Industrial equipment operation data analysis method and system based on artificial intelligence, equipment and medium

Abstract

The invention relates to the technical field of industrial data analysis, and discloses an artificial intelligence-based industrial equipment operation data analysis method and system, equipment and a medium, and the method comprises the steps: carrying out the dynamic noise reduction processing of the multi-source sensor data of the industrial equipment, and dynamically adjusting the size of a filtering window; performing cross-modal feature alignment on the multi-sensor data after noise reduction, and respectively extracting time-frequency domain features; constructing a lightweight state evaluation model based on the equipment maintenance knowledge base, performing equipment state evaluation on the fused features, and outputting a confidence index; performing multi-stage verification on the evaluation result, and dynamically correcting the confidence index according to the verification result; and triggering a control instruction in a grading manner according to the corrected confidence index, and completing closed-loop response from data analysis to equipment control. The method breaks through the technical bottleneck of a traditional method in a dynamic industrial scene on the premise of ensuring privacy security (data are not delivered out of a factory), and provides a standardized solution for predictive maintenance and intelligent control.

Description

Technical Field

[0001] The present invention relates to the field of industrial data analysis technology, and in particular to an artificial intelligence-based industrial equipment operation data analysis method, system, equipment and medium. Background Art

[0002] With the rapid development of Industry 4.0 and smart manufacturing, the automation and intelligence of various industrial equipment in production processes are constantly increasing. The real-time collection and analysis of massive amounts of equipment data has become a critical foundation for accurate prediction, fault diagnosis, and production optimization. Traditional centralized data management models face numerous challenges in terms of privacy protection, data transmission, and compatibility between different devices, making them unable to fully meet the needs of today's distributed industrial scenarios.

[0003] Currently, the application of artificial intelligence technology in the industrial sector is becoming increasingly profound. Various deep learning and machine learning algorithms can effectively extract status characteristics and fault symptoms from massive amounts of data, providing a basis for predictive maintenance and production decision-making. However, factors such as environmental differences in industrial sites, the diversity of equipment types, and uneven data distribution make models susceptible to noise interference and difficult to adapt to dynamic changes in multiple scenarios. At the same time, factories or equipment distributed in different locations face conflicts between data sharing and privacy protection. A solution that balances distributed learning and adaptive evolution is urgently needed. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides an industrial equipment operation data analysis method based on artificial intelligence, which can perform real-time, reliable and high-precision analysis of industrial data that is dispersed, dynamically changing and noisy.

[0006] To solve the above technical problems, the present invention provides the following technical solutions: an artificial intelligence-based industrial equipment operation data analysis method, comprising: dynamically performing noise reduction processing on multi-source sensor data of industrial equipment, and dynamically adjusting the filter window size according to the real-time operating parameters of the equipment during the noise reduction process; performing cross-modal feature alignment on the noise-reduced multi-sensor data, and extracting time-frequency domain features respectively; constructing a lightweight state assessment model based on an equipment maintenance knowledge base, performing equipment state assessment on the fused features and outputting a confidence index; using a physical rule verification engine to perform multi-level verification on the assessment results, and dynamically correcting the confidence index according to the verification results; and triggering control instructions in a hierarchical manner according to the corrected confidence index to complete a closed-loop response from data analysis to equipment control.

[0007] As a preferred solution of the artificial intelligence-based industrial equipment operation data analysis method of the present invention, the dynamic adjustment of the filter window size includes: when the equipment load rate exceeds a preset threshold, using a first time window for data filtering; when the equipment load rate is lower than or equal to the preset threshold, using a second time window for data filtering, wherein the first time window is smaller than the second time window;

[0008] During the filtering process, physical parameter constraints of the equipment are imposed. Specifically, values ​​in the sensor data that exceed the rated operating parameter range of the equipment are truncated; and components in the vibration sensor data whose frequency components exceed the set multiple of the critical speed of the equipment rotor are filtered out.

[0009] As a preferred solution of the artificial intelligence-based industrial equipment operation data analysis method described in the present invention, the cross-modal feature alignment includes performing a fast Fourier transform on the noise-reduced vibration signal to obtain an actual spectrum; and calculating the similarity between the actual spectrum and the theoretical vibration spectrum of the equipment, wherein the theoretical vibration spectrum is derived from the structural parameters provided by the equipment manufacturer.

[0010] As a preferred solution of the artificial intelligence-based industrial equipment operation data analysis method of the present invention, the cross-modal feature alignment further includes dynamic weight allocation during feature fusion, and when the similarity between the actual spectrum and the theoretical spectrum is lower than a set threshold, the weight ratio of the frequency domain feature in the fusion feature is reduced;

[0011] When the peak value of the time domain feature exceeds the statistical range of historical data, the weight ratio of the time domain feature in the fusion feature is increased.

[0012] As a preferred embodiment of the artificial intelligence-based industrial equipment operation data analysis method of the present invention, the lightweight state assessment model is constructed by using an improved decision tree model, wherein when splitting the decision tree nodes, features with a high correlation with the equipment's historical fault records are preferentially selected as split nodes, and each leaf node additionally records the distribution of the evaluation results of the most recent N similar data samples;

[0013] The confidence index calculation method includes obtaining the historical result distribution corresponding to the current evaluation result at the decision tree leaf node, and calculating the confidence value according to the degree of deviation between the current result and the historical distribution. The greater the deviation, the lower the confidence.

[0014] As a preferred solution of the artificial intelligence-based industrial equipment operation data analysis method of the present invention, wherein: the multi-level verification includes primary verification and advanced verification;

[0015] Among them, primary verification is to directly compare the assessment results with the equipment safety threshold; advanced verification is to verify whether the physical relationship between multiple parameters in the assessment results is consistent with the equipment operation mechanism;

[0016] The dynamic correction of the confidence index includes multiplying the confidence index by a first correction coefficient when the evaluation result fails the primary verification; and multiplying the confidence index by a second correction coefficient when the evaluation result fails the advanced verification, wherein the second correction coefficient is smaller than the first correction coefficient.

[0017] As a preferred solution of the artificial intelligence-based industrial equipment operation data analysis method of the present invention, wherein: the trigger control instruction includes directly sending an adjustment instruction to the equipment control system when the confidence index is greater than a first threshold;

[0018] When the confidence index is between the first threshold and the second threshold, a review request is sent to the monitoring terminal and the current control parameters are maintained;

[0019] When the confidence indicator is less than the second threshold, the sensor calibration procedure is triggered and the data analysis process is suspended.

[0020] As a preferred solution of the artificial intelligence-based industrial equipment operation data analysis system of the present invention, it includes: a data processing module, a feature fusion module, and a credibility verification control module;

[0021] The data processing module collects multi-source sensor data in real time, dynamically switches the filter window algorithm according to the load rate, and implements physical parameter constraints;

[0022] The feature fusion module performs cross-modal feature alignment and dynamic fusion on the preprocessed data;

[0023] The credibility verification control module performs physical rule verification and closed-loop control in a hierarchical manner.

[0024] A computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of an industrial equipment operation data analysis method based on artificial intelligence when executing the computer program.

[0025] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of an industrial equipment operation data analysis method based on artificial intelligence.

[0026] The present invention's beneficial effects include dynamically adjusting the filter window and the physical parameters of the device, addressing the issues of effective feature loss or residual noise caused by traditional fixed filter parameters. Combining theoretical spectrum similarity comparison with historical data statistical thresholds, this method overcomes the limitations of single modal features and achieves a closed-loop system for industrial equipment data analysis, with controllable data quality, explainable decision-making processes, and traceable control responses. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0028] Figure 1 A schematic flow chart of an artificial intelligence-based industrial equipment operation data analysis method provided in accordance with one embodiment of the present invention.

[0029] Figure 2 A schematic diagram of the working modules of an artificial intelligence-based industrial equipment operation data analysis system provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0030] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0031] Example 1, reference Figure 1 , which is the first embodiment of the present invention, provides an industrial equipment operation data analysis method based on artificial intelligence, including:

[0032] S1: Dynamically reduce the noise of multi-source sensor data of industrial equipment. During the noise reduction process, the filter window size is dynamically adjusted according to the real-time operating parameters of the equipment.

[0033] Furthermore, the load rate is calculated by collecting the ratio of the current signal of the device in real time to the preset rated current value. When the load rate exceeds 70%, the filter window is set to a fixed duration of 10ms, and the sensor data is processed using a sliding average filter algorithm; when the load rate is lower than or equal to 70%, the filter window is extended to 100ms and switched to a weighted moving average filter, where the weight of the latest data point in the window is 0.6, and the weight of the historical data point decreases successively according to the time attenuation coefficient of 0.4.

[0034] When switching filter windows, a transition buffer mechanism is used: within 50ms after the load rate first exceeds the threshold, the window size is linearly reduced from 100ms to 10ms to avoid data jumps. For scenarios with frequent load rate fluctuations (more than 5 changes per minute), adaptive window mode is enabled, using the average of the last 10 load rate changes as the basis for window adjustment.

[0035] It should be noted that during the filtering process, physical device parameter constraints are applied. Engineering parameters such as the motor's rated power and the maximum allowable bearing speed are read from the device control system to establish a dynamic constraint rule base. For temperature sensor data, if the instantaneous value exceeds 105% of the rated operating temperature upper limit, the data is replaced with the moving average of the previous three sampling periods. For vibration sensor data, frequency components exceeding 1.2 times the rotor's critical speed are calculated in real time. A digital band-stop filter is designed with a center frequency of 1.2 times the critical speed and a bandwidth of ±5 Hz to attenuate the signal in this frequency band by -20 dB. If an excessive frequency is detected at five consecutive sampling points, a speed adjustment suggestion is automatically sent to the device control system, reducing the current speed to below 0.8 times the critical speed.

[0036] S2: Perform cross-modal feature alignment on the denoised multi-sensor data and extract time-frequency domain features respectively.

[0037] Furthermore, the vibration sensor signal is framed and processed with 256 sampling points per frame, with a frame shift of 128 points, and windowed using the Hanning window function. After performing a fast Fourier transform on each frame of data, the first 10 main frequency components and their amplitudes are extracted. The corresponding features, including the theoretical values ​​of gear meshing frequency and bearing ball passing frequency, are retrieved from a pre-generated theoretical spectrum database based on the device model. The similarity between the actual spectrum and the theoretical spectrum is calculated using a segmented energy matching method: the 0-5kHz frequency band is divided into 20 equal-width sub-bands, and the ratio of actual energy to theoretical energy in each sub-band is calculated. If more than 80% of the sub-band ratios fall within the range of 0.7-1.3, it is judged to be highly similar (>75%); otherwise, a spectrum anomaly flag is generated, triggering the adjustment of the frequency domain feature weights.

[0038] It should be noted that when the spectrum similarity is lower than 60%, the initial weight of the frequency domain feature in the fusion is reduced from 0.5 to 0.3, and the weight of the time domain feature is increased from 0.5 to 0.7; if the time domain peak exceeds the 95% percentile of the historical data statistics of the last 24 hours, the time domain weight is increased by an additional 0.1 (not exceeding 0.8). The weight adjustment adopts a smooth transition strategy: the amplitude of each weight change does not exceed 0.2 / second to avoid drastic fluctuations in feature weights. For situations where both low spectrum similarity and excessive time domain peaks exist, a double verification mechanism is activated: the highest abnormality level marks of the time domain and frequency domain features are retained for priority processing by the subsequent state assessment model.

[0039] S3: Build a lightweight status assessment model based on the equipment maintenance knowledge base, evaluate the equipment status based on the fused features, and output a confidence index.

[0040] Furthermore, fault events that occurred in the past three years and their associated sensor data features are extracted from the equipment maintenance log, and the significance score of each feature within 1 hour before the fault occurs is calculated. The significance score is based on the correlation coefficient between the rate of change of the feature value and the time of fault occurrence. The features with the top 30% scores are marked as high priority. In the decision tree generation process, high-priority features are preferentially selected as split nodes. If there are no high-priority features among the optional features of the current node, the Gini coefficient is used for conventional splitting. Each leaf node maintains a ring buffer of length 50, which records the actual status (normal / warning / fault) and feature vector of the samples that have arrived at the node for the last 50 times, which is used for subsequent confidence calculation.

[0041] Furthermore, when a new sample arrives at a decision tree node, the historical state distribution of similar feature vectors is extracted from the circular buffer. If the current predicted state is "faulty," and the fault rate in the most recent 50 records in the buffer is less than 10%, the mean Euclidean distance between the current feature vector and the historical faulty samples is calculated. If the current distance exceeds two standard deviations of the historical mean, the confidence level is set to 50%; if it is within one standard deviation, the confidence level is 80%. For "normal" state predictions, the number of consecutive normal predictions in the buffer is counted. The confidence level increases by 2% for each additional normal prediction, up to a maximum of 95%.

[0042] S4: Use the physical rule verification engine to perform multi-level verification of the evaluation results and dynamically modify the confidence index based on the verification results.

[0043] Furthermore, in the primary verification stage, a safety threshold parameter table is obtained from the equipment control system in real time, including temperature alarm thresholds and vibration velocity effective value thresholds. If the temperature prediction value in the evaluation result exceeds the threshold, it is marked as a primary verification failure. In the advanced verification stage, a multi-parameter association rule library is constructed.

[0044] The parameter combinations in the evaluation results are matched against the rules. If the power calculation deviation exceeds ±8%, or the temperature drop rate and flow rate do not conform to the reasonable range of the corresponding relationship, the system is marked as a high-level verification failure. The verification results are output in three levels: pass, primary failure, and high-level failure.

[0045] Furthermore, when the assessment result is a primary failure, the original confidence is multiplied by a correction factor of 0.6; if it is a high-level failure, it is multiplied by 0.8. If the corrected confidence is lower than 20%, it is forced to be set to 20% as the lower limit. A verification log is generated for each correction operation, recording the failed rule entry, the original confidence, and the corrected value. If three consecutive correction operations involve the same rule entry, the rule base update check process is triggered: the trigger frequency of the rule in the past 30 days is automatically compared with the historical average level. If the deviation exceeds 200%, a rule review request is sent to the maintenance personnel.

[0046] S5: Trigger control instructions based on the revised confidence index, completing the closed-loop response from data analysis to device control

[0047] It should be noted that when the confidence level is greater than 85%, a predefined set of instructions is directly sent to the PLC control system, including reducing the spindle speed to 70% of the current value; starting the auxiliary cooling system to maximum power; and locking the feed rate to maintain the current value.

[0048] When the confidence level is between 60% and 85%, a level 3 alarm message is pushed to the HMI interface and the current control parameters are frozen for 30 seconds pending manual confirmation. During this period, the system recalculates the confidence level every 5 seconds and automatically executes the control command if it rises to above 85% for three consecutive times.

[0049] When the confidence level is less than 60%, the sensor calibration process is executed in sequence. The target sensor is zero-calibrated for 2 seconds. The standard test signal is injected to verify the channel integrity. The data difference before and after calibration is compared. If the deviation is greater than 15%, the sensor fault is marked. Data analysis is suspended during calibration. After the calibration is completed, the cached data of the last 30 seconds is reloaded for compensation analysis.

[0050] Embodiment 2, the second embodiment of the present invention, is different from the previous embodiment in that:

[0051] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0052] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0053] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.

[0054] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0055] Example 3, reference Figure 2 , as one embodiment of the present invention, provides an industrial equipment operation data analysis system based on artificial intelligence, including a data processing module, a feature fusion module, and a credibility verification control module;

[0056] The data processing module collects multi-source sensor data in real time, dynamically switches the filter window algorithm according to the load rate, and implements physical parameter constraints;

[0057] Feature fusion module, which performs cross-modal feature alignment and dynamic fusion on preprocessed data;

[0058] The credibility verification control module performs physical rule verification and closed-loop control in a hierarchical manner.

[0059] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. An artificial intelligence-based industrial equipment operation data analysis method, characterized by: include, Dynamically reduce noise on multi-source sensor data from industrial equipment, dynamically adjusting the filter window size based on the equipment's real-time operating parameters during the noise reduction process. Perform cross-modal feature alignment on the denoised multi-sensor data and extract time-frequency domain features respectively; A lightweight condition assessment model is built based on the equipment maintenance knowledge base to evaluate the equipment condition based on the fused features and output a confidence index. Use a physical rule verification engine to perform multi-level verification of the assessment results and dynamically modify the confidence index based on the verification results; Control instructions are triggered hierarchically based on the revised confidence index, completing a closed-loop response from data analysis to device control.

2. The method for analyzing industrial equipment operation data based on artificial intelligence according to claim 1, wherein: The dynamic adjustment of the filter window size includes: when the device load rate exceeds a preset threshold, using a first time window for data filtering; when the device load rate is lower than or equal to the preset threshold, using a second time window for data filtering, wherein the first time window is smaller than the second time window; During the filtering process, physical parameter constraints of the equipment are imposed. Specifically, values ​​in the sensor data that exceed the rated operating parameter range of the equipment are truncated; and components in the vibration sensor data whose frequency components exceed the set multiple of the critical speed of the equipment rotor are filtered out.

3. The method for analyzing industrial equipment operation data based on artificial intelligence according to claim 2, wherein: The cross-modal feature alignment includes performing a fast Fourier transform on the noise-reduced vibration signal to obtain an actual spectrum; and calculating the similarity between the actual spectrum and a theoretical vibration spectrum of the device, where the theoretical vibration spectrum is derived from structural parameters provided by the device manufacturer.

4. The method for analyzing industrial equipment operation data based on artificial intelligence according to claim 3, wherein: The cross-modal feature alignment also includes dynamic weight allocation during feature fusion, where when the similarity between the actual spectrum and the theoretical spectrum is lower than a set threshold, the weight ratio of the frequency domain feature in the fusion feature is reduced; When the peak value of the time domain feature exceeds the statistical range of historical data, the weight ratio of the time domain feature in the fusion feature is increased.

5. The method for analyzing industrial equipment operation data based on artificial intelligence according to claim 4, wherein: The lightweight state assessment model is constructed by using an improved decision tree model, where features with a high correlation with historical equipment fault records are preferentially selected as split nodes when splitting the decision tree nodes, and each leaf node additionally records the distribution of assessment results of the most recent N similar data samples; The confidence index calculation method includes obtaining the historical result distribution corresponding to the current evaluation result at the decision tree leaf node, and calculating the confidence value according to the degree of deviation between the current result and the historical distribution. The greater the deviation, the lower the confidence.

6. The method for analyzing industrial equipment operation data based on artificial intelligence according to claim 5, wherein: The multi-level verification includes primary verification and advanced verification; Among them, primary verification is to directly compare the assessment results with the equipment safety threshold; advanced verification is to verify whether the physical relationship between multiple parameters in the assessment results is consistent with the equipment operation mechanism; The dynamic correction of the confidence index includes multiplying the confidence index by a first correction coefficient when the evaluation result fails the primary verification; and multiplying the confidence index by a second correction coefficient when the evaluation result fails the advanced verification, wherein the second correction coefficient is smaller than the first correction coefficient.

7. The method for analyzing industrial equipment operation data based on artificial intelligence according to claim 6, wherein: The triggering control instruction includes directly sending an adjustment instruction to the device control system when the confidence index is greater than a first threshold; When the confidence index is between the first threshold and the second threshold, a review request is sent to the monitoring terminal and the current control parameters are maintained; When the confidence indicator is less than the second threshold, the sensor calibration procedure is triggered and the data analysis process is suspended.

8. A system using the artificial intelligence-based industrial equipment operation data analysis method according to any one of claims 1 to 7, characterized in that: Including data processing module, feature fusion module, and credibility verification control module; The data processing module collects multi-source sensor data in real time, dynamically switches the filter window algorithm according to the load rate, and implements physical parameter constraints; The feature fusion module performs cross-modal feature alignment and dynamic fusion on the preprocessed data; The credibility verification control module performs physical rule verification and closed-loop control in a hierarchical manner.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method for analyzing industrial equipment operation data based on artificial intelligence according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for analyzing industrial equipment operation data based on artificial intelligence according to any one of claims 1 to 7 are implemented.

Images