Big data analysis governance system and method based on artificial intelligence

Abstract

The application belongs to the technical field of data analysis, and specifically discloses a big data analysis and management system and method based on artificial intelligence, which comprises an equipment production analysis module, an equipment health evaluation module, a maintenance trigger judgment module and a maintenance decision optimization terminal. The application integrates vibration, temperature, voiceprint and other multi-source heterogeneous data to perform multi-modal data fusion modeling and construct a dynamic threshold correction mechanism, thereby significantly improving the accuracy and adaptability of predictive maintenance. In combination with equipment load and production environment disturbance factors, a dynamic correlation analysis framework is constructed to realize the upgrade from single feature anomaly detection to multi-modal linkage anomaly pattern recognition. Meanwhile, based on real-time working condition threshold self-correction and maintenance decision optimization, the maintenance lag problem under static rules is effectively solved, and through health index dynamic evaluation and resource scheduling cooperation, a closed-loop intelligent maintenance management system is formed.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of data analysis, and specifically relates to a big data analysis and management system and method based on artificial intelligence. BACKGROUND

[0002] Under the background of Industry 4.0 and digital transformation, predictive maintenance of intelligent factories realizes the change from passive response to active intervention through artificial intelligence and big data management system. The traditional maintenance mode is limited by high cost, low efficiency and data fragmentation, etc., thus highlighting the necessity of big data dimension analysis and management.

[0003] The prior art such as the Chinese patent application for invention with the application number 202310780247.9 discloses a device predictive maintenance method based on working condition time sequence analysis, which constructs a working condition data topology structure through a partition analysis, simplifies the monitoring dimension by using a clustering analysis, and realizes the minimal prediction of faults.

[0004] The prior art such as the Chinese patent application for invention with the application number 202110161025.X discloses a substation equipment maintenance method based on task reliability analysis, which quantifies the equipment reliability fluctuation based on a fault tree and sensitivity analysis, and then adjusts the dynamic maintenance cycle to ensure system stability.

[0005] Obviously, both of the above-mentioned solutions attempt to improve the maintenance efficiency through data modeling, but the former focuses on the extraction of time sequence correlation features, and the latter focuses on the control of system-level reliability boundaries, and there are still several deficiencies in the current plant equipment predictive maintenance analysis and management: 1. It relies on single modal data and lacks fusion analysis and management capability of multi-source heterogeneous data such as vibration and image, resulting in certain deviation in the effectiveness and reliability of predictive analysis.

[0006] 2. The current method is based on static threshold or fixed model, which is difficult to respond to the nonlinear characteristics of equipment degradation and production environment disturbance in real time, making the dynamic adaptability insufficient, and thus the timeliness of maintenance triggering is difficult to guarantee. SUMMARY

[0007] In view of this, in order to solve the problems raised in the background art, the present application proposes a big data analysis and management system and method based on artificial intelligence.

[0008] The purpose of the present application can be achieved by the following technical solutions: the present application provides a big data analysis and management system based on artificial intelligence, which comprises the following modules: a device production analysis module, which dynamically adjusts the data reference range under normal operation condition according to the service life of production equipment and historical fault records, and analyzes the linkage change relationship of temperature, vibration and voiceprint characteristics of production equipment in the production process.

[0009] A device health evaluation module calculates a device health index under a time sequence according to the data reference range and the linkage change relationship.

[0010] A maintenance trigger judgment module real-time corrects a maintenance trigger threshold according to current production load data and environment data, and compares the device health index with the corrected maintenance trigger threshold to output a maintenance type.

[0011] A maintenance decision optimization terminal generates an optimization scheme including a spare part preparation list and a maintenance time suggestion according to the maintenance type, a production idle period and a spare part inventory state.

[0012] The application also provides a big data analysis and management method based on artificial intelligence, which comprises the following steps: S1, device production analysis: dynamically adjusting a data reference range under a normal operation condition according to a production device service length and historical failure records, and analyzing linkage change relationships of corresponding temperature, vibration and voiceprint features of production devices in a production process.

[0013] S2, device health evaluation: calculating a device health index under a time sequence according to the data reference range and the linkage change relationship.

[0014] S3, maintenance trigger judgment: real-time correcting a maintenance trigger threshold according to current production load data and environment data, and comparing the device health index with the corrected maintenance trigger threshold to output a maintenance type.

[0015] S4, maintenance decision optimization: generating an optimization scheme including a spare part preparation list and a maintenance time suggestion according to the maintenance type, a production idle period and a spare part inventory state.

[0016] Compared with the prior art, the application has the following beneficial effects: (1) the application greatly improves the accuracy and adaptability of predictive maintenance by means of multi-modal data fusion modeling and dynamic threshold correction mechanism, simultaneously integrates multi-source heterogeneous data such as vibration, temperature and voiceprint, and considers device load and production environment disturbance factors to build a dynamic correlation analysis framework, realizes the leap from only detecting single feature anomaly to identifying multi-modal linkage anomaly mode, and further solves the maintenance lag problem caused by static rules through automatic threshold correction and maintenance decision optimization based on real-time working conditions.

[0017] (2) the application breaks through the limitation of single data mode by means of dynamic correlation modeling of vibration, temperature and voiceprint features, accurately identifies complex failure modes by means of cross-modal anomaly propagation law and intensity coupling relationship, reduces the risk of false positives and false negatives, provides high confidence basis for device health analysis, and ensures the effectiveness and reliability of the predictive analysis results.

[0018] (3) The application dynamically adjusts the triggering threshold based on real-time production load and environmental parameters, adapts to the health state and external disturbance, avoids premature or delayed maintenance caused by traditional static threshold, and ensures the spatio-temporal matching of maintenance instructions and actual health state of the equipment.

[0019] (4) The application quantifies the equipment performance degradation rate and the cumulative effect of multi-feature anomalies through health index calculation under time series, supports progressive capture of early implicit faults, and provides decision transformation support from passive response to active intervention for maintenance strategy.

[0020] (5) The application generates a globally optimal scheme of maintenance timing and resource allocation by combining spare parts inventory, production idle window and other multi-constraint conditions, reduces unplanned downtime loss and redundant spare parts occupation, and realizes the Pareto equilibrium of maintenance cost and equipment reliability. BRIEF DESCRIPTION OF DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the application, the following will briefly introduce the drawings needed to be used in the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor.

[0022] Figure 1 It is a schematic diagram of the system module structure of the application.

[0023] Figure 2 It is a schematic diagram of the method implementation step flow of the application.

[0024] Figure 3 It is a schematic diagram of the analysis flow of the linkage change relationship of each feature of the application. DETAILED DESCRIPTION

[0025] The technical solutions in the embodiments of the application will be described clearly and completely in the following with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only some embodiments of the application, not all embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the application.

[0026] Please refer to Figure 1 The application provides a big data analysis and management system based on artificial intelligence, which comprises: an equipment production analysis module, an equipment health evaluation module, a maintenance triggering judgment module and a maintenance decision optimization terminal.

[0027] In the above, the equipment health evaluation module is connected with the equipment production analysis module and the maintenance triggering judgment module respectively, and the maintenance triggering judgment module is connected with the maintenance decision optimization terminal.

[0028] The device production analysis module dynamically adjusts the data reference range under normal operating conditions according to the service life of the production equipment and the historical failure record, and analyzes the linkage change relationship of the corresponding temperature, vibration and soundprint characteristics of the production equipment in the production process.

[0029] Specifically, the adjustment of the data reference range under the normal operating condition comprises: A1, predicting the remaining service life based on the service life and the historical failure record.

[0030] A2, comparing the service life with the set safe service life and then outputting the planned remaining service life.

[0031] A3, if the difference between the predicted remaining service life and the planned remaining service life is less than 0 and exceeds the deviation range, the adjustment coefficient is assigned as 1, otherwise the ratio of the predicted remaining service life to the safe service life is taken as the adjustment coefficient.

[0032] A4, adjusting the pre-set normal value interval of temperature, vibration and soundprint characteristics through the adjustment coefficient to generate the dynamically adjusted data reference range.

[0033] The embodiment of the application breaks through the limitation of single data mode by establishing the double comparison relationship of the predicted remaining life value and the safe use threshold, and modeling according to the dynamic correlation of vibration, temperature and soundprint characteristics, utilizes the cross-modal abnormal propagation rule and intensity coupling relationship, accurately identifies the composite failure mode, reduces the false alarm and missed alarm risk, provides high confidence basis for equipment health analysis, ensures the effectiveness and reliability of the predictive analysis result, and also solves the adaptability problem of monitoring parameter drift in the equipment performance degradation process.

[0034] Understandably, as to the A1 step, the specific process of predicting the remaining service life is as follows: A11, matching each failure type with a pre-set failure type-impact weight table to obtain the corresponding impact weight of each failure.

[0035] It should be noted that the setting of the impact weight corresponding to the failure type can be quantitatively designed according to severity, propagation speed, repair cost and production loss, etc. For example, bearing overheating, motor insulation failure, etc. are taken as examples for demonstration, and Table 1 can be referred to for details.

[0036] Table 1: Example of failure type-impact weight table

[0037]

[0038] As for how to quantify the influence weight of the design fault type according to factors such as severity, propagation speed, repair cost and production loss, the following steps and logic can be used: extract the frequency of occurrence, repair time, loss amount and other data of past similar faults, and extract the fault manual provided by the reference equipment manufacturer, the risk assessment guidelines published by the industry association, or invite experts in the field to score typical faults, establish a 0-10 point or 0-1 point quantitative standard, convert qualitative description into numerical value, in order to be consistent with the value interval of the present application, 0-1 point quantification can be used, for example, if the local functional abnormality does not affect the overall operation, the severity is recorded as mild, and the severity score is set to 0.1-0.3 points, if the whole machine is paralyzed or causes a safety accident, the severity is recorded as severe, and the severity score is set to 0.8-1 point, if it is between the two, the severity is recorded as moderate, and the severity score is set to 0.3-0.8 points, if the abnormality does not spread for more than 24 hours, the propagation speed is recorded as slow, and the propagation speed score is set to 0.1-0.3 points, if it spreads to the key components within 1 hour, the propagation speed is recorded as extremely fast, and the propagation speed score is set to 0.8-1 point, if it is not the two cases, the propagation speed score is set to 0.3-0.8 points, the repair cost score is set according to the ratio of repair cost to equipment purchase cost, when the repair cost exceeds the equipment purchase cost, the repair cost score is set to 1, if the production capacity decreases by less than 5%, the production loss is recorded as slight, and the production loss score is set to 0.1-0.3 points, if the production capacity is stopped for more than 48 hours or the compensation for breach of contract is greater than the set warning for breach of contract threshold or the production capacity decrease ratio exceeds the set warning production capacity decrease ratio, the production loss is recorded as severe, and the production score is set to 0.8-1 point, otherwise, the ratio of production capacity decrease ratio to set warning production capacity decrease ratio is taken as the production loss score, the weight coefficients of each factor are determined by expert scoring or historical data statistics, for example, severity accounts for 40%, propagation speed accounts for 20%, repair cost accounts for 20%, and production loss accounts for 20%, and finally the final influence weight is obtained by weighted sum of each factor score.

[0039] The data in the table can be calculated according to the severity, propagation speed, repair cost and production loss recorded at the time of the corresponding fault type, according to the calculation method of the final influence weight described above.

[0040] It should be noted that the data in the above table can be expanded according to the specific industry characteristics, for example, adding a blade crack in power generation equipment, adding a pipeline corrosion in petrochemical equipment and other special fault types, and the specific influence weight setting can also be dynamically adjusted according to the industry characteristics.

[0041] A12, calculate the average fault influence weight according to the influence weight of all faults, and calculate the fault influence weight change rate according to the relative deviation value of the influence weight between adjacent faults in the time sequence of historical fault occurrence.

[0042] A13. Compare the timestamps of each failure to obtain the interval between adjacent failures, calculate the average fault interval, and generate the fault interval variation rate based on the dynamic trend of the interval between adjacent failures.

[0043] Among them, the dynamic change trend analysis of the interval between adjacent faults generates the fault interval change rate. The dynamic change trend curve can be generated by linear fitting, and the slope of the curve can be extracted as the fault interval change rate.

[0044] A14. Construct a fault index sequence based on the average fault impact weight, the rate of change of fault impact weight, the average fault interval duration, and the rate of change of fault interval duration.

[0045] A15. Match the fault indicator sequence with the predefined fault interference factor association table and output the matched fault interference factors.

[0046] A16. The remaining service life is predicted by outputting the comprehensive fault interference factor, usage time and safe usage time threshold.

[0047] There is a complex mapping relationship between fault indicators and fault interference factors, which is mainly obtained by matching the fault indicator sequence with a predefined fault interference factor association table. Specifically, the fault indicator sequence constructed in step A14 contains four parameters: , , and The calculation methods for these parameters are explained in detail in steps A12-A13. The fault interference factor association table is a pre-established mapping relationship based on historical equipment operating data, industry experience, or experimental data. That is, it combines historical operating data and other pre-calculated data to obtain the simulated four parameter results under various fault interference levels, and constructs the association table. In other words, the fault interference factor is obtained by matching the four parameter values ​​calculated in real time with the rows in the table, rather than by processing with a simple formula.

[0048] It should be added that the specific formula for calculating the predicted remaining useful life is as follows: , Indicates the predicted remaining useful life. Indicates the threshold for safe usage time. Indicates usage duration. This represents the fault interference factor. To ensure the attenuation setting is reasonable, the fault interference factor needs to be set with boundaries, and the fault interference factor changes dynamically over time.

[0049] For example, the early stage is set when the use time is 10% of the safe use time threshold, the middle stage is set when the use time is less than 70% of the safe use time threshold, and the aging stage is set when the use time is more than 70% of the safe use time threshold. In the early stage, the device is new and has stable overall performance after strict debugging and inspection, and the components work well and have good fault tolerance. At this time, the prediction life is not so strict, specifically, The maximum value can be 0.1. In the middle stage, as the use time increases, the industrial equipment components will gradually wear out, fatigue, and other conditions, and the fault risk will rise. Specifically, The maximum value can be 0.2. In the aging stage, the components are aging and the precision is declining, and the fault is more likely to be triggered. The device is greatly disturbed by the fault and has a significant impact on the remaining useful life. Specifically, The maximum value can be 0.5.

[0050] For another example, the average fault impact weight, fault impact weight change rate, average fault interval time, and fault interval time change rate are respectively denoted as 、 、 and , and it is assumed that the current is in the aging stage. The specific fault index sequence can refer to Table 2.

[0051] Table 2 Fault Index Sequence Example Table

[0052]

[0053] It can be understood that the A3 step needs to be explained. The deviation range is a limit value determined according to the device historical operation data, industry general standards, and device expected performance targets, etc. It is used to measure the deviation degree of the actual remaining life of the device from the planned remaining life.

[0054] When the difference between the predicted remaining useful life and the planned remaining useful life is less than 0 and exceeds the deviation range, it indicates that the actual remaining life of the device is much lower than expected, and the device state has seriously and unacceptably deteriorated, and is in an emergency state that needs maintenance. At this time, the adjustment coefficient is assigned a value of 1, which can quickly switch the maintenance process to the highest priority emergency mode and fully ensure the normal operation of the device.

[0055] When the difference does not appear in the above case, it indicates that the device state is relatively stable. The ratio of the predicted remaining useful life to the safe use time is used as the adjustment coefficient. According to the proportion of the actual remaining life of the device to the safe use time, the maintenance resource investment and maintenance strategy can be scientifically and reasonably adjusted to adapt to the maintenance needs of the device in different health states.

[0056] In a preferred embodiment of the present invention, the following supplement is provided regarding the adjustment by adjusting the coefficient in step A4: When the adjustment coefficient is set to 1, it indicates that the equipment urgently needs maintenance. At this time, it is necessary to make an urgent and significant adjustment to the normal value range of the integrated temperature, vibration, and acoustic characteristics. In this case, the originally set warning critical value range is directly called as the adjustment value range to ensure that abnormal characteristics can be well captured in emergency situations.

[0057] For example, assuming the original temperature difference between adjacent areas is [2℃, 5℃], the original diffusion rate of the temperature anomaly area is [0.1m² / h, 0.3m² / h], and the original diffusion range is [0.5m², 1m²], the corresponding adjustments are as follows: [3℃, 4℃], [0.2m² / h, 0.25m² / h], and [0.6m², 0.8m²].

[0058] The original ranges for the maximum amplitude difference in vibration characteristics are [3mm, 8mm], the original ranges for the duration of vibration amplitude deviation are [0s, 2s], and the original ranges for the maximum percentage are [10%, 20%]. After adjustment, they are as follows: [4mm, 6mm], [0s, 1s], and [12%, 16%].

[0059] The original range of the maximum decibel value in the voiceprint features is [60dB, 80dB], and the original range of the abnormal duration is [0s, 3s]. After adjustment, they are as follows: [65dB, 75dB] and [0s, 1.5s].

[0060] When the adjustment coefficient is the ratio of the predicted remaining lifespan to the safe usage time, the equipment status is relatively stable, and the normal value range of each feature is dynamically adjusted based on this.

[0061] For example, the specific formula for adjusting the lower limit of the normal value range of a feature is as follows: , This represents the lower limit of the adjusted value range. and These represent the upper and lower limits of the normal value range, respectively. Indicates the adjustment factor. This represents the adjustment factor for the lower limit value, and can specifically be 0.4.

[0062] The specific formula for adjusting the upper limit of the normal value range of the feature is as follows: , This represents the upper limit of the adjusted value range. This represents the adjustment factor for the upper limit value, and can specifically be 0.3.

[0063] Taking temperature difference as an example, when the adjustment coefficient is 0.8, the normal range of temperature difference between adjacent areas [1.5℃, 4.5℃] is adjusted as follows: the lower limit of temperature difference between adjacent areas is 1.5 + (4.5 − 1.5) × (1 − 0.8) × 0.4 = 1.74℃, and the upper limit is 4.5 − (4.5 − 1.5) × (1 − 0.8) × 0.4 = 4.26℃.

[0064] For another specific example, please refer to Figure 3 As shown, the specific analysis process of the linkage change relationship is as follows: B1. Infrared images, vibration signals and acoustic signals of each collection position of the production equipment are acquired in real time through infrared thermal imager, vibration sensor and acoustic text acquisition device, and feature extraction and standardization processing are performed.

[0065] B2. Within the same time window, check whether the temperature, vibration, and voiceprint characteristics at the same location exceed the preset threshold. If at least two characteristics exceed the threshold simultaneously, mark it as a synchronous abnormal event and record the abnormal feature combination.

[0066] B3. Construct an abnormal feature correlation matrix by statistically analyzing the abnormal intensity and direction of the three features in the synchronous abnormal event, and map the synchronous abnormal event to the corresponding physical location of the device to generate an abnormal heat map.

[0067] B4. If the anomalies of the three characteristics have a sequential order, time series analysis is used to determine the anomaly propagation path, and a multi-level causal relationship chain is constructed based on the temporal relationship. Then, the starting point and propagation path of the causal chain are marked on the anomaly heatmap.

[0068] It should be added that when the anomalies of the three features do not have a sequential order and at least two features do not exceed the threshold at the same time, it indicates that there is no linkage relationship, and the process can return to step B1 for data collection again.

[0069] Understandably, the feature extraction described in step B1 includes: B11, identifying areas in the infrared image where the temperature exceeds a set threshold, marking them as temperature anomaly areas, and marking their locations.

[0070] For example, the set temperature threshold can be determined comprehensively based on historical temperature data during normal operation of the equipment, industry standards, and the equipment's design parameters. For instance, for a specific piece of equipment, the normal operating temperature range is 30℃-50℃. After long-term monitoring and data analysis, the area with a temperature higher than 60℃ is defined as a temperature anomaly zone.

[0071] B12. Perform continuous multi-frame image comparison analysis on the temperature anomaly area to determine the diffusion speed and range of the temperature anomaly area.

[0072] The specific example of analyzing the diffusion speed of the temperature abnormal area is: selecting continuous multiple frames of images, assuming that the frames are the nth frame, the nth+1 frame, the nth+2 frame, and so on. For two adjacent frames of images, such as the nth frame and the nth+1 frame, the centroid positions of the temperature abnormal areas in the two frames are determined first. The centroid position can be obtained by calculating the weighted average value of the coordinates of each pixel point in the temperature abnormal area, and the weight is the temperature value of the pixel point. The higher the temperature, the greater the weight. Assuming that the centroid coordinates of the temperature abnormal area in the nth frame are (x n, y n), the centroid coordinates of the temperature abnormal area in the nth+1 frame are (x n+1, y n+1), the time interval between the two frames of images is Δt, and the velocity component in the x-axis direction is vx, the calculation formula of the velocity component in the x-axis direction is: vx= (x n+1- x n) / Δt. The velocity component in the y-axis direction is vy, and the calculation formula of the velocity component in the y-axis direction is: vy= (y n+1- y n) / Δt. Then, the actual moving speed of the centroid is calculated through the Pythagorean theorem. This speed value can approximately represent the diffusion speed of the temperature abnormal area in this time period. In order to obtain more accurate diffusion speed, multiple groups of adjacent frames of images can be used to calculate the average diffusion speed. , ), , ), , , , , , , ,

[0073] Further, when analyzing the diffusion range of the temperature abnormal area, the pixel counting method is used to count the number of pixels of the temperature abnormal area in each frame of image. According to the conversion relationship between the pixels of the thermal imaging device and the actual area, the number of pixels is converted into the actual area. Assuming that the actual area represented by each pixel is A, and the number of pixels of the temperature abnormal area is N, then the actual area of the temperature abnormal area is N * A. For example, the area of the temperature abnormal area in the nth frame is N n * A, and the area of the temperature abnormal area in the nth+j frame is N n+j * A. When N n+1 > N n, the diffusion range of the temperature abnormal area is N n+1- N n * A. When N n+1 < N n, the diffusion range of the temperature abnormal area is N n- N n+1 * A. , , , , , , , ,

[0074] B13, calculating the temperature difference between adjacent regions, integrating the temperature difference between adjacent regions, the diffusion speed and range of the temperature abnormal area into temperature characteristics.

[0075] B14, obtaining the maximum amplitude value difference by subtracting the preset fluctuation amplitude value from the maximum amplitude value of the vibration waveform in the vibration signal, and simultaneously counting the duration of the amplitude value difference greater than the fluctuation amplitude difference value, which is recorded as the vibration amplitude deviation duration.

[0076] B15, the proportion of high frequency and low frequency components in the vibration signal is analyzed by frequency domain analysis statistics, and the maximum proportion is extracted, and the maximum amplitude value difference, vibration amplitude deviation duration and maximum proportion are taken as vibration characteristics.

[0077] wherein the energy of each frequency band can be approximately calculated by summing the square of the amplitude value of each frequency point in the frequency band, the high frequency component proportion is the ratio of the high frequency band energy to the sum of the energy of each frequency band, and the low frequency component proportion is the ratio of the low frequency band energy to the sum of the energy of each frequency band.

[0078] B16, the maximum decibel value of the voiceprint signal is extracted, and the duration of the decibel value exceeding the normal noise is counted as the abnormal duration, and then the maximum decibel value and the abnormal duration are taken as the voiceprint characteristics.

[0079] Understandably, in the B4 step, the specific determination of determining the abnormal propagation path by using time series analysis is as follows: using time series analysis methods such as autoregressive integrated moving average model (ARIMA) to model the abnormal time series of the three characteristics respectively, and analyze its own change trend and law. For example, through the ARIMA model to analyze the change trend of temperature abnormal value with time, determine whether the temperature anomaly is gradually rising or suddenly increasing. And ARIMA model is an existing model, its specific analysis process is not described again.

[0080] In a preferred embodiment of the present application, the device overheating leading to failure is taken as an example to demonstrate the construction of the multi-level causal relationship chain and the labeling of the starting point and propagation path of the causal chain in the B4 step, and the specific examples are as follows: temperature feature anomaly at time t1, vibration amplitude sudden increase at time t2, high frequency noise in voiceprint spectrum at time t3. Among them, , , and respectively represent the interval duration between time t1 and time t2, and the interval duration between time t2 and time t3.

[0081] The first level of causality is temperature anomaly → vibration anomaly, wherein the P value of Granger causality test is 0.01, and the value of transfer entropy is 0.35. The second level of causality is vibration anomaly → voiceprint anomaly, and the value of transfer entropy is 0.28, and then the multi-level causal chain is temperature anomaly → vibration anomaly → voiceprint anomaly.

[0082] Taking the temperature anomaly as the starting point of the causal chain, mark the risk path of "overheating-vibration-noise".

[0083] ​​Granger causality test is a common statistical method for determining whether variable A is the Granger cause of variable B, i.e. whether the historical information of variable A has a significant prediction effect on the current value of variable B. The calculation logic of P value = 0.01, transfer entropy = 0.35 and transfer entropy = 0.28 involved in the causality verification of "temperature anomaly → vibration anomaly" and "vibration anomaly → voiceprint anomaly" can refer to the technical principles mentioned in the text and the following prior art supplementary description:

[0084] P value calculation: the temperature values exceeding the preset threshold value are combined to form a temperature anomaly value sequence, and the vibration values exceeding the preset threshold value are combined to form a vibration anomaly value sequence. Stationarity test such as ADF test is performed on the temperature anomaly value sequence and the vibration anomaly value sequence. If it is not stationary, difference processing is performed to convert it into a stationary sequence. A univariate ARIMA model of the vibration anomaly value sequence only depends on its own historical values, and a multivariate ARIMA model of the vibration anomaly value sequence depends on its own historical values and the historical values of the temperature anomaly value sequence.

[0085] Compare the prediction error sum of squares of the univariate ARIMA model and the multivariate ARIMA model to calculate the F statistic, wherein, represents the prediction error sum of squares of the univariate ARIMA model, represents the prediction error sum of squares of the multivariate ARIMA model, is the lag order of the independent variable such as temperature, is the lag order of the dependent variable such as vibration, is the sample size, and the P value refers to the probability of observing the current F value and more extreme values under the null hypothesis that there is no causal relationship between temperature anomaly and vibration anomaly. The F value follows an F distribution with degrees of freedom , and the right tail probability corresponding to the F value can be calculated through the distribution, i.e. the P value.

[0086] Transfer entropy value calculation: the transfer entropy calculation formula is: which is used to measure the prediction contribution of the current state of variable X to the future state of variable Y given the historical state of variable Y.

[0087] The calculation of transfer entropy based on the content of the present application needs to be based on the joint probability distribution of time series. Continuous temperature, vibration and voiceprint feature values are divided into discrete states such as low, medium and high to facilitate the calculation of probability distribution, and is the state of temperature or vibration at t-1, is the state of temperature or vibration at t, is the state of vibration or voiceprint at t, and the following probabilities are calculated through historical data , and , and the transfer entropy value is calculated by substituting the transfer entropy formula.

[0088] It should be noted that the first-order causality determines the causality between the temperature anomaly and the vibration anomaly through Granger causality test and transfer entropy. The P value of Granger causality test is 0.01, which is less than the common significance level such as 0.05, which shows that the temperature anomaly has a significant impact on the vibration anomaly in statistics, that is, the temperature anomaly is one of the reasons for the vibration anomaly. The transfer entropy is 0.35, which shows the degree of information transfer between the two variables, that is, the temperature anomaly transfers a certain amount of information to the vibration anomaly, which further supports the causality.

[0089] Although the second-order causality does not perform Granger causality test, it determines the causality between the vibration anomaly and the voiceprint anomaly through the transfer entropy. The transfer entropy is 0.28, which shows that the vibration anomaly transfers a certain amount of information to the voiceprint anomaly, that is, the vibration anomaly is one of the reasons for the voiceprint anomaly. When the value of the transfer entropy is 0, it means that there is no information transfer between the two variables, and they are independent of each other. When the value of the transfer entropy is greater than 0, it means that the state of one variable can provide a certain amount of prediction information for another variable, that is, there is information transfer.

[0090] The device health evaluation module calculates the device health index under the time sequence according to the data reference range and the linkage change relationship.

[0091] Specifically, the device health index analysis includes: D1, calculating the deviation degree of temperature, vibration and voiceprint features in each time window relative to the data reference range.

[0092] Understandably, the deviation degree is obtained by calculating the difference between the temperature, vibration and voiceprint feature values in each time window and the corresponding data reference range, and comparing it with the amplitude of the reference range.

[0093] D2, setting the device linkage deviation degree based on the linkage change relationship, and analyzing the comprehensive device deviation degree according to the deviation degree of each feature and the device linkage deviation degree.

[0094] D3, if the comprehensive deviation degree is 0, the upper limit value of the preset device health index reference range is taken as the device health index thereof, otherwise the inverse of the comprehensive deviation degree is mapped into the preset device health index reference range through the Sigmoid function, and the mapping value is taken as the device health index thereof.

[0095] It should be noted that the comprehensive deviation degree of 0 means that the characteristics of the device are completely consistent with the normal standard value without any deviation. At this time, the device is in an ideal normal operating state, and in this case, the upper limit value of the preset device health index reference range can be taken as the device health index. On the contrary, the more ideal the device operating state is, the higher the potential failure risk is. The sigmoid function has the characteristic of compressing the input value to the interval (0, 1), and can reasonably compress and convert the reciprocal of a large range of comprehensive deviation degrees.

[0096] D4, binding the device health index of each time window with the window end timestamp, outputting the device health index under the time sequence.

[0097] The embodiment of the application quantifies the device performance degradation rate and the cumulative effect of multi-feature anomalies through health index calculation under the time sequence, supports progressive capture of early implicit failures, and provides decision transformation support from passive response to active intervention for maintenance strategies.

[0098] Further, the setting process of the device linkage deviation degree in the D2 step comprises: D21, calculating the linkage strength between temperature, vibration and voiceprint features through the Pearson correlation coefficient to generate linkage weights.

[0099] It should be noted that the Pearson correlation coefficient is an existing function, and the specific calculation formula is not shown.

[0100] D22, if a synchronous abnormal event is triggered, obtaining an abnormal trigger weight according to the preset weight table matched with the abnormal feature combination, and analyzing the maximum weight according to the abnormal trigger weight and the linkage weight.

[0101] It should be noted that the preset weight table is prepared in advance based on a large amount of historical data and field expert experience. The table gives corresponding weight values for different abnormal feature combinations, reflecting the influence degree of each abnormal feature combination on the device failure or abnormal state.

[0102] D23, if no abnormal event is triggered, performing Granger causality test on the multi-level causal relationship chain, distributing causal weights according to the absolute value proportion of the Granger coefficient, and generating the maximum weight by comprehensively combining the linkage weight and the causal weight.

[0103] It should be noted that the Granger causality test is a common technical means, and the specific test process is not described again, and the maximum weight generated by comprehensively combining the linkage weight and the causal weight can be obtained by calculating the mean value.

[0104] D24, calculating the output device linkage deviation degree by weighted summation based on the maximum weight and the deviation degree of each feature.

[0105] The maintenance trigger judgment module corrects the maintenance trigger threshold in real time according to current production load data and environment data, and outputs the maintenance type by comparing the equipment health index with the corrected maintenance trigger threshold.

[0106] Specifically, the process of correcting the maintenance trigger threshold in real time comprises the following steps: H1, extracting equipment utilization and production plan from current production load data, and comprehensively analyzing and calculating production urgency.

[0107] H2, the current environment data and the adaptive environment data under the normal operation condition of the equipment are correspondingly deviated to obtain the environment deviation degree.

[0108] It should be noted that the calculation principle of the environment deviation degree is the same as that of the deviation degree in the above D1 step, and will not be supplemented and explained.

[0109] H3, generating a correction factor after normalizing the production urgency and the environment deviation degree, and adjusting the preset maintenance trigger threshold based on the correction factor.

[0110] Further, the specific output process of the maintenance type comprises: E1, matching the equipment health index under the current time window with the trigger equipment health index threshold corresponding to each maintenance type, and if the matching is successful, outputting the matched maintenance type.

[0111] E2, otherwise, constructing an equipment health index change curve with the time window as the horizontal coordinate and the equipment health index as the vertical coordinate, calculating the slope of the curve as the equipment health index change rate.

[0112] E3, if the equipment health index change rate is less than 0 and its absolute value exceeds the preset warning value, outputting the component detection maintenance type.

[0113] The embodiment of the application dynamically adjusts the trigger threshold based on real-time production load and environment parameters, adapts to the health state and external disturbance, avoids premature or delayed maintenance caused by traditional static threshold, and ensures the spatio-temporal matching of maintenance instructions and actual health state of the equipment.

[0114] The maintenance decision optimization terminal generates an optimization scheme containing a spare part preparation list and a maintenance time suggestion according to the maintenance type, the production idle period and the spare part inventory state.

[0115] The embodiment of the application generates a globally optimal scheme of maintenance opportunity and resource allocation by combining multiple constraint conditions such as spare part inventory and production idle window, reduces non-planned downtime loss and redundant spare part occupation, and realizes the Pareto equilibrium of maintenance cost and equipment reliability.

[0116] For example, the specific generation of the optimization scheme is as follows: check the production schedule, select suitable continuous idle windows, check the spare parts inventory status, and determine the required spare parts inventory quantity, availability and replenishment cycle.

[0117] If spare parts are sufficient and a certain production idle period matches the maintenance time requirement, a spare parts preparation list is generated according to the maintenance type. The preventive maintenance list includes routine and easily damaged parts, and the production idle period is also used as the maintenance time. The component inspection and maintenance list includes professional inspection tools and small test replacement parts, and the maintenance time is planned according to the inspection process and duration.

[0118] If spare parts need to be replenished, a plan should be developed based on the replenishment cycle and subsequent production downtime. Preventative maintenance should follow the replenishment cycle for regular consumable parts, coordinating subsequent downtime and clarifying procurement time and transportation methods. For component inspection and maintenance, the difficulty and cycle of replenishing specialized spare parts should be considered; advance communication with suppliers for customized or expedited procurement should be maintained, and maintenance times should be adjusted according to new timeframes, planning inspection and component replacement procedures.

[0119] This invention significantly improves the accuracy and adaptability of predictive maintenance by leveraging multimodal data fusion modeling and a dynamic threshold correction mechanism. It integrates heterogeneous data from multiple sources, including vibration, temperature, and acoustic signatures, and considers equipment load and environmental disturbances to build a dynamic correlation analysis framework. This achieves a leap from detecting only single-feature anomalies to identifying multimodal linked anomaly patterns. Furthermore, it automatically corrects thresholds and optimizes maintenance decisions based on real-time operating conditions, effectively mitigating the maintenance lag problem caused by static rules.

[0120] Please see Figure 2 As shown, the present invention also provides a big data analysis and governance method based on artificial intelligence. The method includes: S1, equipment production analysis: dynamically adjusting the data reference range under normal operating conditions according to the usage time of production equipment and historical fault records, and analyzing the linkage relationship of temperature, vibration and acoustic characteristics of production equipment during the production process.

[0121] S2. Equipment Health Assessment: Calculate the equipment health index in the time series based on the data reference range and the linkage relationship.

[0122] S3. Maintenance Trigger Judgment: Adjust the maintenance trigger threshold in real time based on the current production load data and environmental data, and output the maintenance type by comparing the equipment health index with the adjusted maintenance trigger threshold.

[0123] S4. Maintenance Decision Optimization: Based on maintenance type, production idle time, and spare parts inventory status, generate an optimization plan that includes a spare parts preparation list and maintenance time suggestions.

[0124] The above formula is obtained by collecting a large amount of data for software simulation of a recent real situation, and preset parameters in the formula are set by a person skilled in the art according to actual conditions.

[0125] The above embodiments can be realized wholly or partially by software, hardware, firmware or any other combination. When realized by software, the above embodiments can be realized wholly or partially in the form of a computer program product.

[0126] Those skilled in the art can realize that the modules and algorithm steps of the examples described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solutions. Those skilled in the art can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0127] In addition, each functional module in each embodiment of the present application can be integrated in one processing module, or each module can exist physically alone, or two or more modules can be integrated in one module.

[0128] The above is merely specific embodiments of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0129] Finally, the above is merely preferred embodiments of the present application, and is not used to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be covered within the protection scope of the present application.

Claims

1. An artificial intelligence-based big data analysis governance system, characterized by, The system comprises the following modules: A device production analysis module dynamically adjusts the data reference range under normal operating conditions according to the service life of the production equipment and historical fault records, and analyzes the linkage change relationship of the corresponding temperature, vibration and soundprint characteristics of the production equipment in the production process; A device health assessment module calculates the device health index under the time sequence according to the data reference range and the linkage change relationship; A maintenance trigger judgment module real-time corrects the maintenance trigger threshold according to the current production load data and environmental data, and compares the device health index with the corrected maintenance trigger threshold to output the maintenance type; A maintenance decision optimization terminal generates an optimization scheme including a spare parts preparation list and a maintenance time suggestion according to the maintenance type, production idle period and spare parts inventory status; The device health index analysis comprises: Calculating the deviation of temperature, vibration and soundprint characteristics in each time window relative to the data reference range; Setting the device linkage deviation based on the linkage change relationship, analyzing the comprehensive device deviation based on the deviation of each characteristic and the device linkage deviation; If the comprehensive deviation is 0, the upper limit value of the preset device health index reference range is taken as the device health index, otherwise the inverse of the comprehensive deviation is mapped into the preset device health index reference range through a Sigmoid function, and the mapped value is taken as the device health index; Binding the device health index of each time window with the window end timestamp, and outputting the device health index under the time sequence. 2.The artificial intelligence-based big data analysis governance system of claim 1, wherein: The adjustment of the data reference range under normal operating conditions comprises: Predicting the remaining service life based on the service life and historical fault records; Comparing the service life with the set safe service life to output the planned remaining service life; If the difference between the predicted remaining service life and the planned remaining service life is less than 0 and exceeds the deviation range, the adjustment coefficient is assigned as 1, otherwise the ratio of the predicted remaining life to the safe service life is taken as the adjustment coefficient; Adjusting the pre-set normal value range of temperature, vibration and soundprint characteristics through the adjustment coefficient to generate the dynamically adjusted data reference range. 3.The artificial intelligence-based big data analytics governance system of claim 2, wherein: The specific process of predicting the remaining service life is as follows: Matching each fault type with a preset fault type-impact weight table to obtain the corresponding impact weight of each fault; Calculating the average fault impact weight according to the impact weight of all faults, and calculating the relative deviation value of the impact weight according to the time sequence of the historical fault occurrence and the impact weight of the adjacent faults; Comparing the timestamps of each fault occurrence to obtain the interval length of adjacent faults, calculating the average fault interval length and generating the fault interval length change rate according to the dynamic change trend of the interval length of adjacent faults; Building a fault index sequence according to the average fault impact weight, the fault impact weight change rate, the average fault interval length and the fault interval length change rate; Matching the fault index sequence with a predefined fault disturbance factor association table to output the matched fault disturbance factor; Outputting the predicted remaining service life by comprehensively considering the fault disturbance factor, the service life and the safe service life threshold. 4.The artificial intelligence-based big data analysis governance system of claim 1, wherein: The specific analysis process of the linkage change relationship is as follows: Real-time acquisition of infrared images, vibration signals and acoustic signals of each collection position of the production equipment through infrared thermal imager, vibration sensor and acoustic fingerprint collection device, and feature extraction and standardization processing are performed; In the same time window, check whether the temperature, vibration and acoustic features of the same position exceed the preset threshold value, and if at least two features exceed the threshold value at the same time, mark it as a synchronous abnormal event and record the abnormal feature combination; Statistical abnormal intensity and direction of three features in synchronous abnormal event to construct abnormal feature correlation matrix, and map the synchronous abnormal event to the corresponding physical position of the equipment to generate abnormal heat map; If there is a sequence of abnormalities of the three features, determine the abnormal propagation path by time series analysis, and construct a multi-level causal relationship chain based on the time sequence relationship, and then mark the starting point and propagation path of the causal chain in the abnormal heat map. 5.The artificial intelligence-based big data analytics governance system of claim 4, wherein: The feature extraction includes: Identify the area where the temperature exceeds the set threshold value in the infrared image, and record it as a temperature abnormal area and mark the position; Compare and analyze the diffusion speed and range of the temperature abnormal area for continuous multiple frames of images; Calculate the temperature difference between each adjacent area, and integrate the temperature difference between adjacent areas, the diffusion speed and range of the temperature abnormal area into temperature features; Determine the maximum amplitude value difference by subtracting the preset fluctuation amplitude value from the maximum amplitude value of the vibration waveform in the vibration signal, and simultaneously count the duration of the amplitude value difference greater than the fluctuation amplitude difference value, which is recorded as the vibration amplitude deviation duration; Statistically analyze the proportion of high and low frequency components in the vibration signal by frequency domain analysis, and extract the maximum proportion, which is used as the vibration feature together with the maximum amplitude value difference, vibration amplitude deviation duration and maximum proportion; Extract the maximum decibel value of the acoustic fingerprint signal, and count the duration of the decibel value exceeding the normal noise, which is recorded as the abnormal duration, and then use the maximum decibel value and abnormal duration as the acoustic feature. 6.The artificial intelligence-based big data analysis governance system of claim 1, wherein: The setting process of the equipment linkage deviation degree includes: Calculate the linkage strength between temperature, vibration and acoustic features by Pearson correlation coefficient to generate linkage weight; If a synchronous abnormal event is triggered, obtain the abnormal trigger weight according to the abnormal feature combination matching the preset weight table, and analyze the maximum weight according to the abnormal trigger weight and the linkage weight; If no abnormal event is triggered, perform Granger causality test on the multi-level causal relationship chain, distribute the causal weight according to the absolute value proportion of the Granger coefficient, and generate the maximum weight by integrating the linkage weight and the causal weight; Calculate the output equipment linkage deviation degree based on the maximum weight and the deviation degree of each feature.

7. The artificial intelligence-based big data analytics governance system of claim 1, wherein: The process of real-time correction of maintenance trigger threshold includes: Extract the equipment utilization and production plan from the current production load data, and comprehensively analyze and calculate the production urgency; Calculate the environmental deviation degree by comparing the current environmental data with the adaptive environmental data under the normal operating condition of the equipment; Generate a correction factor based on the production urgency and environmental deviation degree, and adjust the preset maintenance trigger threshold based on the correction factor.

8. The artificial intelligence-based big data analytics governance system of claim 1, wherein: The specific output process of the maintenance type includes: Match the equipment health index in the current time window with the trigger equipment health index threshold corresponding to each maintenance type, and if the matching is successful, output the matched maintenance type; Otherwise, a device health index change curve is constructed with time window as the horizontal coordinate and device health index as the vertical coordinate, and a curve slope is calculated as a device health index change rate; If the device health index change rate is less than 0 and the absolute value exceeds a preset warning value, the output component detects a maintenance type.

9. A method for big data analysis governance based on artificial intelligence, characterized in that: The method comprises: S1, device production analysis: dynamically adjusting the data reference range under normal operation conditions according to the service life of the production equipment and the historical fault records, and analyzing the linkage change relationship of the corresponding temperature, vibration and voiceprint characteristics of the production equipment in the production process; S2, device health evaluation: calculating the device health index under the time sequence according to the data reference range and the linkage change relationship; S3, maintenance trigger judgment: real-time correcting the maintenance trigger threshold according to the current production load data and environmental data, and comparing the device health index with the corrected maintenance trigger threshold to output the maintenance type; S4, maintenance decision optimization: generating an optimization scheme including a spare parts preparation list and a maintenance time suggestion according to the maintenance type, the production idle period and the spare parts inventory state.

Images