaly response during the target anomaly period is determined as the process amplification index at the target anomaly moment.

9. The predictive maintenance system for petroleum resin production line equipment based on digital twins according to claim 1, characterized in that, The data processing module calculates the deterioration index at the target anomaly time based on the process expansion index at the target anomaly time, specifically including: The duration of the target abnormal period, the average duration of the target period, and the growth period in the initial screening abnormal period of the device to be analyzed are obtained. In each growth period, the abnormality of the monitoring indicator at time b is greater than the abnormality of the monitoring indicator at time b-1. The difference between the degree of abnormality of monitoring indicator c at time d and the degree of abnormality of monitoring indicator c at time d-1 is determined as the abnormal growth amount of monitoring indicator c at time d. Obtain the abnormal growth amount of monitoring indicator c at each time point; Obtain the average abnormal growth of each monitoring indicator at each time point to obtain the average abnormal growth. The time duration is determined by the ratio of the duration of the target abnormal period to the average duration of the target period. The deterioration index of the target abnormal moment is determined by multiplying the duration of time, the number of moments in the growth period, the average abnormal growth, and the process expansion index of the target abnormal moment.

Images