Model predictive control device risk management system

Abstract

The application relates to the technical field of equipment risk management and control, and particularly discloses a model predictive control-based equipment risk management and control system, which comprises a data acquisition and processing module, a data analysis module, a model construction and prediction module and an analysis and control module, wherein the data acquisition and processing module is used for collecting equipment operation state information, and obtaining an equipment health condition score through standardized processing and weighted calculation; the data analysis module distinguishes linear and nonlinear characteristic signals according to a score change trend; the model construction and prediction module selects a corresponding algorithm to establish a prediction model according to different types of signals, and realizes prediction of a future equipment health state; and the analysis and control module formulates a reasonable equipment maintenance plan based on a prediction result, and dynamically adjusts production arrangement. The application can effectively prevent faults from occurring and reduce the occurrence probability of unplanned shutdown events through continuous monitoring and intelligent analysis of the equipment state, so that the purpose of improving production safety and economic benefits is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of equipment risk management, and particularly relates to a model predictive control-based equipment risk management and control system. BACKGROUND

[0002] With the continuous improvement of industrial automation level, the stable operation of production equipment has become one of the key factors for efficient production and sustainable development of enterprises. However, due to the wear and potential failure caused by long-time operation of the equipment, it may lead to production interruption and even economic loss. Therefore, how to effectively manage and control the risk of the equipment has become a problem to be solved. The existing equipment risk management and control technology mostly relies on after-event maintenance or regular maintenance, which is not only inefficient, but also often fails to discover potential failures in time, resulting in an increased probability of sudden equipment shutdown. Therefore, developing a method capable of real-time monitoring of equipment state and risk assessment based on a prediction model is of great significance for improving equipment management efficiency and reducing unplanned downtime.

[0003] The existing equipment maintenance technology mainly includes after-event maintenance, regular maintenance and state-based maintenance. After-event maintenance is to repair the equipment after failure, which is simple and direct, but lacks preventive measures, easily leading to long-term unstable state of the equipment and increasing maintenance cost. Regular maintenance is to repair the equipment according to a predetermined period, which can prevent some failures, but lacks pertinence, easily causing resource waste. State-based maintenance is to monitor the equipment operation data in real time through sensors and assess the equipment state in combination with data analysis means to determine whether maintenance is needed, which is an advanced method, but still faces challenges such as complex data processing, difficulty in constructing prediction model and high implementation cost. Although these methods have advantages such as real-time monitoring and prediction improving fault warning capability and being helpful to preventive maintenance, they also have significant disadvantages such as complex data processing, difficulty in constructing prediction model and high implementation cost.

[0004] In view of the problems existing in the prior art, the present application proposes a model predictive control-based equipment risk management and control system, which aims to realize effective management of equipment risk through model predictive control technology. The core of the present application is to comprehensively analyze whether there is an abnormal situation by monitoring the vibration signal, temperature and current data of the equipment in real time, and if an abnormality is detected, further analyze whether there is a rule (such as linear or nonlinear) in the data, and select a suitable method to construct a device risk prediction model, obtain the risk trend based on the model, and analyze the length of time from the current time to the risk arrival time according to the trend, and finally decide whether to need to be regulated and controlled according to whether the remaining use time of the equipment exceeds the risk arrival time. SUMMARY

[0005] The purpose of the present application is to provide a model-based predictive control device risk management system, which mainly includes the following steps: by collecting the state data of the device (such as vibration signal, temperature and current), and integrating and standardizing the data, the comprehensive score of the device is calculated by the preset weight coefficient, and then the data abnormal signal or data normal signal is generated; based on the data abnormal signal, the changing trend of the device state data with time is further analyzed, and the linear trend signal or nonlinear trend signal is generated; according to the nonlinear trend signal, the device risk prediction model is constructed, and the future state of the device is predicted; based on the prediction result, the device risk trend curve is constructed, the expected service life of the device is obtained, and the device is adjusted according to the state signal of the device. In addition, the monitoring period is divided into a plurality of equal length time subunits and time small units, the state data of the device is collected at the middle time point of each time small unit, and the data is standardized by calculating the mean and standard deviation of the device state data in each time subunit, the comprehensive score is calculated by using the weight coefficient and compared with the threshold value, so as to judge whether the device state is abnormal. For the data abnormal signal, it is further judged whether the vibration signal, temperature or current data shows a linear trend, and a suitable prediction model is selected accordingly: when the data shows a linear trend, a linear regression model is used for prediction, and when the data shows a nonlinear trend, a random forest model is used to construct a device risk assessment model to predict the future state change of the device. Based on the prediction model, the present application predicts the risk trend in the future period of time by collecting the real-time state data of the device, and constructs the risk trend curve of the device, compares the expected service life of the device with the remaining working time, and decides whether maintenance measures need to be taken in advance to ensure that the device can continue to operate without immediate intervention, or timely maintenance to prolong the service life of the device and reduce the risk. In summary, the present application realizes real-time monitoring of device state data, intelligent analysis and prediction, and dynamic control strategy, which not only improves the efficiency of device management, but also effectively prolongs the service life of the device, reduces the unplanned downtime, and improves the safety and economy of the production process.

[0006] The purpose of the present application can be realized by the following technical solutions:

[0007] A model-based predictive control device risk management system, comprising a data acquisition and processing module, a data analysis module, a model construction and prediction module, and an analysis and control module;

[0008] The data acquisition and processing module is used for collecting the state data of the device;

[0009] The state data of the device is integrated and standardized;

[0010] According to the standardized state data, the comprehensive score of the device is calculated, and the data abnormal signal and the data normal signal are generated;

[0011] The data analysis module further analyzes the change trend of equipment status data over time based on data anomaly signals and generates linear trend signals and nonlinear trend signals;

[0012] The model building prediction module builds a risk prediction model based on linear trend signals and nonlinear trend signals;

[0013] The analysis and control module obtains the remaining working time of the equipment based on the risk prediction model, performs comparative analysis, and generates equipment operation signals, where the equipment operation signals include equipment normal working signals and equipment maintenance signals.

[0014] As a further technical solution of the present invention, the status data of the device includes a vibration intensity value, a temperature value and a current value;

[0015] Define the monitoring cycle and subdivide it into multiple equal-length time subunits and smaller time units, and collect device status data at the middle point of each time unit;

[0016] The data collected in each time sub-unit are integrated and processed to calculate the mean value of each state data in the time period, which is recorded as the state data value;

[0017] Based on the status data values ​​in each time subunit, the mean and standard deviation of the status data of the equipment in the monitoring period are calculated.

[0018] As a further technical solution of the present invention, based on the mean and standard deviation of each state data of the device within the monitoring period, the mean of each state data corresponding to the monitoring period is subtracted from each state data value within the time subunit, and then divided by the standard deviation of each state data corresponding to the monitoring period, to obtain the standard value of each state data within each time subunit;

[0019] Among them, the standard values ​​of various status data in the time sub-unit include vibration intensity standard value, temperature standard value and current standard value.

[0020] As a further technical solution of the present invention, the standard values ​​of the status data of all time sub-units obtained within the monitoring period are integrated, and the formula DF i =w1×WD i +w2×ZD i +w3×DL i Calculate the comprehensive score DF based on equipment vibration, temperature and current in each time subunit i ;

[0021] Wherein, w1 is the weight coefficient assigned to the temperature standard value WD, w2 is the weight coefficient assigned to the vibration standard value ZD, w3 is the weight coefficient assigned to the current standard value DL, i is the number of time sub-units, i = 1, 2,..., n, WD i is the temperature standard value of the i th time sub-unit, ZD i and DL i Similarly.

[0022] As a further technical solution of the present application is: based on the comprehensive score DF, the comprehensive score DF is compared with the score threshold value;

[0023] If the comprehensive score is greater than the score threshold value, a data anomaly signal is generated;

[0024] If the comprehensive score is less than or equal to the score threshold value, a data normal signal is generated.

[0025] As a further technical solution of the present application is: the process of analyzing the trend of the device state data over time to generate linear trend signals and nonlinear trend signals is:

[0026] Based on the data anomaly signal, the standard values of all state data obtained within the monitoring period are integrated and processed in time sequence to obtain time series of state data, including vibration intensity sequence, temperature sequence and current sequence;

[0027] Obtain the linear reference value XC, and compare the linear reference value XC with the linear reference threshold value:

[0028] If the linear reference value XC is greater than or equal to the linear reference threshold value, the time series is marked as a linear sequence;

[0029] If the linear reference value XC is less than the linear reference threshold value, the time series is marked as a nonlinear sequence;

[0030] In all time series, the number of linear sequences is counted and compared with the number of time series to obtain a linear number ratio and mark it as XL;

[0031] In all time series, the linear reference value corresponding to the nonlinear sequence is subtracted from the linear reference threshold value, and the absolute value of the difference is taken to obtain the linear reference deviation value corresponding to the nonlinear sequence. The linear reference deviation values corresponding to all nonlinear sequences are summed and averaged to obtain the linear reference deviation mean value, and the linear reference deviation mean value is compared with the linear reference threshold value to obtain a linear deviation ratio and mark it as XP;

[0032] The obtained linear number ratio XL and linear deviation ratio XP are data processed by the formula: to obtain the linear performance value XB, wherein z1 and z2 are both preset proportion coefficients;

[0033] Comparing the linear performance value XB with a linear performance threshold value;

[0034] If the linear performance value XB is greater than the linear performance threshold value, a linear trend signal is generated;

[0035] If the linear performance value XB is less than or equal to the linear performance threshold value, a nonlinear trend signal is generated.

[0036] As a further technical solution of the present application, the linear reference value XC is obtained in the following manner: based on any one time sequence, a sequence change curve is constructed according to the time sequence, and the sequence change curve is divided into a plurality of change sub-curves;

[0037] The two end points of the sequence change curve are connected by a straight line to obtain a change reference line;

[0038] The slope value of the reference line and the slope value of the change sub-curve are measured, and the slope value of the change sub-curve is compared with the slope value of the reference line in terms of positive and negative:

[0039] If the slope value of the change sub-curve and the slope value of the reference line are the same in terms of positive and negative, the change sub-curve is marked as a same-direction sub-curve;

[0040] If the slope value of the change sub-curve and the slope value of the reference line are different in terms of positive and negative, the change sub-curve is marked as a non-same-direction sub-curve;

[0041] The number of same-direction sub-curves is counted and subjected to ratio processing with the number of change sub-curves to obtain the number ratio of same-direction sub-curves and mark it as SL;

[0042] The slope values of all same-direction sub-curves are obtained and integrated into a slope data group, the variance value of the slope data group is calculated, and marked as FC;

[0043] The linear reference value XC is obtained by the formula: wherein a1 and a2 are both preset proportion coefficients.

[0044] As a further technical solution of the present application, based on the linear trend signal, a linear model is selected as a risk prediction model;

[0045] Based on the nonlinear trend signal, a nonlinear model is selected as a risk prediction model.

[0046] As a further technical solution of the present application, based on the risk prediction model, a risk threshold value is input into the risk prediction model to obtain the time when the device state reaches the risk threshold value, and the time when the device state reaches the risk threshold value is subjected to difference processing with the current time to obtain the remaining working time of the device.

[0047] As a further technical solution of the present application, the remaining working time of the equipment is acquired, and the expected use time of the equipment is compared with the remaining working time of the equipment.

[0048] If the expected use time of the equipment is less than or equal to the remaining working time of the equipment, a normal working signal of the equipment is generated.

[0049] If the expected use time of the equipment is greater than the remaining working time of the equipment, it is considered that the equipment needs to take maintenance measures in advance, and a maintenance signal of the equipment is generated.

[0050] The present application has the following beneficial effects:

[0051] (1) The present application quickly and accurately identifies potential equipment failures by collecting state data of the equipment and calculating a comprehensive score through a preset weight coefficient, effectively reduces the frequency of manual inspection and the misjudgment rate, and significantly improves the efficiency of equipment management. Secondly, the equipment risk prediction model is constructed according to the nonlinear trend signal, accurately predicts the future possible failure conditions, makes the maintenance plan in advance, avoids the unplanned shutdown events caused by sudden failures, and further prolongs the overall service life of the equipment.

[0052] (2) The present application constructs a device risk trend curve based on the prediction results, monitors the current and future health status of the equipment in real time, and takes preventive maintenance measures, which significantly reduces the unplanned downtime loss caused by equipment failure, enhances the safety and economy of the production process, and ensures the smooth operation of the production equipment through the mean and standard deviation calculation of the state data of the equipment in each time subunit and the comparative analysis with the threshold value. Under the premise of safety, the present application optimizes the resource allocation and further improves the overall economic benefits of the production process.

[0053] (3) The present application flexibly adapts to different types of equipment and their state data characteristics, selects the most suitable prediction model by judging whether the vibration signal, temperature or current data shows a linear trend, ensures the accuracy of the prediction results, expands the application scope of the system, and at the same time, through real-time state data acquisition and risk trend prediction, and comparing the expected use time and the remaining working time, the present application can dynamically decide whether to take maintenance measures in advance, ensure the continuous operation of the equipment without immediate intervention, or timely maintenance to prolong the service life of the equipment and reduce the risk, thereby realizing the transformation from passive response to active prevention and control, greatly improving the stability and economy of the production system, and having significant application value and social benefits. BRIEF DESCRIPTION OF DRAWINGS

[0054] The present application will be further described below in conjunction with the drawings.

[0055] Figure 1is a specific step flowchart of the model predictive control equipment risk management and control system based on the present application;

[0056] Figure 2 is a flowchart of generating data normal signal and data abnormal signal in the present application. DETAILED DESCRIPTION

[0057] The technical solutions in the embodiments of the present application will be clearly and completely described with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of protection of the present application.

[0058] Embodiment 1

[0059] Please refer to Figure 1 The present application is a model predictive control equipment risk management and control system based on the present application, which comprises a data acquisition and processing module, a data analysis module, a model construction and prediction module, and an analysis and control module.

[0060] The data acquisition and processing module is used for acquiring state data of the equipment.

[0061] The state data of the equipment is integrated and standardized processed.

[0062] According to the standardized state data, the comprehensive score of the equipment is calculated by combining the preset weight coefficient, and the data abnormal signal and the data normal signal are generated.

[0063] The data analysis module further analyzes the change trend of the equipment state data with time based on the data abnormal signal, and generates a linear trend signal and a nonlinear trend signal.

[0064] The model construction and prediction module constructs a risk prediction model based on the linear trend signal and the nonlinear trend signal.

[0065] The analysis and control module obtains the remaining working time of the equipment based on the risk prediction model, and performs comparative analysis to generate an equipment operation signal, wherein the equipment operation signal comprises an equipment normal working signal and an equipment maintenance signal.

[0066] The data acquisition and processing module:

[0067] Function: responsible for acquiring state data such as vibration intensity, temperature and current of the equipment, and performing integration and standardization processing.

[0068] Output: standard value of equipment state data, comprehensive score, data abnormal / normal signal.

[0069] The data analysis module:

[0070] Function: Based on the data anomaly signal, analyze the trend of the device state data over time, and generate a linear or nonlinear trend signal.

[0071] Output: Linear / nonlinear trend signal.

[0072] Model construction prediction module:

[0073] Function: Based on the linear trend signal and the nonlinear trend signal, construct a risk prediction model.

[0074] Output: Risk prediction model.

[0075] Analysis and control module:

[0076] Function: Based on the risk prediction model, obtain the remaining working time of the device, and perform comparative analysis to generate a device operation signal.

[0077] Output: Device operation signal.

[0078] Embodiment 2

[0079] As shown in the following table, this embodiment describes in detail how to collect and process device state data within a monitoring period: Figure 2

[0080] Divide the monitoring period into several time sub-units of equal length, divide the time sub-unit into several time small-units of equal length, and collect the device state data at the middle time point of each time small-unit;

[0081] Wherein, the monitoring period is a preset time period, for example, 1 hour, the time sub-unit of equal length is obtained by equal time division method, the time length of each time sub-unit is a preset length, for example, 1 minute, the time length of the time small-unit is a preset length, for example, 1 second, the middle time point of the time small-unit is the time point when the time small-unit is at the middle time, for example, if the time length of the time small-unit is 1 second, the middle time point of the time small-unit is 0.5 seconds, the device state data includes the vibration intensity value, temperature value and current value of the device;

[0082] Integrate the device state data collected at the middle time point of the time small-unit in each time sub-unit respectively to obtain the vibration intensity value unit group, temperature value unit group and current value unit group in the time sub-unit;

[0083] ​The vibration intensity value unit group, the temperature value unit group and the current value unit group are processed according to a mean value calculation formula respectively to obtain the vibration intensity unit mean value, the temperature unit mean value and the current unit mean value of the equipment in the time subunit, which are respectively recorded as the unit vibration signal value, the unit temperature value and the unit current value;

[0084] It should be noted that in the risk control analysis of the equipment, the equipment may have instantaneous changes, that is, the collected equipment state data is instantaneous data, and the instantaneous data cannot accurately represent the overall state of the equipment. The average value acquisition method has high accuracy for the equipment;

[0085] The unit vibration signal value, the unit temperature value and the unit current value in the monitoring period are integrated and processed to obtain the vibration intensity value group, the temperature value group and the current value group;

[0086] The vibration intensity value group, the temperature value group and the current value group are processed according to a mean value calculation formula respectively to obtain the vibration signal mean value, the temperature mean value and the current mean value of the equipment in the monitoring period;

[0087] The vibration intensity value group, the temperature value group and the current value group are processed according to a standard deviation calculation formula to obtain the vibration intensity standard deviation, the temperature value standard deviation and the current value standard deviation of the equipment in the monitoring period;

[0088] Based on the mean value and the standard deviation of the equipment state data, the values in the vibration intensity value group, the temperature value group and the current value group are respectively subtracted from the corresponding mean value and divided by the standard value to obtain the standard value of the equipment state data;

[0089] The standard value of the equipment state data includes the vibration intensity standard value, the temperature standard value and the current standard value.

[0090] It should be noted that the data standardization is to make the data of different dimensions in the same order of magnitude, prevent the data from being inconsistent, and thus avoid the influence of some features on the model being too large.

[0091] The comprehensive score DF based on the vibration, temperature and current of the equipment is calculated by the formula DF i =w1×WD i +w2×ZD i +w3×DL i ; i ;

[0092] Wherein, w1 is the weight coefficient allocated to the temperature standard value WD, w2 is the weight coefficient allocated to the vibration standard value ZD, w3 is the weight coefficient allocated to the current standard value DL, i is the number of time subunits, i = 1, 2,..., n, WD i is the temperature standard value of the i-th time subunit, ZD i and DLi Similarly;

[0093] The comprehensive score DF is compared with a score threshold value;

[0094] If the comprehensive score is greater than the score threshold value, it is considered that the data of the equipment in the time subunit is in an abnormal condition, and a data abnormal signal is generated;

[0095] If the comprehensive score is less than or equal to the score threshold value, it is considered that the data of the equipment in the time subunit is in a normal state, and a data normal signal is generated;

[0096] It should be noted that in this embodiment, the monitoring period (such as 1 hour) is first divided into several equal-length time subunits (such as 1 minute), each time subunit is further subdivided into shorter time subunits (such as 1 second), and equipment state data is collected at the middle time point (such as 0.5 seconds) of each time subunit. Then, the state data in each time subunit is integrated, and the mean and standard deviation of the vibration intensity, temperature and current are calculated to eliminate the influence of dimension. Finally, a preset weight coefficient (w1, w2, w3) is used to calculate the comprehensive score DF, which is compared with the threshold value to generate a data abnormal / normal signal.

[0097] Embodiment 3

[0098] This embodiment further illustrates how to analyze the trend of the equipment state data according to the data abnormal signal and select a suitable prediction model:

[0099] In one specific embodiment:

[0100] All state data standard values obtained in the monitoring period are integrated in time sequence to obtain a time sequence of state data, including a vibration intensity sequence, a temperature sequence and a current sequence;

[0101] Based on any one time sequence, a sequence change curve is constructed according to the time sequence, and the sequence change curve is divided into several change sub-curves;

[0102] The two end points of the sequence change curve are connected by a straight line to obtain a change reference line;

[0103] The slope value of the reference line and the slope value of the change sub-curve are measured, and the slope value of the change sub-curve is compared with the slope value of the reference line in terms of positive and negative:

[0104] If the slope value of the change sub-curve and the slope value of the reference line are the same in terms of positive and negative, the change sub-curve is marked as a same-direction sub-curve;

[0105] If the slope value of the change sub-curve and the slope value of the reference line are different in terms of positive and negative, the change sub-curve is marked as a non-same-direction sub-curve;

[0106] It should be noted that the same positive and negative indicates that the slope value of the change sub-curve is the same as the slope value of the reference line or the same as the negative;

[0107] The number of the same direction sub-curve is counted, and a ratio processing is performed with the number of the change sub-curve to obtain a number ratio of the same direction sub-curve, and marked as SL;

[0108] The slope values of all the same direction sub-curves are obtained, and integrated into a slope data group, and a variance value of the slope data group is calculated and marked as FC;

[0109] The linear reference value XC is obtained through the formula: , wherein a1 and a2 are both preset proportion coefficients;

[0110] In some embodiments, the linear reference value XC is compared with a linear reference threshold value:

[0111] If the linear reference value XC is greater than or equal to the linear reference threshold value, the time sequence is marked as a linear sequence;

[0112] If the linear reference value XC is less than the linear reference threshold value, the time sequence is marked as a nonlinear sequence;

[0113] In all time sequences, the number of linear sequences is counted, and a ratio processing is performed with the number of time sequences to obtain a linear number ratio, and marked as XL;

[0114] In all time sequences, a difference processing is performed between the linear reference value corresponding to the nonlinear sequence and the linear reference threshold value, and an absolute value is taken to obtain a linear reference deviation value corresponding to the nonlinear sequence, all linear reference deviation values corresponding to the nonlinear sequence are summed and averaged to obtain a linear reference deviation mean value, and a ratio processing is performed with the linear reference threshold value to obtain a linear deviation ratio, and marked as XP;

[0115] The obtained linear number ratio XL and the linear deviation ratio XP are data processed through the formula: to obtain a linear performance value XB, wherein z1 and z2 are both preset proportion coefficients;

[0116] In some embodiments, the linear performance value XB is compared with a linear performance threshold value;

[0117] If the linear performance value XB is greater than the linear performance threshold value, a linear trend signal is generated;

[0118] If the linear performance value XB is less than or equal to the linear performance threshold value, a nonlinear trend signal is generated;

[0119] Based on the linear trend signal, a linear model is selected to predict the future state of the equipment, such as a linear regression model or an autoregressive moving average model;

[0120] Based on the nonlinear trend signal, a nonlinear model is selected to predict the future state of the equipment, such as a random forest model;

[0121] It should be noted that the random forest model is suitable for processing cases with multiple input variables and can automatically screen important features;

[0122] In another specific embodiment, if the standard value of the equipment state data presents a nonlinear trend, a nonlinear trend signal is generated, and a risk assessment model of the equipment is constructed by a random forest model, specifically:

[0123] Based on the nonlinear trend signal, the time series of the equipment state data in the monitoring period is obtained;

[0124] Based on the time series of the equipment state data, a fixed-size sliding window is added to the vibration intensity sequence, temperature sequence, or current sequence;

[0125] The standard value of the state data in the sliding window is processed by moving average to obtain the time series characteristics of the state data in the sliding window, including the mean, standard deviation, maximum value, and minimum value of the state data in the sliding window;

[0126] Statistical feature extraction processing is performed on the vibration intensity sequence, temperature sequence, or current sequence to obtain the statistical features of the standard value of the state data of the equipment in the vibration intensity sequence, temperature sequence, or current sequence, including the mean, median, standard deviation, maximum value, and minimum value of the standard value of the state data;

[0127] Through the formula: The temperature sequence and the vibration intensity sequence are jointly analyzed to obtain the combination characteristics of the joint distribution of temperature and vibration;

[0128] Where Ti is the temperature standard value, Vi is the vibration intensity standard value, T, is the average value of the temperature standard value, V, is the average value of the vibration intensity standard value, and WZL is the correlation coefficient between temperature and vibration, i.e. the joint distribution characteristics of temperature and vibration, i.e. the combination characteristics;

[0129] The time series characteristics, statistical characteristics, and combination characteristics of the equipment state data standard value are integrated to obtain the feature matrix X of the equipment state data standard value;

[0130] A corresponding label y is added to each data point in the feature matrix;

[0131] The feature matrix X and the label y are integrated into a data set D;

[0132] A prediction model is constructed using a random forest regression, and the dataset D is divided into a training set, a validation set and a test set for training the model, to obtain a mature device risk prediction model;

[0133] The feature matrix X is simulated and predicted by the device risk prediction model, to obtain the state change of the device in the future period of time;

[0134] It should be noted that the embodiment provides a risk prediction model for simulating and predicting the risk trend of the device in a certain period of time in the future based on a nonlinear trend, by constructing a nonlinear model. The health state of the device in the future period of time is simulated and predicted, specific risk indicators are provided, the influence of the current device state on the production safety is understood by the decision maker, and a basis is provided for subsequent decision making.

[0135] Embodiment 4

[0136] This embodiment describes how to perform real-time state monitoring and future risk trend prediction based on the risk prediction model:

[0137] In a specific embodiment:

[0138] Based on the risk prediction model, the risk threshold is input into the risk prediction model, to obtain the time when the device state reaches the risk threshold. The time when the device state reaches the risk threshold is processed by subtracting the current time, to obtain the remaining working time length of the device;

[0139] The remaining working time length of the device is obtained, and the expected use time of the device is compared with the remaining working time length of the device;

[0140] If the expected use time of the device is less than or equal to the remaining working time length of the device, it is considered that the device can continue to operate without immediate intervention, and a device normal working signal is generated;

[0141] If the expected use time of the device is greater than the remaining working time length of the device, it is considered that the device needs to take maintenance measures in advance, and a device maintenance signal is generated;

[0142] It should be noted that the expected use time represents the time length that the device needs to continue to use according to the use demand;

[0143] Based on the device normal working signal, the current operation mode is maintained, and the device does not need to be regulated and controlled;

[0144] Based on the device maintenance signal, planned maintenance, replacement of parts, adjustment of operation parameters to reduce load or enhance monitoring frequency and other measures are taken on the device.

[0145] The technical solution of the present application is as follows: first, by simulating the health status of the equipment in the future, specific risk indicators are provided to help understand the impact of the current equipment state on production safety and provide a basis for subsequent decision-making; then, by determining the time required for the equipment to develop from the current state to a high-risk state, a time window is provided during which it is determined whether to take measures to avoid or mitigate risks, ensuring sufficient time to arrange necessary maintenance activities to prevent sudden failures from affecting production; subsequently, by estimating how long the equipment can still operate normally without failure, a scientific basis is provided for equipment maintenance and replacement planning, optimizing asset management, reducing unnecessary maintenance costs, and avoiding failures caused by excessive aging of the equipment; next, by comparing the remaining usage time of the equipment with the length of time required to reach a high-risk state, it can be determined whether action needs to be taken to extend the life of the equipment or reduce risks, ensuring the economic and effective nature of maintenance activities and avoiding over-maintenance or under-maintenance; finally, based on the results of the comparative analysis, it is determined whether and how to adjust the operating conditions of the equipment, extending the service life of the equipment, reducing failure rates, ensuring the continuity and stability of production, improving the overall performance of the equipment, and reducing production costs.

[0146] The working principle of the present application is as follows: by collecting the state data (such as vibration signal, temperature and current) of the equipment, and integrating and standardizing the data, the comprehensive score of the equipment is calculated through the preset weight coefficient, and then the data abnormal signal or data normal signal is generated; based on the data abnormal signal, the changing trend of the equipment state data over time is further analyzed, and the linear trend signal or nonlinear trend signal is generated; according to the nonlinear trend signal, the equipment risk prediction model is constructed, and the future state of the equipment is predicted; based on the prediction result, the equipment risk trend curve is constructed, the expected service life of the equipment is obtained, and the equipment is regulated according to the state signal of the equipment. In addition, the monitoring period is divided into a plurality of equal length time subunits and time small units, the state data of the equipment is collected at the middle time point of each time small unit, and the data is standardized by calculating the mean and standard deviation of the equipment state data in each time subunit, the comprehensive score is calculated by using the weight coefficient and compared with the threshold value, so as to judge whether the equipment state is abnormal. For the data abnormal signal, it is further judged whether the vibration signal, temperature or current data shows a linear trend, and the appropriate prediction model is selected accordingly: when the data shows a linear trend, the linear regression model is used for prediction, and when the data shows a nonlinear trend, the random forest model is used to construct the equipment risk assessment model to predict the future state change of the equipment. Based on the prediction model, the present application predicts the risk trend in the future period of time by collecting the real-time state data of the equipment, and constructs the risk trend curve of the equipment, compares the expected service life of the equipment with the remaining working time, and decides whether to take maintenance measures in advance to ensure that the equipment can continue to run without immediate intervention, or to maintain in time to prolong the service life of the equipment and reduce the risk. In summary, the present application realizes real-time monitoring of equipment state data, intelligent analysis and prediction, and dynamic regulation strategy, which not only improves the equipment management efficiency, but also effectively prolongs the service life of the equipment, reduces the unplanned downtime, and improves the safety and economy of the production process.

[0147] The above describes one embodiment of the present application in detail, but the above description is only a preferred embodiment of the present application, and cannot be considered as limiting the scope of the present application. Any equivalent changes and improvements made within the scope of the present application shall still belong to the scope of the present application.

Claims

1. A risk management and control system for equipment based on model prediction control, characterized in that: It includes data acquisition and processing module, data analysis module, model building and prediction module and analysis and control module; The data acquisition and processing module is used to collect status data of the equipment; Integrate and standardize equipment status data; Based on the standardized status data, the comprehensive score of the device is calculated and the data abnormality signal and data normal signal are generated; The data analysis module further analyzes the change trend of equipment status data over time based on data anomaly signals and generates linear trend signals and nonlinear trend signals; The model building prediction module builds a risk prediction model based on linear trend signals and nonlinear trend signals; All the standard values ​​of the status data obtained during the monitoring period are integrated and processed in chronological order to obtain the time series of the status data; Based on any time series, a sequence change curve is constructed according to the time series, and the sequence change curve is divided into several change sub-curves; Connect the two end points of the sequence change curve with a straight line to obtain the change reference line; If the slope of the changing sub-curve has the same sign as the slope of the reference line, the changing sub-curve is marked as a same-direction sub-curve; If the slope of the changing sub-curve is different from the slope of the reference line in sign, the changing sub-curve is marked as a non-co-directional sub-curve; Count the number of same-direction sub-curves and compare them with the number of changing sub-curves to obtain the number ratio of same-direction sub-curves, which is marked as SL. Obtain the slope values ​​of all the same-direction sub-curves and integrate them into a slope data group. Calculate the variance value of the slope data group and mark it as FC. By formula: Obtain a linear reference value XC, where a1 and a2 are both preset proportional coefficients; If the linear reference value XC is greater than or equal to the linear reference threshold, the time series is marked as a linear series; If the linear reference value XC is less than the linear reference threshold, the time series is marked as a nonlinear series; In all time series, the number of linear series is compared with the number of time series to obtain the linear number ratio, which is marked as XL; The linear reference value corresponding to the nonlinear sequence is subtracted from the linear reference threshold and the absolute value is taken to obtain the linear reference deviation value. The linear reference deviation mean is obtained by summing up all the linear reference deviation values ​​and taking the average value. The linear reference deviation mean is then compared with the linear reference threshold to obtain the linear deviation ratio, which is marked as XP. By formula: Obtain the linear performance value XB, where z1 and z2 are both preset proportional coefficients; If the linear performance value XB is greater than the linear performance threshold, a linear trend signal is generated and the linear regression model or the autoregressive moving average model is selected; If the linear performance value XB is less than or equal to the linear performance threshold, a nonlinear trend signal is generated and a random forest model is generated; The analysis and control module obtains the remaining working time of the equipment based on the risk prediction model, performs comparative analysis, and generates equipment operation signals, where the equipment operation signals include normal equipment operation signals and equipment maintenance signals; Based on nonlinear trend signals, the time series of equipment status data within the monitoring period is obtained; Based on the time series of device status data, a fixed-size sliding window is added to the vibration intensity series, temperature series, or current series. Perform sliding average processing on the standard value of the state data in the sliding window to obtain the time series characteristics of the state data in the sliding window, including the mean, standard deviation, maximum and minimum values ​​of the state data in the sliding window; Performing statistical feature extraction processing on the vibration intensity sequence, temperature sequence, or current sequence to obtain statistical features of the state data standard values ​​of the equipment in the vibration intensity sequence, temperature sequence, or current sequence, including the mean, median, standard deviation, maximum value, and minimum value of the state data standard values; By formula: The temperature series and vibration intensity series are jointly analyzed to obtain the combined characteristics of the joint distribution of temperature and vibration; Where Ti is the standard value of temperature, Vi is the standard value of vibration intensity, T, is the average value of the standard value of temperature, V, is the average value of the standard value of vibration intensity, and WZL is the correlation coefficient between temperature and vibration, that is, the joint distribution characteristics of temperature and vibration, that is, the combined characteristics; The time series characteristics, statistical characteristics and combined characteristics of the standard value of the equipment status data are integrated to obtain the characteristic matrix X of the standard value of the equipment status data; Add a corresponding label y to each data point in the feature matrix; Integrate the feature matrix X and label y into the dataset D; A prediction model was constructed using random forest regression, and the dataset D was divided into training set, validation set, and test set for model training, resulting in a mature equipment risk prediction model.

2. The risk management and control system for equipment based on model predictive control according to claim 1, characterized in that: The status data of the equipment includes vibration intensity value, temperature value and current value; Define the monitoring cycle and subdivide it into multiple equal-length time subunits and smaller time units, and collect device status data at the middle point of each time unit; The data collected in each time sub-unit are integrated and processed to calculate the mean value of each state data in the time period, which is recorded as the state data value; Based on the status data values ​​in each time subunit, the mean and standard deviation of the status data of the equipment in the monitoring period are calculated.

3. The risk management and control system for equipment based on model predictive control according to claim 2, characterized in that: Based on the mean and standard deviation of each status data of the equipment during the monitoring period, the mean of each status data corresponding to the monitoring period is subtracted from the value of each status data in the time subunit, and then divided by the standard deviation of each status data corresponding to the monitoring period to obtain the standard value of each status data in each time subunit; Among them, the standard values ​​of various status data in the time sub-unit include vibration intensity standard value, temperature standard value and current standard value.

4. The risk management and control system for equipment based on model predictive control according to claim 3, characterized in that: Integrate the standard values ​​of each state data of all time sub-units obtained during the monitoring period, and use the formula Calculate the comprehensive score based on equipment vibration, temperature and current in each time subunit ; Among them, w1 is the weight coefficient assigned to the temperature standard value WD, w2 is the weight coefficient assigned to the vibration standard value ZD, w3 is the weight coefficient assigned to the current standard value DL, i is the number of time subunits, i=1,2,...,n, is the standard temperature value of the i-th time subunit, Same thing.

5. The risk management and control system for equipment based on model predictive control according to claim 4, characterized in that: Based on the comprehensive score , compare the comprehensive score DF with the score threshold; If the comprehensive score is greater than the score threshold, a data anomaly signal is generated; If the comprehensive score is less than or equal to the score threshold, a data normal signal is generated.

6. The risk management and control system for equipment based on model predictive control according to claim 1, characterized in that: The time series of status data includes vibration intensity series, temperature series and current series.

7. The risk management and control system for equipment based on model predictive control according to claim 6, characterized in that: Based on the risk prediction model, the risk threshold is input into the risk prediction model to obtain the time when the equipment status reaches the risk threshold. The time when the equipment status reaches the risk threshold is subtracted from the current time to obtain the remaining working time of the equipment.

8. The risk management and control system for equipment based on model predictive control according to claim 7, characterized in that: Obtain the remaining working time of the device and compare the expected usage time of the device with the remaining working time of the device; If the expected usage time of the device is less than or equal to the remaining working time of the device, a normal working signal of the device is generated; If the expected usage time of the equipment is greater than the remaining working time of the equipment, it is considered that the equipment needs to take maintenance measures in advance and an equipment maintenance signal is generated.

Images