Multi-source production data fusion processing method based on industrial information model

Abstract

This application provides a method for multi-source production data fusion processing based on an industrial information model, belonging to the field of data processing technology. The method includes: simulating equipment operation based on the processing product sequence, material attribute sequence, and process type sequence of a CNC machining equipment within a future preset time zone to obtain the equipment operating load sequence; performing equipment fault early warning timeliness analysis to determine the appropriate early warning timeliness coefficient, and setting an appropriate monitoring parameter set based on the appropriate early warning timeliness coefficient; filtering multi-source production monitoring data according to the appropriate monitoring parameter set to obtain an appropriate production monitoring data sequence set; extracting abnormal data to obtain an abnormal production monitoring data sequence set, and calculating the average proportion of abnormal data; and calling an equipment fault prediction plugin to perform fault prediction and fault early warning based on the abnormal production monitoring data sequence set. This solves the technical problem in existing technologies where it is difficult to effectively integrate multi-source data and the accuracy of equipment fault early warning is insufficient.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of data processing, and in particular to a multi-source production data fusion processing method based on an industrial information model. BACKGROUND

[0002] In industrial production processes, the operation of devices such as numerical control machining generates multi-source production data such as operation parameters, machining processes, and process parameters. These data sources are scattered, have various formats, and often contain redundant information and noise.

[0003] However, traditional data processing methods are difficult to effectively integrate multi-source data, resulting in insufficient accuracy of device fault early warning, and are difficult to meet the needs of industrial production for efficient and accurate data analysis and decision support.

[0004] Therefore, there is an urgent need for a multi-source production data fusion processing method based on an industrial information model to achieve systematic integration and accurate processing of multi-source production data. SUMMARY

[0005] The present application provides a multi-source production data fusion processing method based on an industrial information model to address the technical problem of insufficient accuracy of device fault early warning in the prior art.

[0006] The technical solution of the present application to solve the above technical problem is as follows:

[0007] The present application provides a multi-source production data fusion processing method based on an industrial information model, comprising:

[0008] Simulating device operation according to the machining product sequence, material attribute sequence, and process type sequence of the numerical control machining device in a future preset time zone to obtain a device operation load sequence;

[0009] Performing device fault early warning timeliness analysis based on the device operation load sequence and process type sequence to determine an adaptive early warning timeliness coefficient, and setting an adaptive monitoring parameter set based on the adaptive early warning timeliness coefficient;

[0010] In the monitoring time window of the preset time zone, filtering multi-source production monitoring data according to the adaptive monitoring parameter set to obtain an adaptive production monitoring data sequence set;

[0011] Extracting abnormal data from the adaptive production monitoring data sequence set to obtain an abnormal production monitoring data sequence set, and calculating an abnormal data proportion mean value;

[0012] Calling a device fault prediction plug-in based on the adaptive monitoring parameter set and abnormal data proportion mean value, and performing fault prediction and fault early warning based on the abnormal production monitoring data sequence set.

[0013] The beneficial effects of the present application are:

[0014] Compared with the prior art, firstly, according to the machining product sequence, material attribute sequence and process type sequence of the numerical control machining equipment in the future preset time zone, the equipment running simulation is carried out, the equipment running load sequence is obtained, the load pressure distribution of the equipment in the future preset time zone can be predicted in advance, which provides reliable data support for subsequent steps of determining the adaptive early warning time coefficient, setting the monitoring parameter and the like. Secondly, according to the equipment running load sequence and the process type sequence, the adaptive early warning time coefficient is determined by analyzing the equipment fault early warning timeliness, and the adaptive monitoring parameter set is set based on the adaptive early warning time coefficient, which realizes the dynamic adaptation of the equipment fault early warning emergency degree and the monitoring parameter quantity, and improves the early warning response efficiency and accuracy. Thirdly, in the monitoring time window of the preset time zone, the adaptive monitoring parameter set is used to filter the multi-source production monitoring data to obtain the adaptive production monitoring data sequence set, which simplifies the redundant information in the multi-source production monitoring data, guarantees the time continuity and parameter pertinence of the data, and provides efficient and targeted data source support for subsequent accurate equipment fault early warning. Further, the abnormal production monitoring data sequence set is obtained by extracting the abnormal data from the adaptive production monitoring data sequence set, and the average of the abnormal data proportion is calculated, the abnormal information in the adaptive production monitoring data is accurately extracted, and the overall abnormal level of the adaptive production monitoring is quantified, which provides reliable data support for the accurate judgment of the subsequent equipment fault risk. Finally, based on the adaptive monitoring parameter set and the average of the abnormal data proportion, the equipment fault prediction plug-in is called, the fault prediction and fault early warning are carried out according to the abnormal production monitoring data sequence set, which improves the accuracy of fault prediction and the reliability of early warning, and realizes the accurate judgment and timely early warning of equipment fault.

[0015] Through the above technical scheme, the present application realizes efficient integration and accurate processing of multi-source production data, effectively improves the timeliness, accuracy and reliability of equipment fault early warning, and provides strong support for stable operation of numerical control machining equipment and efficient decision-making of industrial production. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 The flowchart of the multi-source production data fusion processing method based on the industrial information model provided by the present application is shown.

[0017] Figure 2 The flowchart of the adaptive equipment fault prediction plug-in construction in the multi-source production data fusion processing method based on the industrial information model provided by the present application is shown. DETAILED DESCRIPTION

[0018] With reference to the accompanying drawings, the technical solutions in the embodiments of the present application will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts fall within the scope of the present application.

[0019] In the description of the present application, the terms "first", "second" are used only for the purpose of description, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first", "second" can explicitly or implicitly include one or more of the features. In the description of the present application, the meaning of "a plurality of" is two or more, unless otherwise explicitly and specifically limited.

[0020] In the description of the present application, the term "for example" is used to indicate "as an example, illustration or explanation". Any embodiment described as "for example" in the present application is not necessarily interpreted as more preferred or more advantageous than other embodiments. The following description is given in order to enable any person skilled in the art to implement and use the present application. In the following description, details are listed for the purpose of explanation. It should be understood that those skilled in the art can recognize that the present application can be implemented without using these specific details. In other examples, well-known structures and processes will not be described in detail in order to avoid unnecessary details making the description of the present application obscure. Therefore, the present application is not intended to be limited to the shown embodiments, but is consistent with the broadest scope of the principles and characteristics disclosed.

[0021] As shown in the embodiments, the embodiments of the present application provide a multi-source production data fusion processing method based on an industrial information model, comprising: Figure 1

[0022] S10: Simulating the equipment operation according to the machining product sequence, the material attribute sequence and the process type sequence of the numerical control machining equipment in the future preset time zone, and obtaining the equipment operation load sequence.

[0023] In the production process of the numerical control machining equipment, the differences in machining products, material attributes and process types will cause dynamic changes in equipment load, and the equipment load state is directly related to the fault risk.

[0024] However, the traditional method is difficult to predict the future operation load of the equipment in advance, resulting in lack of pertinence in subsequent fault warning and monitoring, and therefore it is necessary to obtain the load sequence by simulating the future operation state of the equipment, so as to provide a reliable data basis for accurately formulating a warning strategy.

[0025] ​To solve the above problems, the application first simulates the equipment operation according to the machining product sequence, material attribute sequence and process type sequence of the numerical control machining equipment in the future preset time zone to obtain the equipment operation load sequence. The future preset time zone is a production time range set in advance, such as 12 hours, 1 shift, etc., which is the time boundary of subsequent equipment operation simulation, and the person skilled in the art can dynamically set it according to the actual situation.

[0026] The machining product sequence refers to the list and processing order of all products that need to be processed by the equipment in the future preset time zone, such as processing part A from 9:00 to 10:30 and part B from 10:30 to 12:00, etc. The machining product sequence determines the processing task type and processing rhythm of the equipment.

[0027] The material attribute sequence refers to the material characteristic parameters corresponding to the machining product sequence, such as material hardness, density, size tolerance, surface roughness requirement, etc. For example, part A is processed with 45 steel with a hardness of HB200, and part B is processed with aluminum alloy with a hardness of HB80, etc. The material attribute directly affects the processing difficulty and equipment load.

[0028] The process type sequence refers to the specific process steps and types corresponding to each type of product processing, such as turning, milling, drilling, grinding, tool changing, etc. For example, the process of part A is: turning outside circle → milling key groove → drilling → grinding. Different processes have significant differences in equipment operating parameters and equipment load.

[0029] For example, if the future preset time zone is determined to be 12 hours according to the production plan arrangement, the machining product sequence, material attribute sequence and process type sequence within the next 12 hours are obtained, and then the future production operation process of the equipment is simulated in the computer through industrial information models such as digital twin model and equipment simulation model. During the simulation process, the specific operating state of the equipment when processing different products, handling different materials and executing different processes is restored. Finally, the load data of the equipment at each time node within the next 12 hours is output, such as 9:00 load 60%, 9:10 load 75%, 9:20 load 70%, …, 21:00 load 55%, forming the equipment operation load sequence, which intuitively reflects the load peak, trough and stable state of the equipment at different production periods, and clearly presents the dynamic change rule of the load.

[0030] It should be noted that the industrial information model used for equipment operation simulation is a known technology in the art, and the person skilled in the art can select or construct it according to the actual application scenario, and the application will not be described in detail.

[0031] To sum up, compared with the prior art, the application simulates the equipment operation according to the machining product sequence, material attribute sequence and process type sequence of the numerical control machining equipment in the future preset time zone, and obtains the equipment operation load sequence. In this way, through the equipment operation load sequence, the load pressure distribution of the equipment in the future preset time zone can be predicted in advance, and reliable data support is provided for subsequent steps of determining the adaptive early warning timeliness coefficient, setting the monitoring parameter, etc.

[0032] S20: determining an adaptive early warning timeliness coefficient according to the equipment operation load sequence and the process type sequence, and setting an adaptive monitoring parameter set based on the adaptive early warning timeliness coefficient.

[0033] In industrial production, traditional equipment failure early warning often uses fixed monitoring parameters and unified early warning response time limit, which is difficult to adapt to the difference in failure risk under different equipment operation load states and process rhythm, and is prone to problems such as early warning lag or data processing redundancy.

[0034] In view of the above problems, the application determines an adaptive early warning timeliness coefficient according to the equipment operation load sequence and the process type sequence, and sets an adaptive monitoring parameter set based on the adaptive early warning timeliness coefficient.

[0035] Specifically, step S20 in the method comprises:

[0036] calculating the equipment load mean and the equipment load fluctuation coefficient from the equipment operation load sequence, wherein the equipment load fluctuation coefficient is the ratio of the equipment load standard deviation of the equipment operation load sequence to the equipment load mean;

[0037] calculating and obtaining the process switching frequency according to the process type sequence;

[0038] setting the ratio of the preset standard equipment load to the equipment load mean as a first timeliness compensation coefficient;

[0039] setting the ratio of the preset standard equipment load fluctuation coefficient to the equipment load fluctuation coefficient as a second timeliness compensation coefficient;

[0040] setting the ratio of the preset standard process switching frequency to the process switching frequency as a third timeliness compensation coefficient;

[0041] weighting and fusing the first timeliness compensation coefficient, the second timeliness compensation coefficient and the third timeliness compensation coefficient to obtain a comprehensive timeliness compensation coefficient;

[0042] multiplying the comprehensive timeliness compensation coefficient and the preset standard early warning time limit to obtain an adaptive early warning timeliness coefficient.

[0043] In the embodiments of the present application, first, the device load mean value and the device load fluctuation coefficient are calculated according to the device running load sequence. The device load fluctuation coefficient is the ratio of the device load standard deviation of the device running load sequence to the device load mean value. The larger the device load standard deviation, the more intense the load fluctuation, and the larger the device load fluctuation coefficient, the higher the device failure risk. For example, using the device running load sequence in step S10, first, the device load mean value is calculated, for example, device load mean value = (60% + 75% + 70% + … + 55%) / (12 x 6) = 70.83%, and then the device load standard deviation of the device running load sequence is calculated, for example, 7.14%, and then the device load fluctuation coefficient = 7.14% / 70.83% = 0.1008, which quantifies the load stability of the numerical control machining device in the future preset time zone.

[0044] Secondly, the process switching frequency is calculated according to the process type sequence. For example, the total switching times in the future preset time zone can be counted from the process type sequence, and then the process switching frequency is calculated in combination with the time length of the future preset time zone, which reflects the degree of compact production rhythm. For example, in the future 12 hours, the total process switching times are 6 times, and then the process switching frequency = 6 times / 12 hours = 0.5 times / hour.

[0045] Thirdly, the ratio of the preset standard device load to the device load mean value is set as the first time compensation coefficient. The preset standard device load refers to the ideal load benchmark value calibrated in combination with the design parameters of the numerical control machining device, the industry operation specification and the historical optimal operation data, which can be flexibly set by the person skilled in the art according to the device model, the processing scene and the production demand. For example, if the preset standard device load is 80% and the device load mean value is 70.83%, then the first time compensation coefficient = 80% / 70.83% ≈ 1.129. The first time compensation coefficient reflects the deviation degree of the actual load of the device from the standard load. The smaller the first time compensation coefficient, the closer the actual load to the standard load, the greater the device running load pressure, and the higher the failure risk, so the emergency response level of the failure warning needs to be further improved.

[0046] Further, a ratio of the preset standard equipment load fluctuation coefficient and the equipment load fluctuation coefficient is set as a second aging compensation coefficient. The preset standard equipment load fluctuation coefficient refers to a standard load fluctuation reference value calibrated based on the running characteristics of the numerical control machining equipment, the industry stable running standard, and historical failure-free running data. A person skilled in the art can reasonably set the preset standard equipment load fluctuation coefficient according to the equipment type, the machining process requirement, and the production stability requirement. For example, if the preset standard equipment load fluctuation coefficient is 0.08 and the equipment load fluctuation coefficient is 0.1008, the second aging compensation coefficient = 0.08 / 0.1008 ≈ 0.793. The second aging compensation coefficient reflects the deviation degree of the actual load fluctuation state of the equipment from the standard stable state. The smaller the second aging compensation coefficient is, the more intense the actual load fluctuation is, the worse the equipment running stability is, and the higher the probability of failure is, so the emergency response level of the failure warning needs to be further improved.

[0047] Further, a ratio of the preset standard process switching frequency and the process switching frequency is set as a third aging compensation coefficient. The preset standard process switching frequency refers to a standard process switching reference frequency calibrated according to the production efficiency requirement of the numerical control machining equipment, the process optimization standard, and historical stable production data. A person skilled in the art can flexibly set the preset standard process switching frequency according to the product machining process, the equipment working condition, and the production plan. For example, if the preset standard process switching frequency is 0.3 times / hour and the process switching frequency is 0.5 times / hour, the third aging compensation coefficient = 0.3 times / hour / 0.5 times / hour = 0.6. The third aging compensation coefficient reflects the deviation degree of the actual process switching rhythm of the equipment from the standard smooth rhythm. The smaller the third aging compensation coefficient is, the more frequent the actual process switching is, the more intense the equipment working condition changes, and the higher the failure induction risk is, so the emergency response level of the failure warning needs to be further improved.

[0048] Further, the first aging compensation coefficient, the second aging compensation coefficient, and the third aging compensation coefficient are weighted and fused to obtain a comprehensive aging compensation coefficient. The weights of the first aging compensation coefficient, the second aging compensation coefficient, and the third aging compensation coefficient can be dynamically set according to the importance of the three coefficients to the equipment failure risk, and the sum of the weights of the three coefficients is 1. For example, the weights of the first aging compensation coefficient, the second aging compensation coefficient, and the third aging compensation coefficient are set as 0.3, 0.4, and 0.3, respectively. A person skilled in the art can flexibly adjust the weights according to the actual production scene and the equipment characteristics.

[0049] Exemplarily, if the first timeliness compensation coefficient is 1.129, the second timeliness compensation coefficient is 0.793, the third timeliness compensation coefficient is 0.6, and the weights of the three are 0.3, 0.4, and 0.3 respectively, then the comprehensive timeliness compensation coefficient = 1.129*0.3 + 0.793*0.4 + 0.6*0.3 = 0.8359. The comprehensive timeliness compensation coefficient comprehensively reflects the influence of the equipment load deviation, the load stability deviation, and the process switching rhythm deviation, and can comprehensively reflect the comprehensive demand of the overall operation state of the equipment on the timeliness of fault early warning.

[0050] Finally, the product of the comprehensive timeliness compensation coefficient and the preset standard warning time limit is taken as the adaptive warning timeliness coefficient. The preset standard warning time limit refers to the basic warning response time benchmark standardized in combination with the fault handling period of the numerical control machining equipment, industry safety specifications, and production interruption loss evaluation. Those skilled in the art can flexibly set it according to the equipment type, processing importance, and fault influence range. Exemplarily, if the preset standard warning time limit is 1 hour and the comprehensive timeliness compensation coefficient is 0.8359, then the adaptive warning timeliness coefficient = 1 hour*0.8359 ≈ 0.84 hours, about 50 minutes, indicating that the actual operation state of the equipment leads to a higher fault risk than the ideal state, and the warning response time needs to be shortened. The adaptive warning timeliness coefficient comprehensively reflects the emergency degree demand of the overall operation state of the equipment on the fault early warning. The smaller the adaptive warning timeliness coefficient is, the higher the warning emergency degree is, and the risk positioning needs to be completed quickly.

[0051] Further, the step S20 in the method further comprises:

[0052] obtaining a preset monitoring parameter set, wherein the preset monitoring parameter set comprises equipment operation monitoring parameters, processing process monitoring parameters, and processing technology parameters;

[0053] based on the historical operation log of the same type of numerical control machining equipment, respectively evaluating the correlation degrees of a plurality of preset monitoring parameters in the preset monitoring parameter set and equipment faults, and arranging the plurality of preset monitoring parameters in descending order of correlation degree to generate a preset monitoring parameter sequence;

[0054] based on the adaptive warning timeliness coefficient, matching to obtain an adaptive monitoring index number, wherein the adaptive monitoring index number and the adaptive warning timeliness coefficient are positively correlated;

[0055] selecting from the front to the back in the preset monitoring parameter sequence according to the adaptive monitoring index number to obtain an adaptive monitoring parameter set.

[0056] In the embodiments of the present application, a preset monitoring parameter set is first acquired. The preset monitoring parameter set includes device operation monitoring parameters, processing process monitoring parameters and processing parameters: the device operation monitoring parameters can reflect the parameters of the device running state, such as spindle speed, motor current, bearing temperature, lubricating oil pressure, cooling system flow, etc.; the processing process monitoring parameters can reflect the parameters of the real-time state of the processing operation, such as cutting force, machining vibration amplitude, workpiece surface temperature, tool cutting sound frequency, etc.; the processing parameters refer to the set parameters related to the processing scheme, such as feed amount, cutting depth, tool type and wear amount, machining precision deviation, etc.; the three types of parameters can capture the device running state from different dimensions. Exemplarily, the preset monitoring parameter set of a certain numerical control milling machine includes 20 parameters covering device operation monitoring parameters, processing process monitoring parameters and processing parameters, such as bearing temperature, motor current, cutting force, feed amount, etc., providing a comprehensive parameter pool for subsequent screening.

[0057] Secondly, based on the historical running logs of the same type of numerical control machining equipment, the correlation degrees of a plurality of preset monitoring parameters in the preset monitoring parameter set and the equipment failure are respectively evaluated, and the plurality of preset monitoring parameters are arranged in descending order of correlation degrees to generate a preset monitoring parameter sequence. Exemplarily, based on the historical running logs of the same type of numerical control machining equipment, the number of times of equipment failure within a preset time (such as 1 hour) after the preset monitoring parameter appears abnormal can be counted, and then the ratio of the number of times of equipment failure to the total number of times of abnormality of the preset monitoring parameter is calculated as the correlation degree of the preset monitoring parameter and the equipment failure. The higher the correlation degree, the higher the probability of equipment failure after the abnormality of the preset monitoring parameter, and the greater the reference value for failure warning. After the correlation degree evaluation is completed, all preset monitoring parameters are arranged in descending order of correlation degrees to generate a preset monitoring parameter sequence.

[0058] For example, through the historical running logs of the same type of numerical control machining equipment, it is calculated that the failure rate within 1 hour after the abnormality of the bearing temperature is 85%, the failure rate within 1 hour after the abnormality of the motor current is 80%, …, and the failure rate within 1 hour after the abnormality of the environmental humidity is 5%. Therefore, the preset monitoring parameter sequence can be represented as: bearing temperature, motor current, …, environmental humidity. The earlier the preset monitoring parameter in the preset monitoring parameter sequence, the higher the correlation degree with the equipment failure and the greater the reference value for failure warning, which should be selected preferentially.

[0059] Again, the number of adaptive monitoring indicators is obtained based on matching the adaptive early warning timeliness coefficient. The number of adaptive monitoring indicators and the adaptive early warning timeliness coefficient are positively correlated. The greater the adaptive early warning timeliness coefficient, the more time the early warning response has, the lower the risk of the equipment, and the more adaptive monitoring indicators that can be configured to cover potential risks as comprehensively as possible. Conversely, the smaller the adaptive early warning timeliness coefficient, the more urgent the early warning response time, and the fewer adaptive monitoring indicators that should be configured to improve response efficiency.

[0060] For example, based on the device failure history data, the importance of the production task, and the real-time cost of parameter monitoring, a person skilled in the art can construct a preset mapping rule for the adaptive early warning timeliness coefficient. For example, when the adaptive early warning timeliness coefficient is ≤0.6 hours (extremely urgent), 40% of the total number of preset monitoring parameters in the preset monitoring parameter sequence is selected; when the adaptive early warning timeliness coefficient is between 0.6 hours and 0.9 hours (urgent), 50% of the total number is selected; when the adaptive early warning timeliness coefficient is between 0.9 hours and 1.2 hours (moderate), 75% of the total number is selected; and when the adaptive early warning timeliness coefficient is >1.2 hours (extremely moderate), 100% of the total number is selected.

[0061] For example, if the adaptive early warning timeliness coefficient is 0.84 hours and the preset monitoring parameter sequence contains a total of 20 parameters, according to the above preset mapping rule, it is determined that 50% of the total number of preset monitoring parameters in the preset monitoring parameter sequence should be selected, so the number of adaptive monitoring indicators = 20 x 50% = 10, thereby achieving accurate adaptation of the early warning urgency and the number of monitoring indicators.

[0062] Finally, the adaptive monitoring parameter set is obtained by selecting from the front to the back in the preset monitoring parameter sequence according to the number of adaptive monitoring indicators. The reason for selecting from the front to the back in the preset monitoring parameter sequence is that the preset monitoring parameter sequence is sorted by correlation, and the earlier the preset monitoring parameter, the higher the correlation with the device failure and the greater the reference value for failure warning, which should be selected first. For example, if the number of adaptive monitoring indicators is 10, 10 preset monitoring parameters such as bearing temperature and motor current are selected from the front to the back in the preset monitoring parameter sequence to form the adaptive monitoring parameter set.

[0063] In summary, compared with the prior art, the present application determines the adaptive early warning timeliness coefficient based on the device operation load sequence and the process type sequence, and sets the adaptive monitoring parameter set based on the adaptive early warning timeliness coefficient. In this way, the dynamic adaptation of the device failure early warning urgency and the number of monitoring parameters is achieved, and the early warning response efficiency and accuracy are improved.

[0064] S30: filtering, in a monitoring time window of the preset time zone, the multi-source production monitoring data according to the adaptive monitoring parameter set to obtain an adaptive production monitoring data sequence set.

[0065] The multi-source production monitoring data obtained by the foregoing steps includes all parameters in the preset monitoring parameter sequence, and the data amount is large and mixed with redundant information with low correlation with equipment failure. The adaptive monitoring parameter set is a core parameter set with high correlation with equipment failure, which is filtered based on the adaptive early warning timeliness coefficient and can accurately match the actual demand for failure early warning.

[0066] Therefore, the application filters, in a monitoring time window of the preset time zone, the multi-source production monitoring data according to the adaptive monitoring parameter set to obtain an adaptive production monitoring data sequence set. The monitoring time window is the smallest data collection unit obtained by dividing the preset time zone by a fixed time interval, such as 10 minutes, which can be dynamically adjusted by a person skilled in the art according to the equipment operation characteristics, monitoring accuracy requirements and data processing capacity.

[0067] Exemplarily, in a 12-hour preset time zone, 72 continuous monitoring time windows are divided according to a fixed time interval of 10 minutes, and the multi-source production monitoring data collected in the preset time zone is filtered according to the adaptive monitoring parameter set. In each monitoring time window, only the data in the adaptive parameter set such as bearing temperature and motor current are retained, and redundant parameters with low correlation with equipment failure are removed to obtain an adaptive production monitoring data sequence set. The adaptive production monitoring data sequence set not only ensures the time continuity of the data and can clearly reflect the parameter change trend over time, but also reduces data redundancy by simplifying the parameters, thereby providing efficient and targeted data source support for subsequent accurate equipment failure early warning.

[0068] In summary, compared with the prior art, the application filters, in a monitoring time window of the preset time zone, the multi-source production monitoring data according to the adaptive monitoring parameter set to obtain an adaptive production monitoring data sequence set. In this way, the redundant information in the multi-source production monitoring data is simplified, the time continuity of the data and the parameter targeting are ensured, and efficient and targeted data source support is provided for subsequent accurate equipment failure early warning.

[0069] S40: extracting abnormal data from the adaptive production monitoring data sequence set to obtain an abnormal production monitoring data sequence set, and calculating an average abnormal data proportion.

[0070] In industrial production monitoring, the adaptive production monitoring data sequence set has removed redundant information, but still contains mixed information of normal fluctuation data and abnormal data, which is easy to be disturbed by normal data when directly used for failure early warning.

[0071] Meanwhile, abnormal data of a single parameter is difficult to reflect the abnormal degree of the overall operation state of the equipment, and if there is a lack of systematic extraction of abnormal data and quantitative means of overall abnormal level, it may lead to false alarm or missed alarm of fault.

[0072] To solve the above problems, the application extracts abnormal production monitoring data sequence set from the adaptive production monitoring data sequence set, and calculates the average proportion of abnormal data.

[0073] Specifically, step S40 in the method comprises:

[0074] Randomly selecting any adaptive production monitoring data sequence in the adaptive production monitoring data sequence set as a to-be-processed adaptive production monitoring data sequence;

[0075] Selecting the first data in the to-be-processed adaptive production monitoring data sequence as the first abnormal data, and selecting the adjacent data of the first abnormal data as the second monitoring data;

[0076] Calculating the deviation of the first abnormal data and the second monitoring data to obtain the first data deviation, if the first data deviation is greater than a preset deviation threshold, setting the second monitoring data as the second abnormal data, and continuing the data deviation calculation and abnormal data selection from the second abnormal data as the starting point;

[0077] If the first data deviation is less than or equal to the preset deviation threshold, discarding the second monitoring data, and continuing the data deviation calculation and abnormal data selection from the first abnormal data as the starting point, until the to-be-processed adaptive production monitoring data sequence is traversed, outputting the first abnormal data sequence as the first abnormal production monitoring data sequence, and adding it to the abnormal production monitoring data sequence set.

[0078] In the embodiment of the application, first, any adaptive production monitoring data sequence in the adaptive production monitoring data sequence set is randomly selected as a to-be-processed adaptive production monitoring data sequence. For example, if the adaptive production monitoring data sequence set contains 10 adaptive production monitoring parameters such as bearing temperature and motor current, a adaptive production monitoring data sequence corresponding to the bearing temperature, such as [35℃, 42℃, 43℃, 36℃, …, 40℃], is randomly selected as the to-be-processed adaptive production monitoring data sequence.

[0079] Secondly, the first data in the to-be-processed adaptive production monitoring data sequence is selected as the first abnormal data, and the adjacent data of the first abnormal data is selected as the second monitoring data. For example, the first data in the to-be-processed adaptive production monitoring data sequence, such as 35℃, is selected as the first abnormal data, and the adjacent data of the first abnormal data, such as 42℃, is selected as the second monitoring data.

[0080] Again, the deviation calculation is performed on the first abnormal data and the second monitoring data to obtain a first data deviation, if the first data deviation is greater than a preset deviation threshold, the second monitoring data is set as the second abnormal data, and the data deviation calculation and the abnormal data selection are continued from the second abnormal data as a starting point. Optionally, the deviation calculation can be performed by calculating the absolute difference value of the first abnormal data and the second monitoring data as the first data deviation. The preset deviation threshold is a critical value set according to the reasonable fluctuation range of the parameter when the equipment is normally running, for example, the preset deviation threshold of the bearing temperature can be set to 5℃. For example, if the first abnormal data is 35℃ and the second monitoring data is 42℃, the first data deviation is 7℃, which is greater than the preset deviation threshold 5℃, and therefore 42℃ is set as the second abnormal data, and subsequently the data deviation is calculated from 42℃ as a starting point and the adjacent data 43℃.

[0081] Finally, if the first data deviation is less than or equal to the preset deviation threshold, the second monitoring data is discarded, and the data deviation calculation and the abnormal data selection are continued from the first abnormal data as a starting point, until the processing of the adaptive production monitoring data sequence is completed, the first abnormal data sequence is output as the first abnormal production monitoring data sequence, and is added to the abnormal production monitoring data sequence set. The first abnormal data sequence is a sequence composed of all continuous abnormal data extracted from the processing adaptive production monitoring data sequence; the abnormal production monitoring data sequence set contains all abnormal production monitoring data sequences, such as bearing temperature abnormal sequence, motor current abnormal sequence, etc.

[0082] For example, after the processing of the adaptive production monitoring data sequence is completed, the first abnormal data sequence [35℃, 42℃, 36℃, …] is obtained as the first abnormal production monitoring data sequence, which is added to the abnormal production monitoring data sequence set.

[0083] Further, the step S40 in the method further comprises:

[0084] The abnormal data proportion of each adaptive monitoring parameter in the abnormal production monitoring data sequence set is counted, and the mean value of the abnormal data proportions is calculated.

[0085] In the embodiment of the application, the abnormal data proportion of each adaptive monitoring parameter in the abnormal production monitoring data sequence set is counted first, that is, the ratio of the abnormal data amount of each adaptive monitoring parameter to the total data amount of the parameter is calculated, and the abnormal data proportion of the adaptive monitoring parameter can quantify the abnormal frequency of each adaptive monitoring parameter.

[0086] For example, if the adaptive monitoring parameter "bearing temperature" has a total of 72 window data in 12 hours, and the abnormal data sequence contains 18 abnormal data, the abnormal data proportion of the bearing temperature = 18 / 72 = 25%.

[0087] Secondly, the proportion of abnormal data of each adaptive monitoring parameter in the abnormal production monitoring data sequence set is counted according to the same logic and method, and then the average value is calculated to obtain the average value of the proportion of abnormal data.

[0088] For example, if there are 10 adaptive monitoring parameters, and the proportion of abnormal data of each adaptive monitoring parameter is 25%, 12.5%, …, and 18% respectively, then the average value of the proportion of abnormal data is (25%+12.5%+…+18%) / 10=15%, and the higher the average value of the proportion of abnormal data, the higher the overall abnormality of the equipment.

[0089] In summary, compared with the prior art, the present application extracts abnormal data from the adaptive production monitoring data sequence set to obtain an abnormal production monitoring data sequence set, and calculates the average value of the proportion of abnormal data. In this way, the abnormal information in the adaptive production monitoring data is accurately extracted, and the overall abnormality level of the adaptive production monitoring is quantitatively integrated, providing reliable data support for the accurate judgment of subsequent equipment fault risk.

[0090] S50: calling an equipment fault prediction plug-in based on the adaptive monitoring parameter set and the average value of the proportion of abnormal data, and performing fault prediction and fault warning according to the abnormal production monitoring data sequence set.

[0091] In traditional equipment fault prediction, a fixed model is often used without combining real-time monitoring parameter characteristics and dynamically adjusting the prediction strategy based on the overall level of abnormal data, resulting in poor model adaptability, large prediction deviation of single model, insufficient warning accuracy, and other problems.

[0092] To solve the above problems, the present application calls an equipment fault prediction plug-in based on the adaptive monitoring parameter set and the average value of the proportion of abnormal data, and performs fault prediction and fault warning according to the abnormal production monitoring data sequence set.

[0093] Specifically, step S50 in the method comprises:

[0094] obtaining an adaptive equipment fault prediction plug-in based on the adaptive monitoring parameter set, wherein the adaptive equipment fault prediction plug-in comprises K fault prediction units, and K is an integer greater than or equal to 10;

[0095] multiplying the ratio of the average value of the proportion of abnormal data to the historical maximum average value of the proportion of abnormal data in a historical time range by K to obtain an adaptive unit selection quantity P, wherein P is greater than or equal to 3 and less than or equal to K;

[0096] randomly selecting P fault prediction units from the K fault prediction units, performing fault prediction according to the abnormal production monitoring data sequence set, and calculating the average value of the P fault prediction probabilities to obtain a fault prediction probability average value;

[0097] If the average fault prediction probability exceeds a preset fault probability threshold, a fault warning is given to the numerical control machining equipment.

[0098] In the embodiments of the present application, first, an adaptive device fault prediction plug-in is obtained based on matching of an adaptive monitoring parameter set. The adaptive device fault prediction plug-in includes K fault prediction units, K being an integer greater than or equal to 10, and each fault prediction unit being an independently trained prediction model that can independently output a fault prediction probability.

[0099] Secondly, the ratio of the average abnormal data proportion to the historical maximum average abnormal data proportion in a historical time range is multiplied by K to obtain an adaptive unit selection number P, wherein P is greater than or equal to 3 and less than or equal to K. The historical maximum average abnormal data proportion refers to the maximum average abnormal proportion before the device fails under the same working condition in the past, which is used to quantify the relative severity of the current abnormal level. For example, if the average abnormal data proportion is 15%, the historical maximum average abnormal data proportion is 30%, and K = 10, then P = 15% / 30% x 10 = 5 (rounded), that is, 5 fault prediction units are randomly selected from the K fault prediction units.

[0100] Thirdly, P fault prediction units are randomly selected from the K fault prediction units, fault prediction is performed according to the abnormal production monitoring data sequence set, and the average of the P fault prediction probabilities is calculated to obtain the average fault prediction probability. For example, 5 fault prediction units are randomly selected from the K fault prediction units, the abnormal production monitoring data sequence set is input, and 5 fault prediction probabilities are output: 85%, 90%, 78%, 82%, and 86%. Then the average is calculated to obtain the average fault prediction probability = (85% + 90% + 78% + 82% + 86%) / 5 = 84.2%. The average fault prediction probability integrates the prediction results of multiple fault prediction units, improving the reliability and robustness of the results.

[0101] Finally, if the average fault prediction probability exceeds a preset fault probability threshold, a fault warning is given to the numerical control machining equipment. The preset fault probability threshold refers to a critical probability value for determining whether to trigger a device fault warning, such as 80%. The skilled person can dynamically adjust the setting according to the device fault risk tolerance. For example, if the preset fault probability threshold is 80% and the average fault prediction probability is 84.2%, which exceeds the preset fault probability threshold, a fault warning is given to the numerical control machining equipment.

[0102] Specifically, as shown in FIG. 1, the construction method of the adaptive device fault prediction plug-in includes: Figure 2

[0103] ​Based on the historical running logs of the same type of numerical control machining equipment, a plurality of sample monitoring data sets are collected with the adaptive monitoring parameter set as a constraint, and the proportion of fault events of different sample monitoring data sets in the historical future period is counted as a sample fault prediction probability, to obtain a plurality of sample fault prediction probabilities;

[0104] The plurality of sample monitoring data sets and the plurality of sample fault prediction probabilities are used as training data and are equally divided into K parts, and K sample training sets are obtained by randomly selecting with replacement.

[0105] The K sample training sets are used to train the deep learning model to convergence respectively, to generate K fault prediction units, and to combine to obtain an adaptive equipment fault prediction plug-in.

[0106] In the embodiments of the present application, first, based on the historical running logs of the same type of numerical control machining equipment, a plurality of sample monitoring data sets are collected with the adaptive monitoring parameter set as a constraint, and the proportion of fault events of different sample monitoring data sets in the historical future period is counted as a sample fault prediction probability, to obtain a plurality of sample fault prediction probabilities. The sample fault prediction probability refers to the proportion of actual faults of the equipment corresponding to the sample monitoring data. For example, 10 numerical control machining equipment of the same type as the equipment to be monitored are selected, and the historical running logs of the past year are extracted, and 10 adaptive monitoring parameters such as bearing temperature and motor current are used as constraints to collect 5000 sample monitoring data sets. For each sample monitoring data set, the proportion of the number of faults of the equipment within 12 hours after collection to the total running time is counted. If the equipment has a fault for 6 hours within 12 hours after collection of a sample, the sample fault prediction probability is 6 / 12=50%, and finally 5000 corresponding sample fault prediction probabilities are obtained.

[0107] Secondly, the plurality of sample monitoring data sets and the plurality of sample fault prediction probabilities are used as training data and are equally divided into K parts, and K sample training sets are obtained by randomly selecting with replacement. The purpose of randomly sampling with replacement is to enhance the diversity of the training set and reduce the influence of a single sample on the model, so that each fault prediction unit can learn different features of the data. For example, 5000 training data are equally divided into K parts, and 6-8 parts are randomly selected from the K parts to form the first sample training set by sampling with replacement, and the process is repeated to obtain K sample training sets.

[0108] Finally, the K sample training sets are used to train the deep learning model to convergence respectively, to generate K fault prediction units, and to combine to obtain an adaptive equipment fault prediction plug-in. For example, the fault prediction unit can be constructed based on a convolutional neural network (CNN) architecture, which mainly consists of an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer. Specifically:

[0109] The input layer is configured to receive structured data in the abnormal production monitoring data sequence set and convert the structured data into a tensor format meeting the network input requirement.

[0110] The convolution layer is configured to perform sliding convolution operation on the input tensor by using a preset size of convolution kernel to extract local correlation features in the data. For example, a 3*3 convolution kernel is used to cover 3 continuous time windows horizontally and 3 monitoring parameters vertically, and a plurality of convolution operations are performed to capture mutation features of the parameters over time, cross-parameter linkage abnormal features, and the like, thereby providing key feature support for fault prediction.

[0111] The pooling layer is configured to perform dimension reduction processing on the feature map output by the convolution layer, and retain key features and reduce data dimensions by using maximum pooling or average pooling. For example, a 2*2 pooling window is used to retain core features such as parameter abnormal trend and time series mutation point, reduce the calculation complexity of the subsequent layer, and avoid overfitting.

[0112] The fully connected layer is configured to map the high-dimensional feature vector output by the pooling layer to a low-dimensional feature space, and integrate global features by using a ReLU activation function. For example, a fully connected layer with 128 neurons is set to fuse the local features and global trend features extracted by the convolution and pooling to provide comprehensive feature support for the final prediction.

[0113] The output layer is configured to convert the features output by the fully connected layer into probability values in the range of 0-1 by using a single neuron structure and a sigmoid activation function, as a fault prediction probability, directly reflecting the possibility of the corresponding equipment failure caused by the input abnormal production monitoring data sequence, and providing a quantitative basis for subsequent fault warning judgment.

[0114] For example, the adaptive equipment fault prediction plug-in can be trained by the following technical path: 1. Data preparation: the training process of each fault prediction unit is the same, and for example, 1 sample training set is randomly selected from K sample training sets, and the sample training set is divided into a training set, a validation set, and a test set according to a ratio of 7:1.5:1.5. 2. Model training: the sample monitoring data set in the training set is used as the input feature, and the corresponding sample fault prediction probability is used as the supervision label. The mean square error is used as the loss function, the model parameters are iteratively updated by using the Adam optimizer, the validation set loss is calculated every round in the training, and if the validation set loss does not decrease significantly (for example, the decrease amplitude is less than 0.001) for 10 consecutive rounds, the model is considered to be converged, the training is stopped, and a trained fault prediction unit is obtained. 3. Model combination: repeat the above data preparation and training process to generate K independent fault prediction units. The K generated fault prediction units are integrated according to the parallel calling logic, that is, the input data can be simultaneously transmitted to each unit and independently output the results, and the adaptive equipment fault prediction plug-in is obtained by combination.

[0115] In summary, compared with the prior art, the application calls a device fault prediction plug-in based on the adaptive monitoring parameter set and the average proportion of abnormal data, and performs fault prediction and fault warning according to the abnormal production monitoring data sequence set. In this way, the accuracy of fault prediction and the reliability of warning are improved, and accurate judgment and timely warning of device faults are realized.

[0116] In summary, the embodiments of the application have at least the following technical effects:

[0117] Compared with the prior art, the application first simulates the device operation according to the processing product sequence, material attribute sequence and process type sequence of the numerical control processing equipment in the future preset time zone, and obtains the device operation load sequence. In this way, through the device operation load sequence, the load pressure distribution of the device in the future preset time zone can be predicted in advance, and reliable data support is provided for subsequent steps of determining the adaptive warning timeliness coefficient, setting the monitoring parameters, etc.

[0118] Secondly, the application determines the adaptive warning timeliness coefficient according to the device operation load sequence and the process type sequence, and sets the adaptive monitoring parameter set based on the adaptive warning timeliness coefficient. In this way, the dynamic adaptation of the emergency degree of device fault warning and the number of monitoring parameters is realized, and the response efficiency and accuracy of the warning are improved.

[0119] Thirdly, the application filters the multi-source production monitoring data according to the adaptive monitoring parameter set in the monitoring time window of the preset time zone to obtain an adaptive production monitoring data sequence set. In this way, the redundant information in the multi-source production monitoring data is simplified, the time continuity and parameter pertinence of the data are ensured, and efficient and targeted data source support is provided for subsequent accurate device fault warning.

[0120] Further, the application extracts abnormal data from the adaptive production monitoring data sequence set to obtain an abnormal production monitoring data sequence set, and calculates the average proportion of abnormal data. In this way, the abnormal information in the adaptive production monitoring data is accurately extracted, and the overall abnormal level of the adaptive production monitoring is quantified, providing reliable data support for the accurate judgment of the device fault risk.

[0121] Finally, the application calls a device fault prediction plug-in based on the adaptive monitoring parameter set and the average proportion of abnormal data, and performs fault prediction and fault warning according to the abnormal production monitoring data sequence set. In this way, the accuracy of fault prediction and the reliability of warning are improved, and accurate judgment and timely warning of device faults are realized.

[0122] Through the technical solution, efficient integration and accurate processing of multi-source production data are realized, time efficiency, accuracy and reliability of equipment fault early warning are effectively improved, and strong support is provided for stable operation of the numerical control machining equipment and efficient decision-making of industrial production.

[0123] It should be noted that in the above embodiments, the description of each embodiment has its own emphasis, and the parts not described in detail in a certain embodiment can be referred to the related description of other embodiments.

[0124] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0125] The present application is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing devices to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices produce a device that implements the flow Figure 1 The function specified in one flow or multiple flows and / or blocks Figure 1 The function specified in one flow or multiple flows and / or blocks

[0126] These computer program instructions can also be stored in a computer-readable memory capable of guiding a computer or other programmable data processing device to work in a specific way, so that the instructions stored in the computer-readable memory produce a product including instruction devices that implement the flow Figure 1 The function specified in one flow or multiple flows and / or blocks Figure 1 The function specified in one flow or multiple flows and / or blocks

[0127] These computer program instructions can also be loaded into a computer or other programmable data processing device, so that a series of operation steps are performed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide a process for implementing the flow Figure 1 The function specified in one flow or multiple flows and / or blocks Figure 1 The function specified in one flow or multiple flows and / or blocks

[0128] While the preferred embodiments of the application have been described, additional variations and modifications can be made to these embodiments by those skilled in the art once they have the benefit of the present disclosure without departing from the spirit and scope of the application.

[0129] Obviously, numerous modifications and variations of the present application are possible in light of the above teachings. It is therefore to be understood that within the scope of the appended claims, the application can be practiced otherwise than as specifically described herein.

Claims

1. A multi-source production data fusion processing method based on an industrial information model, characterized in that the method comprises the steps of: The method comprises the following steps: According to the processing product sequence, material attribute sequence and process type sequence of the numerical control machining equipment in the future preset time zone, the equipment operation simulation is carried out, and the equipment operation load sequence is obtained; According to the equipment operation load sequence and the process type sequence, the equipment fault early warning timeliness analysis is carried out to determine the adaptive early warning timeliness coefficient, and the adaptive monitoring parameter set is set based on the adaptive early warning timeliness coefficient; In the monitoring time window of the preset time zone, the adaptive monitoring parameter set is used for screening the multi-source production monitoring data to obtain an adaptive production monitoring data sequence set; The adaptive production monitoring data sequence set is subjected to abnormal data extraction to obtain an abnormal production monitoring data sequence set, and the average proportion of abnormal data is calculated; Based on the adaptive monitoring parameter set and the average proportion of abnormal data, the equipment fault prediction plug-in is called, and the fault prediction and fault early warning are carried out according to the abnormal production monitoring data sequence set; Wherein, according to the equipment operation load sequence and the process type sequence, the equipment fault early warning timeliness analysis is carried out to determine the adaptive early warning timeliness coefficient, comprising: According to the equipment operation load sequence, the equipment load average and the equipment load fluctuation coefficient are calculated, wherein the equipment load fluctuation coefficient is the ratio of the equipment load standard deviation of the equipment operation load sequence to the equipment load average; According to the process type sequence, the process switching frequency is calculated and obtained; The ratio of the preset standard equipment load to the equipment load average is set as the first timeliness compensation coefficient; The ratio of the preset standard equipment load fluctuation coefficient to the equipment load fluctuation coefficient is set as the second timeliness compensation coefficient; The ratio of the preset standard process switching frequency to the process switching frequency is set as the third timeliness compensation coefficient; The first timeliness compensation coefficient, the second timeliness compensation coefficient and the third timeliness compensation coefficient are weighted and fused to obtain a comprehensive timeliness compensation coefficient; The product of the comprehensive timeliness compensation coefficient and the preset standard early warning time limit is taken as the adaptive early warning timeliness coefficient.

2. The industrial information model based multi-source production data fusion processing method according to claim 1, characterized in that, Based on the adaptive early warning timeliness coefficient, the adaptive monitoring parameter set is set, comprising: Obtaining a preset monitoring parameter set, wherein the preset monitoring parameter set comprises equipment operation monitoring parameters, processing process monitoring parameters and processing parameters; Based on the historical operation log of the same type of numerical control machining equipment, the correlation degrees of several preset monitoring parameters in the preset monitoring parameter set and equipment faults are respectively evaluated, and the several preset monitoring parameters are arranged in descending order of correlation degree to generate a preset monitoring parameter sequence; Based on the adaptive early warning timeliness coefficient, the adaptive monitoring index number is matched and obtained, wherein the adaptive monitoring index number and the adaptive early warning timeliness coefficient are positively correlated; According to the adaptive monitoring index number, the adaptive monitoring parameter set is obtained by selecting from the front to the rear in the preset monitoring parameter sequence. 3.The industrial information model based multi-source production data fusion processing method according to claim 1, characterized in that, In the adaptive production monitoring data sequence set, any adaptive production monitoring data sequence is randomly selected as a to-be-processed adaptive production monitoring data sequence; ​ Set the first data in the to-be-processed adaptive production monitoring data sequence as a first abnormal data, and set the adjacent data of the first abnormal data as a second monitoring data; Perform deviation calculation on the first abnormal data and the second monitoring data to obtain a first data deviation. If the first data deviation is greater than a preset deviation threshold, the second monitoring data is set as a second abnormal data, and the data deviation calculation and abnormal data selection are continued starting from the second abnormal data. If the first data deviation is less than or equal to the preset deviation threshold, the second monitoring data is discarded, and the data deviation calculation and abnormal data selection are continued starting from the first abnormal data, until the to-be-processed adaptive production monitoring data sequence is traversed, and a first abnormal data sequence is output as a first abnormal production monitoring data sequence and added to the abnormal production monitoring data sequence set. 4.The industrial information model based multi-source production data fusion processing method according to claim 1, characterized in that, Statistically calculate the abnormal data proportion of the several adaptive monitoring parameters in the abnormal production monitoring data sequence set, and perform mean value calculation on the several abnormal data proportions to obtain an abnormal data proportion mean value. 5.The industrial information model based multi-source production data fusion processing method according to claim 1, characterized in that, Call a device fault prediction plug-in based on the adaptive monitoring parameter set and the abnormal data proportion mean value, and perform fault prediction and fault warning based on the abnormal production monitoring data sequence set, including: Match and obtain an adaptive device fault prediction plug-in based on the adaptive monitoring parameter set, wherein the adaptive device fault prediction plug-in includes K fault prediction units, and K is an integer greater than or equal to 10; Multiply the ratio of the abnormal data proportion mean value to the historical maximum abnormal data proportion mean value in a historical time range by K to obtain an adaptive unit selection quantity P, wherein P is greater than or equal to 3 and less than or equal to K; Randomly select P fault prediction units from the K fault prediction units, perform fault prediction based on the abnormal production monitoring data sequence set, and perform mean value calculation on the P fault prediction probabilities to obtain a fault prediction probability mean value. If the fault prediction probability mean value exceeds a preset fault probability threshold, perform fault warning on the numerical control machining equipment.

6. The industrial information model based multi-source production data fusion processing method according to claim 5, characterized in that, The construction method of the adaptive device fault prediction plug-in includes: Based on the historical operation logs of similar numerical control machining equipment, collect multiple sample monitoring data sets based on the adaptive monitoring parameter set as a constraint, and statistically calculate the fault event proportion of different sample monitoring data sets in a historical future period as a sample fault prediction probability to obtain multiple sample fault prediction probabilities; Divide the multiple sample monitoring data sets and multiple sample fault prediction probabilities into K parts as training data, and randomly select them with replacement to obtain K sample training sets; Use the K sample training sets to train a deep learning model to convergence respectively, generate K fault prediction units, and combine to obtain an adaptive device fault prediction plug-in.

Images