A data element-based smart factory production process supervision method

Abstract

The application relates to a kind of intelligent factory production process supervision methods based on data element, belong to production data processing analysis technical field.The method is analyzed to the value condition and fluctuation condition of each kind of operating data under each kind of equipment in historical fault-free operation cycle, accurately determines the influence size of each kind of operating data on the state representation of corresponding equipment, and the fusion state coefficient representing the overall state of corresponding equipment is obtained by comprehensively considering the influence of all kinds of operating data on the state representation of corresponding equipment, the threshold range corresponding to the equipment type is constructed according to the fusion state coefficient of each kind of equipment, and more accurate fault discrimination of corresponding type equipment can be completed in subsequent operation cycle based on the constructed threshold range, and the water treatment product production supervision efficiency is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of production data processing analysis, and in particular relates to a smart factory production process monitoring method based on data elements. BACKGROUND

[0002] In the production process of traditional water treatment products, the monitoring of equipment production status is mostly realized based on a combination of manual inspection and regular sampling. However, due to the large production site and the scattered arrangement of equipment in each link, the traditional monitoring mode has the limitation of abnormal response lag, such as an average lag of 2.3 hours in discovering equipment failure and a quality defect tracing period of up to 48 hours, which can cause significant production loss.

[0003] Under this background, as the core carrier of Industry 4.0, the smart factory can realize the digital mapping and intelligent collaboration of production elements by integrating technologies such as the Internet of Things, big data, and artificial intelligence, providing the possibility to solve the lagging problem in the monitoring of traditional water treatment product production processes.

[0004] Although the current water treatment product production monitoring method effectively alleviates the lagging problem of abnormal response through the construction of a smart factory, due to the long production cycle and slow state change of each production equipment, the analysis and identification of abnormal patterns in the existing water product production monitoring process are still realized by comparing the monitoring values with fixed thresholds, resulting in insufficient identification accuracy of abnormal states and lagging and inefficient monitoring of equipment failure.

[0005] That is, the current water treatment product production monitoring process has the technical problem of insufficient accuracy of abnormal identification, resulting in low monitoring efficiency. SUMMARY

[0006] Therefore, the present application provides a smart factory production process monitoring method based on data elements to solve the technical problem of insufficient monitoring efficiency due to insufficient accuracy of abnormal identification.

[0007] The smart factory production process monitoring method based on data elements provided by the present application comprises:

[0008] The historical data set of each type of running data in the historical fault-free running period of the target category equipment is obtained at a set interval, and the running data from the collection time to the end time of the running period is selected as the reference data set from the historical data set at a preset interval time after the start time of the running period.

[0009] determine a stability coefficient of any time point in the current kind of running data in the benchmark data set according to stability of running data of a set time length before any time point and similarity of running data of the set time length before any time point and running data of the same period in other historical failure-free running periods of the target category device;

[0010] determine a reference coefficient of any time point in the current kind of running data in the benchmark data set according to the occurrence frequency of the monitoring value of any time point and the stability coefficient of any time point, take the monitoring value corresponding to the time point with the maximum reference coefficient as the reference value of the current kind of running data, and determine a deviation coefficient of each time point according to the difference between the monitoring value of each time point and the reference value;

[0011] take the ratio of the variance of the current kind of running data in the benchmark data set and the sum of variances of all kinds of running data as the influence weight of the current kind of running data, and determine a fusion state coefficient of the target category device at any time point according to the influence weight of each kind of running data and the deviation coefficient of each kind of running data at any time point;

[0012] determine a threshold range according to the fusion state coefficient of each time point, and complete fault judgment in combination with the fusion state coefficient of the target category device at the current time point in the current running period.

[0013] Further, the determination of the stability coefficient of any time point comprises:

[0014] in the current kind of running data in the benchmark data set, calculate the variance of running data of a set time length before any time point as a first variance, and calculate the variance of the difference between each adjacent running data of the set time length before any time point as a second variance, take the reciprocal of the sum of the first variance and a constant 1 as a first reciprocal, take the reciprocal of the sum of the second variance and the constant 1 as a second reciprocal, and take the average of the first reciprocal and the second reciprocal as an average reciprocal;

[0015] in the current kind of running data in the benchmark data set, calculate the similarity between a data sequence composed of running data of a set time length before any time point and a data sequence composed of running data of the same period in any other historical failure-free running period of the target category device, and take the normalized value of the product of the obtained similarity and the average reciprocal as the stability coefficient of any time point.

[0016] Further, the determination of the reference coefficient of any time point comprises:

[0017] in the current kind of running data in the benchmark data set, count the number of time points at which the monitoring value is equal to the monitoring value of any time point as the occurrence frequency of the monitoring value of any time point, and take the product of the occurrence frequency of the monitoring value of any time point and the stability coefficient of any time point as the reference coefficient of any time point.

[0018] Further, the determining of the deviation coefficient at each time point comprises:

[0019] In the current operation data of the benchmark data set, an absolute value of a difference between the monitoring value at any time point and a reference value of the current operation data is calculated, and an inverse of a sum of the absolute value and a constant 1 is recorded as the deviation coefficient at any time point.

[0020] Further, the determining of the fusion state coefficient of the target category device at any time point comprises:

[0021] In the benchmark data set, a product of the deviation coefficient of any operation data at any time point and an influence weight of the any operation data is calculated, and a sum of products corresponding to all operation data at the any time point is taken as the fusion state coefficient of the target category device at the any time point.

[0022] Further, the determining of the threshold range according to the fusion state coefficient at each time point comprises:

[0023] In the benchmark data set, a slope of the fusion state coefficient formed by the fusion state coefficient at any time point and the fusion state coefficient at a previous time point of the any time point is calculated and recorded as a first slope, a slope of the fusion state coefficient formed by the fusion state coefficient at the any time point and the fusion state coefficient at a next time point of the any time point is calculated and recorded as a second slope, and an absolute value of a difference between the first slope and the second slope is calculated.

[0024] In the benchmark data set, a similarity between a fusion state coefficient sequence corresponding to a set time length before the any time point and a fusion state coefficient sequence corresponding to a set time length after the any time point is calculated, and a product of the calculated similarity and the absolute value of the difference between the first slope and the second slope is recorded as a segmentation coefficient at the any time point.

[0025] A time point corresponding to a set number of segmentation coefficients with the largest values is taken as a segmentation time point, a whole time period corresponding to the benchmark data set is segmented into a corresponding number of time periods by the segmentation time point, and a threshold range corresponding to each time period is determined according to a mean value and a fluctuation degree of the fusion state coefficient at each time point in each time period.

[0026] Further, the determining of the threshold range corresponding to each time period according to the mean value and the fluctuation degree of the fusion state coefficient at each time point in each time period comprises:

[0027] The fluctuation degree is: ,

[0028] wherein, is the fluctuation degree of the vth time period, denotes the fusion state coefficient at the rth time point in the vth time period, denotes the sum of all fusion state coefficients in the vth time period, denotes the number of fusion state coefficients in the vth time period, To prevent the denominator from being 0 and not affecting the calculation result of a small number, the fraction is meaningful.

[0029] The product of the fluctuation degree of the current time period and the mean of the fusion state coefficient of each time in the current time period is taken as the fluctuation value of the current time period, the sum of the mean of the fusion state coefficient of each time in the current time period and the fluctuation value of the current time period is taken as the upper limit value of the threshold range corresponding to the current time period, and the difference between the mean of the fusion state coefficient of each time in the current time period and the fluctuation value of the current time period is taken as the lower limit value of the threshold range corresponding to the current time period.

[0030] Further, the fusion state coefficient of the target category device at the current time in the current running period is combined to complete fault judgment, comprising:

[0031] The time period in which the current time in the current running period falls is recorded as the corresponding time period, and when the fusion state coefficient of the target category device at the current time in the current running period exceeds the threshold range corresponding to the corresponding time period, the fault warning of the target category device is performed.

[0032] Compared with the prior art, the present application has the following beneficial effects:

[0033] The present application analyzes the value and fluctuation of various running data of each device in the historical fault-free running period, accurately determines the influence of each running data on the state representation of the corresponding device, and comprehensively considers the influence of all kinds of running data on the state representation of the corresponding device to obtain a fusion state coefficient representing the overall state of the corresponding device. According to the fusion state coefficient of each device, a threshold range corresponding to the device type is constructed, and based on the constructed threshold range, more accurate fault discrimination of the corresponding type of device can be completed in the subsequent running period, and the production supervision efficiency of the water treatment product is improved. BRIEF DESCRIPTION OF DRAWINGS

[0034] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0035] Figure 1 is a flowchart of a smart factory production process monitoring method based on data elements provided by the first embodiment of the present application. DETAILED DESCRIPTION

[0036] The overall concept of the present application is:

[0037] The state changes of various devices in the aquatic product production process are relatively regular and similar in different failure-free operation periods, and the failure of the device is judged by taking a single historical failure-free operation period as a reference for the current operation period. Specifically, the states of the devices in the aquatic product production process are fused and characterized by fusing all kinds of operation data of a certain type of device, so as to realize the failure judgment of the single type of device by using multiple data elements. While multiple data elements can improve the judgment accuracy, the fusion of multiple data elements makes the single type of device correspond to only one judgment parameter, so that the judgment is easier to realize. Finally, the fusion state threshold range of each type of device is constructed by using the failure-free operation period, and the accurate abnormality recognition of each type of device is completed.

[0038] In order to further illustrate the technical scheme of the present application, the following will be described by specific embodiments.

[0039] In the description of the present application, the reference "one embodiment" or "some embodiments" means that the specific features, structures or characteristics described in connection with the embodiment are included in one or more embodiments of the present application. Therefore, the statements "in one embodiment", "in some embodiments", "in other some embodiments", "in other some embodiments" and the like appearing in different places in the specification are not necessarily all referring to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form, and the terms "include", "contain", "have" and their variants mean "include but not limited to", unless otherwise specifically emphasized.

[0040] It should be understood that the size of the serial number of each step in the following embodiments does not mean the order of execution, and the execution order of each process should be determined according to its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0041] Method embodiment:

[0042] Referring to Figure 1 , it is a flowchart of a smart factory production process monitoring method based on data elements provided by the first embodiment of the present application, as Figure 1 shown, the method can include the following steps:

[0043] S101, acquire historical data set of each kind of operation data in historical failure-free operation period of the target category device at a set interval, and select operation data from the historical data set from the collection time to the end time of the operation period as the reference data set, with the collection time being a time after a preset interval length from the start time of the operation period.

[0044] The purpose of the embodiment is to accurately determine the failure of the related equipment in the water treatment product production process by analyzing the operation state of the equipment, and the threshold range used for determination needs to be determined by analyzing various operation data of the corresponding equipment in the historical failure-free operation period. Therefore, first, various operation data in the historical failure-free operation period of the target category device is acquired at a set interval, wherein the set interval can be set by the operator according to the size of the operation period and the specific requirements of the determination accuracy, and the set interval is preferably 1 second in the embodiment, that is, various operation data is collected from the historical failure-free operation period of the target category device every 1 second, and the historical data set is formed after traversing the operation period.

[0045] Taking the key equipment in the water treatment product production process as an example, the operation data mainly includes: membrane filtration equipment, operation data is transmembrane pressure difference (pressure transmitter collects), water production flow (electromagnetic flowmeter or turbine flowmeter collects); water pump system, operation data is flow (integrated flow collector collects), pressure (pressure sensor collects); aeration fan, operation data is air volume (pitot tube or thermal mass flowmeter collects), air pressure (pressure transmitter collects); electrical control system, operation data is voltage (voltage sensor collects), current (current sensor collects).

[0046] Considering that the probability of failure of each equipment in the starting stage of the water treatment product production period is very low, and at the same time in order to reduce the calculation amount of subsequent analysis, the embodiment eliminates a certain length of data at the beginning of the historical data set to form the reference data set, that is, the collection time is a time after a preset interval length from the start time of the operation period, and the operation data from the collection time to the end time of the operation period in the historical data set is selected as the reference data set. The value of the preset interval length can be set by the operator according to the length of the water treatment product production period and the failure generation of each equipment, and the embodiment will not be specifically exemplified.

[0047] S102, in the current kind of operation data of the reference data set, according to the stability of the operation data of a set time length before any time and the similarity of the operation data of a set time length before any time and the operation data of the same period in other historical failure-free operation periods of the target category device, the stability coefficient of any time is determined.

[0048] In the production process of water treatment equipment, there are multiple related equipment and links. The present scheme takes part of the key equipment of the core link as the analysis object. Since the parameter data types of each equipment are more and not unified, for the convenience of analysis, the multi-source parameters of each equipment can be unified and fused to obtain the fusion state coefficient of each equipment, so as to realize the comprehensive representation of the state of the single equipment by all the running data of the single equipment.

[0049] In order to realize comprehensive representation, it is necessary to determine the representation of each running data to the equipment state respectively. In this embodiment, the concept realized by the representation process is to find the most stable state value of single running data first, and then determine the deviation of each time according to the deviation of each time of the running data compared with the most stable state value in the historical fault-free running period, so as to realize the comprehensive representation of the running state of single equipment at a certain time by the deviation of each running data at the same time.

[0050] Specifically, to find the most stable state value of single running data, it is necessary to judge the stability of single running data at any time. Taking membrane filtration equipment as an example:

[0051] In a single water treatment product production cycle, the operating conditions, water quality, treatment target and other treatment condition factors remain unchanged, so the transmembrane pressure difference and water flow of the membrane filtration equipment are fluctuated in a fixed small range, and the data change at the same time in different historical fault-free running periods (i.e. the time after experiencing the same length of time from the start of the running period in different running periods) has the similarity characteristic. However, the above fluctuation characteristics and similarity characteristics are different in degree for different running data, so in order to realize more fine judgment of equipment failure, the present embodiment first determines the stability coefficient of any time in the current running data of the reference data set according to the stability of the running data of any time before the set time length and the similarity of the running data of any time before the set time length to the running data of the same period in other historical fault-free running periods of the target category equipment. Further, the stability coefficient is preferably

[0052] Wherein, represents the stability coefficient of the i-th running data of the target category equipment in the reference data set at the t-th time, is the variance of the i-th running data in the time period containing the t-th time and the t-th time being the n-th time in the sequence, wherein in order to avoid interference caused by too much data, n is set to 10, represents the local stability of the selected data, The smaller the value is, the greater the local stability is, Let represent the variance of the differences between adjacent data points within a time interval including time t, where time t is the nth time in the sequence. Indicates the local stability of the selected data. The smaller the value, the greater the stability of local changes. This represents the local comprehensive stability of the selected data. The greater the local comprehensive stability, the greater the probability that the i-th type of running data is in a stable state at time t, and the greater its reference significance. This represents the similarity between a data sequence consisting of the i-th type of operational data within a time period including time t, where time t is the nth time in the sequence, and a data sequence consisting of operational data from the same period in any other historical fault-free operating cycle of the target category equipment. Here, "same period" refers to a time period selected from any other historical fault-free operating cycle where the interval between the start time of the aforementioned time period and the start time of the operating cycle is the same, and the time span is also equal to the time span of the aforementioned time period. The similarity of the i-th type of operational data within the two time periods can be obtained by DTW (Data Transmission Wavelength Wrapping). A higher similarity indicates a higher consistency between the data's current position and historical performance, thus increasing the likelihood that the data is within the normal fluctuation range and its reference value. Normalization refers to methods such as linear normalization, norm normalization, etc.

[0053] It should be noted that for the earlier moments in the benchmark dataset, their corresponding... The running data selected during the calculation may exceed the range of the benchmark dataset. However, since the benchmark dataset is obtained by discarding data from the beginning of the historical dataset after a preset interval, for earlier moments in the benchmark dataset, its... The calculation can also be practically achieved, as long as it ensures The set duration of the running data selected during the calculation should not exceed the preset interval duration used in the construction of the benchmark dataset.

[0054] Therefore, the stability coefficient of the i-th type of operating data of the target category equipment at each time point can be obtained in the selected historical fault-free operating cycle. It should be noted that since the wear and tear of the equipment increases with the increase of usage time, the historical fault-free operating cycle selected in this embodiment is preferably the most recent historical fault-free operating cycle.

[0055] S103, in the current type of operation data of the benchmark dataset, determine the reference coefficient for any time based on the frequency of occurrence of the monitoring value at any time and the stability coefficient at any time, take the monitoring value corresponding to the time with the largest reference coefficient as the reference value of the current type of operation data, and determine the deviation coefficient for each time based on the difference between the monitoring value and the reference value at each time.

[0056] After obtaining the stability coefficients of the i-th type of operating data at all time points, it is considered that the more the specific value of the operating data repeats in the entire historical failure-free operation period or the reference data set, the closer it is to the middle value of the fluctuation range, and the greater the reference significance. Therefore, on the basis of the stability coefficients, the occurrence frequency of the monitoring value of the operating data at each time point is also counted. Specifically, in the current type of operating data of the reference data set, the number of time points at which the monitoring value is equal to the monitoring value at any time point is counted as the occurrence frequency of the monitoring value at any time point.

[0057] Then, based on the occurrence frequency of the monitoring value at any time point in the current type of operating data of the reference data set, the reference coefficient of the i-th type of operating data at any time point is calculated:

[0058]

[0059] wherein, represents the reference coefficient of the i-th type of operating data of the target category device at the t-th time point, represents the stability coefficient of the i-th type of operating data of the target category device at the t-th time point, represents the occurrence frequency of the monitoring value of the i-th type of data of the target category device at the t-th time point.

[0060] After obtaining the reference coefficients of the i-th type of operating data of the target category device at each time point, the monitoring value of the operating data corresponding to the time point with the maximum reference coefficient is taken as the reference value of the i-th type of operating data, denoted as .

[0061] The reference value represents the most stable state value of the i-th type of operating data in the historical failure-free operation period, or in other words, represents the middle value of the fluctuation range of the i-th type of operating data. After determining the middle value of the fluctuation range, the difference between the monitoring value at each time point and the reference value needs to be calculated to determine the fluctuation size of the i-th type of operating data when the target category device is failure-free in the historical failure-free operation period, thereby providing a prerequisite basis for subsequently constructing an accurate and targeted fault judgment threshold range for the target category device. Specifically, the difference between the monitoring value at each time point and the reference value is represented by constructing a deviation coefficient:

[0062] wherein, represents the deviation coefficient of the i-th type of operating data at the t-th time point, represents the monitoring value of the i-th type of operating data at the t-th time point, The reference value of the i-th type of operation data, the smaller the difference between the monitoring value at the current time and the reference value, the greater the deviation coefficient of the operation data at the current time, and the greater the fluctuation degree of the i-th type of operation data of the target category device even in the fault-free state.

[0063] S104, the ratio of the variance of the current operation data in the reference data set to the sum of the variances of all operation data is taken as the influence weight of the current operation data, and the influence weight of each operation data and the deviation coefficient of each operation data at any time are used to determine the fusion state coefficient of the target category device at any time.

[0064] Since in the fault-free operation period, the normal operation performance of each operation data of the target category device also shows the characteristics that the data fluctuates around a certain range, and according to the short board effect, when the fluctuation of a certain type of operation data is more obvious, the overall state of the target category device is more easily affected by this type of operation data, therefore, the operation data with more obvious overall fluctuation has greater influence on the device, and in the process of fusing each operation data to realize the comprehensive representation of the target category device, greater influence weight should be allocated, and thus the influence weight is constructed as follows:

[0065] Wherein, The influence weight of the i-th type of operation data of the target category device, The overall variance of the i-th type of data of the target category device in the reference data set, The number of types of operation data of the target category device, the greater the overall variance of the i-th type of data of the target category device in the reference data set, the more obvious the fluctuation of the data, the greater the influence, and the greater the influence weight in the process of realizing fusion representation.

[0066] Based on the influence weight, the fusion state coefficient of the device at each time is:

[0067] Wherein, The fusion state coefficient of the target category device at the t-th time, The influence weight of the i-th type of operation data of the target category device, The deviation coefficient of the i-th type of operation data of the target category device at the t-th time.

[0068] S105, according to the fusion state coefficient at each time, determine the threshold range, and combine the fusion state coefficient of the target category device at the current time in the current operation period to complete fault judgment.

[0069] The fusion state coefficient of each time point actually represents the overall state of the target category device at each time point in the time period corresponding to the reference data set, so the average state condition and the state floating degree of the target category device at all time points can be determined to determine the threshold range of the target category device in the subsequent running period for which the device state is to be determined.

[0070] Further, in a preferred embodiment, considering that in actual operation, due to the combined effects of factors such as fluctuations in water quality, adjustments in operating conditions, and membrane fouling processes within a single water treatment product production cycle, there are significant stage changes in actual water treatment production, such as the start-up stage, in which the operating pressure slowly rises from low pressure (0.5-1.0 MPa) to the design pressure (1.5-6.0 MPa), and the transmembrane pressure difference (TMP) rises from the initial value (0.05-0.1 MPa) to the stable value (0.1-0.15 MPa); in the stable operation stage, the operating pressure, water conductivity, and concentrated water flow fluctuate by ≤±5%, but the TMP may rise by 0.01-0.03 MPa per month due to slight fouling or pollution, the TMP fluctuation is ≤±0.01 MPa, the water turbidity is stable, but the backwashing period may gradually shorten (from every 4 hours to every 2 hours) due to membrane hole blockage, etc. Therefore, it is necessary to divide the overall period corresponding to the reference data set according to the fusion state coefficient at each time point, and construct different threshold ranges based on the divided time periods for subsequent device fault discrimination.

[0071] First, the segmentation coefficient of the segmentation time point between different stages at each time point is determined according to the fusion state coefficient at each time point:

[0072] wherein, denotes the segmentation coefficient of the target category device at the t time point, denotes the slope of the fusion state coefficient composed of the fusion state coefficient at the t time point and the fusion state coefficient at the previous time point of the t time point, denotes the slope of the fusion state coefficient composed of the fusion state coefficient at the t time point and the fusion state coefficient at the previous time point of the t time point, denotes the slope difference, and the greater the slope difference, the greater the probability that the fusion state coefficient at the t time point is a turning point, the greater the probability that it is a stage endpoint, and the greater the segmentation coefficient, The similarity between the fusion state coefficient sequence corresponding to the n time points before the t time point and the fusion state coefficient sequence corresponding to the n time points after the t time point is represented by DTW. In order to avoid interference caused by too much data, n is set to 10. If the similarity is low, it means that the data before and after the t time point is more likely to have different trends, and the probability of being a stage endpoint is greater, and the segmentation coefficient is greater.

[0073] It should be noted that, considering that the possibility of generating a stage endpoint near the start time and near the end time within the time period corresponding to the reference data set is small, and in order to ensure the validity of the above segmentation coefficient, the calculation of the segmentation coefficient for the first n time points and the last n time points within the time period corresponding to the reference data set is discarded.

[0074] After obtaining the segmentation coefficient of each time point, the time points corresponding to the set number of segmentation coefficients with the maximum value are selected as the segmentation time points, and the overall time period corresponding to the reference data set is segmented into a corresponding number of time periods by the segmentation time points. The value of the set number can be set by the operator according to the accuracy requirement of the equipment fault discrimination.

[0075] As a preferred, after obtaining the segmentation coefficient of each time point, the segmentation coefficients of all time points are arranged in order from small to large, and the specific value of the segmentation coefficient corresponding to Q3 is obtained by quartile calculation. Because the segmentation coefficient at the stage endpoint is greater than that at the non-stage endpoint, the segmentation coefficient corresponding to the stage endpoint is often arranged at the rear position of the order sequence. The mean value of all segmentation coefficients greater than or equal to Q3 is calculated and recorded as the segmentation coefficient threshold. The time points corresponding to all segmentation coefficients greater than or equal to the segmentation coefficient threshold are recorded as the segmentation time points, and the corresponding number of time periods is divided.

[0076] Then, the mean value of the fusion state coefficient of each time period is calculated as the reference state coefficient of the target category equipment in the time period , The reference state coefficient of the jth equipment in the vth time period is represented, and the reference state coefficient of each time period is taken as the basic warning threshold of the time period.

[0077] However, due to fluctuations in actual data, the calculated fusion state coefficient fluctuates, so it is necessary to set a dynamic warning threshold range according to the specific value of the fusion state coefficient of each time period. Each time period is analyzed separately, and the overall fluctuation degree of the fusion state coefficient data of each time period is obtained. For time periods with large overall fluctuation degree, set a larger dynamic warning range, and for time periods with small overall fluctuation, set a smaller dynamic warning range. The overall fluctuation degree is calculated as follows:

[0078] wherein, the fluctuation degree of the vth time period, the fusion state coefficient at the rth moment in the vth time period, the sum of all fusion state coefficients in the vth time period, the number of fusion state coefficients in the vth time period, the average difference between the sum calculated by taking the fusion state coefficient at each moment as the mean value and the actual sum, the smaller the difference, the smaller the fluctuation degree of the fusion state coefficient in the vth time period, the lower the difference between each other, the similar the numerical value, the greater the overall stability, and the smaller the overall fluctuation, In order to prevent the denominator from being 0 and not affecting the calculation result, and to ensure that the fraction is meaningful, the value is preferably 0.0001.

[0079] After obtaining the overall fluctuation degree, the dynamic threshold range is , that is, . Thus, the early warning threshold range of the target category device in each time period can be obtained, and then the time period in which the current moment in the current running period falls is recorded as the corresponding time period. When the fusion state coefficient of the target category device at the current moment in the current running period exceeds the threshold range corresponding to the corresponding time period, the fault of the target category device is warned, and a warning notification is immediately sent to the staff for checking and maintenance.

[0080] It is easy to understand that the current moment in the current running period should be determined in the same way as in the above early warning range construction process, that is, the moment after the start moment of the running period by a preset interval is taken as the starting point, so that each moment in the current running period corresponds to each moment in the above time period, that is, the ith moment in the current running period corresponds to the ith moment in the historical running period without failure.

[0081] In addition, the determination of the fusion state coefficient of the target category device at the current moment in the current running period is also from the moment after the start moment of the current running period by a preset interval as the running data collection moment, and the running data between the running data collection moment and the current moment is used to determine the fusion state coefficient at the current moment. In the determination process, the stable coefficient can be selected to be recalculated, the occurrence frequency of the monitoring value of each kind of running data at each moment can be used in the calculation of the reference coefficient based on the stable coefficient, and then the calculation of the reference coefficient is completed, and further the deviation coefficient of each kind of running data at each moment is obtained, and the influence weight of each kind of running data is also used in the above steps. Finally, the fusion state coefficient of the target type device at the current moment is obtained based on the reacquired deviation coefficient and the used influence weight, and the fault discrimination of the target type device is completed.

[0082] The embodiment of the present application analyzes the value and fluctuation of various operation data of each device in the historical fault-free operation period, accurately determines the influence of each operation data on the corresponding device, and obtains the fusion state coefficient representing the overall state of the corresponding device by comprehensively considering the influence of all kinds of operation data on the corresponding device. The threshold range corresponding to the device type is constructed according to the fusion state coefficient of each device. Based on the constructed threshold range, more accurate fault discrimination of the corresponding device type can be completed in the subsequent operation period, and the production supervision efficiency of the water treatment product is improved.

[0083] The above embodiments are only used to illustrate the technical solutions of the present application, but not limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that the technical solutions recorded in the foregoing embodiments can be modified, or some technical features can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A data-element based smart factory production process monitoring method, characterized in that, The method includes: A historical dataset is constructed by acquiring each type of operational data during the historical fault-free operation cycle of the target category equipment at a set interval. The acquisition time is the time after the start time of the operation cycle at a preset interval. The operational data from the acquisition time to the end time of the operation cycle is selected from the historical dataset as the benchmark dataset. In the current type of operating data in the benchmark dataset, the stability coefficient at any given time is determined based on the stability of the operating data for a set duration prior to any given time, and the similarity between the operating data for a set duration prior to any given time and the operating data for the same period in other historical fault-free operating cycles of the target category equipment. In the current type of operational data of the benchmark dataset, the reference coefficient for any given moment is determined based on the frequency of occurrence of the monitoring value at any given moment and the stability coefficient at any given moment. The monitoring value corresponding to the moment with the largest reference coefficient is used as the reference value for the current type of operational data. The deviation coefficient for each moment is determined based on the difference between the monitoring value and the reference value at each moment. The ratio of the variance of the current type of operating data in the benchmark dataset to the sum of the variances of all types of operating data is used as the influence weight of the current type of operating data. The fusion state coefficient of the target category device at any time is determined by the influence weight of each type of operating data and the deviation coefficient of each type of operating data at any time. The threshold range is determined based on the fusion state coefficient at each moment, and the fault judgment is completed by combining the fusion state coefficient of the target category device at the current moment in the current operating cycle. Determining the stability coefficient at any given time includes: In the current running data of the benchmark dataset, the variance of the running data before any time point and the set time period are calculated and denoted as the first variance, and the variance of the difference between each adjacent running data before any time point and the set time period are calculated and denoted as the second variance. The reciprocal of the sum of the first variance and the constant 1 is denoted as the first reciprocal, and the reciprocal of the sum of the second variance and the constant 1 is denoted as the second reciprocal. The mean of the first reciprocal and the second reciprocal is denoted as the average reciprocal. In the current type of operating data in the benchmark dataset, calculate the similarity between the data sequence consisting of operating data for a set duration before any time and the data sequence consisting of operating data during the same period in any other historical fault-free operating cycle of the target category equipment. The normalized value of the product of the obtained similarity and the reciprocal of the mean is recorded as the stability coefficient at any time. The determination of the fusion state coefficient of the target category device at any given time includes: In the benchmark dataset, the product of the deviation coefficient of any type of operating data at any time and the influence weight of the any type of operating data is calculated, and the sum of the products corresponding to all types of operating data at any time is used as the fusion state coefficient of the target category device at that time.

2. The data element based smart factory production process monitoring method as claimed in claim 1, wherein, The determination of the reference coefficient at any given time includes: In the current running data of the benchmark dataset, the number of times when the statistical monitoring value is equal to the monitoring value at any given time is taken as the frequency of occurrence of the monitoring value at any given time. The product of the frequency of occurrence of the monitoring value at any given time and the stability coefficient at any given time is taken as the reference coefficient at any given time.

3. The data element based smart factory production process monitoring method as claimed in claim 1, wherein, The determination of the deviation coefficient at each time point includes: In the current kind of running data of the benchmark data set, the absolute value of the difference between the monitoring value at any time and the reference value of the current kind of running data is calculated, and the reciprocal of the sum of the absolute value and a constant 1 is recorded as the deviation coefficient at any time.

4. The data element based smart factory production process monitoring method as claimed in claim 1, wherein, The threshold range is determined according to the fusion state coefficient at each time, and includes: In the benchmark data set, the slope of the fusion state coefficient formed by the fusion state coefficient at any time and the fusion state coefficient at the previous time of any time is calculated and recorded as a first slope, and the slope of the fusion state coefficient formed by the fusion state coefficient at the next time of any time and the fusion state coefficient at any time is calculated and recorded as a second slope, and the absolute value of the difference between the first slope and the second slope is calculated; In the benchmark data set, the similarity between the fusion state coefficient sequence corresponding to a set time length before any time and the fusion state coefficient sequence corresponding to a set time length after any time is calculated, and the product of the calculated similarity and the absolute value of the difference between the first slope and the second slope is recorded as a segmentation coefficient at any time; The time at which the set number of segmentation coefficients with the maximum value correspond is taken as a segmentation time, the overall time period corresponding to the benchmark data set is segmented into a corresponding number of time periods by the segmentation time, and the threshold range corresponding to each time period is determined according to the mean value and fluctuation degree of the fusion state coefficient at each time in each time period.

5. The data element based smart factory production process monitoring method as claimed in claim 4, wherein, The threshold range corresponding to each time period is determined according to the mean value and fluctuation degree of the fusion state coefficient at each time in each time period, and includes: The fluctuation degree is: ， wherein, is the fluctuation degree of the vth time period, denotes the fusion state coefficient at the rth time instant in the vth time period, denotes the sum of all fusion state coefficients in the vth time period, denotes the number of fusion state coefficients in the vth time period, is a small number to prevent the denominator from being zero and not affecting the calculation result, for ensuring the meaning of the fraction; The product of the fluctuation degree of the current time period and the mean value of the fusion state coefficient at each time in the current time period is taken as the fluctuation value of the current time period, the sum of the mean value of the fusion state coefficient at each time in the current time period and the fluctuation value of the current time period is taken as the upper limit value of the threshold range corresponding to the current time period, and the difference between the mean value of the fusion state coefficient at each time in the current time period and the fluctuation value of the current time period is taken as the lower limit value of the threshold range corresponding to the current time period.

6. The data-element based smart factory production process monitoring method according to claim 4 or 5, characterized in that, The fault judgment is completed by combining the fusion state coefficient of the target category device at the current time in the current running period, and includes: The time period into which the current time in the current running period falls is recorded as a corresponding time period, and the fault warning of the target category device is performed when the fusion state coefficient of the target category device at the current time in the current running period exceeds the threshold range corresponding to the corresponding time period.

Images