An intelligent manufacturing system based on digital twins

Abstract

The present invention belongs to the field of intelligent manufacturing technology. The present invention provides an intelligent manufacturing system based on digital twins, including: obtaining intelligent manufacturing production line data in a historical period, and performing production line operation characteristic analysis to evaluate whether the intelligent manufacturing production line is in a healthy state; if it is in an unhealthy state, performing a fluctuation analysis on the production efficiency of the production line in multiple historical periods, identifying the production efficiency change pattern of the unhealthy production line in multiple historical periods, and judging whether a production efficiency fluctuation signal is generated; if a production efficiency fluctuation signal is generated, analyzing the dominant frequency when the production efficiency fluctuation signal is generated and the vibration frequency when the vibration signal of the fault part is generated through Fourier transform, and judging the degree of coupling; the present invention locates the fault source that causes the production efficiency fluctuation, converts passive maintenance into active intervention, effectively avoids sudden shutdowns, and significantly reduces the risk of unplanned shutdowns.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent manufacturing technology, and specifically relates to an intelligent manufacturing system based on digital twins. Background Art

[0002] With the rapid development of industrial and intelligent manufacturing technologies, the complexity and intelligence level of production line equipment have been significantly improved. In the field of intelligent manufacturing, production line health management and fault prediction are the core links to ensure production continuity and efficiency.

[0003] However, existing technologies have the following defects. On the one hand, for smart manufacturing production lines, the status of the production lines is mainly evaluated by manual inspections, and there is a lack of integrated analysis of multi-dimensional data such as the distribution of fault locations and production efficiency, resulting in highly subjective and inaccurate health assessment results. On the other hand, the maintenance mode for production lines is mainly post-repair or regular inspections, with a high risk of sudden downtime. Fault location relies on "trial and error troubleshooting", making it difficult to quickly locate the specific source of the fault. In addition, existing technologies have not integrated digital twin technology, and cannot dynamically simulate the wear process of key components and predict the remaining life, resulting in maintenance decisions lagging behind the actual degradation status of the equipment, and the risk of sudden downtime is difficult to effectively control.

[0004] To this end, the present invention provides an intelligent manufacturing system based on digital twins. Summary of the Invention

[0005] In order to make up for the deficiencies of the prior art, at least one technical problem raised in the background technology is solved.

[0006] The technical solution adopted by the present invention to solve the technical problem is: an intelligent manufacturing system based on digital twins, which specifically includes the following modules:

[0007] Health Assessment Module: This module obtains historical data on smart manufacturing production lines and analyzes their operational characteristics to assess whether they are in a healthy state. This data includes the number of production line failures, the location of the failures, and the output of the production line.

[0008] Efficiency Fluctuation Analysis Module: If the system is in an unhealthy state, the production efficiency of the production line in the historical period is determined based on the output of the unhealthy production line in the historical period and the production time corresponding to the output of the production line. The system also performs a fluctuation analysis on the production efficiency of the production line in multiple historical periods to determine whether a production efficiency fluctuation signal should be generated.

[0009] Coupling-related modules: If a production efficiency fluctuation signal is generated, the corresponding production line fault location in the unhealthy production line is extracted. The dominant frequency when the production efficiency fluctuation signal is generated and the vibration frequency when the vibration signal of the fault location are generated are analyzed through Fourier transform to determine the degree of coupling.

[0010] Wear analysis module: If the degree of coupling is high, the digital twin platform is used to simulate the wear process of the fault part with a high degree of coupling, and a long-short-term memory network is used to construct a wear degradation model. The wear volume threshold of the key parts is input into the wear degradation model to determine the remaining wear period.

[0011] As a further solution of the present invention: the specific process of evaluating whether the intelligent manufacturing production line is in a healthy state is:

[0012] Based on any production line, analyze the intelligent manufacturing production line data within the historical period to obtain the production line failure rate ratio and production line failure location ratio;

[0013] The comprehensive health score is obtained by weighted summing the ratio of production line failure times and the ratio of production line failure locations.

[0014] If the comprehensive health score is greater than or equal to the comprehensive health score threshold, it means that the smart manufacturing production line is in an unhealthy state.

[0015] As a further solution of the present invention: the process of obtaining the production line failure ratio is as follows:

[0016] Count the number of production line failures in the historical period, perform subtraction processing on the number of failures compared with the standard value, and then perform ratio processing on the number of failures compared with the standard value to obtain the production line failure ratio;

[0017] As a further solution of the present invention: the process of obtaining the production line failure location ratio is as follows:

[0018] The number of production line fault locations within the historical period is counted, and the difference between the number and the standard value of the number of production line fault locations is processed. The difference is then compared with the standard value of the number of production line fault locations to obtain the production line fault location ratio.

[0019] As a further solution of the present invention: the specific process of determining whether to generate a production efficiency fluctuation signal is as follows:

[0020] The intelligent manufacturing production line in an unhealthy state is recorded as an unhealthy production line. The mean and standard deviation formulas are used to calculate the production efficiency mean and production efficiency standard deviation of the current unhealthy production line.

[0021] The efficiency fluctuation value of the current unhealthy production line is obtained by comparing the standard deviation of the production efficiency of the current unhealthy production line with the mean production efficiency.

[0022] If the efficiency fluctuation value of the current unhealthy production line is greater than or equal to the efficiency fluctuation threshold, a production efficiency fluctuation signal is generated.

[0023] As a further solution of the present invention: the specific process of determining the degree of coupling is as follows:

[0024] Extract the fault part of the production line when the production efficiency fluctuation signal is generated, collect the vibration signal of the fault part of the production line, and analyze the production efficiency fluctuation signal. and vibration signals of faulty parts of the production line Perform Fourier transform and analyze to obtain amplitude correlation value and frequency difference ratio;

[0025] Perform weighted summation of the amplitude correlation value and the frequency difference ratio to obtain a coupling correlation value;

[0026] If the coupling correlation value is less than the coupling correlation threshold, it means that the dominant frequency when the production efficiency fluctuation signal is generated is highly coupled with the vibration frequency when the vibration signal of the faulty part of the current production line is generated.

[0027] As a further solution of the present invention: the process of obtaining the amplitude-related value is:

[0028] Signals of production efficiency fluctuations and vibration signals of faulty parts of the production line Perform Fourier transform to obtain frequency domain signal and , for frequency domain signals and The cross-spectral density of the production efficiency fluctuation signal and the vibration signal is calculated by multiplying the conjugate of ;

[0029] Calculate the amplitude of the cross-spectral density corresponding to the current frequency based on the cross-spectral density of the production efficiency fluctuation signal and the vibration signal;

[0030] The amplitude of the current frequency cross-spectral density is subtracted from the amplitude standard value, and the absolute value is taken. The difference is then compared with the amplitude standard value to obtain the amplitude correlation value.

[0031] As a further solution of the present invention: the process of obtaining the frequency difference ratio is:

[0032] Obtain the frequency domain amplitude spectrum of the production efficiency fluctuation signal after Fourier transformation, and extract the frequency with the largest amplitude from it, which is the dominant frequency of the production efficiency fluctuation signal;

[0033] Obtain the frequency domain amplitude spectrum of the vibration signal at the fault location after Fourier transformation to determine the vibration frequency of the fault location;

[0034] The dominant frequency of production efficiency is subtracted from the vibration frequency of the fault location, and the absolute value is taken to obtain the frequency difference. The frequency difference is then compared with the standard value of the frequency difference to obtain the frequency difference ratio.

[0035] As a further solution of the present invention: the process of obtaining the vibration frequency of the fault location is:

[0036] The frequency domain amplitude spectrum of the vibration signal at the fault location is obtained by Fourier transform. The fault characteristic frequency is extracted through time-frequency analysis of the vibration signal. The frequency characteristic frequency is subtracted from the standard frequency value and the absolute value is taken. The frequency error ratio is then obtained by comparing the frequency characteristic frequency with the standard frequency value.

[0037] If the frequency error ratio is within the frequency error ratio range, the frequency standard value corresponding to the frequency signal is the vibration frequency of the fault location; otherwise, the fault characteristic frequency is extracted through time-frequency analysis of the vibration signal, which is the vibration frequency of the fault location.

[0038] As a further solution of the present invention: the process of obtaining the remaining wear period is:

[0039] According to Archard wear theory, the fault location with high coupling degree is recorded as the key location, and the wear volume of the key location is calculated;

[0040] The wear volume of key parts in multiple historical periods is used as a training set to train the long short-term memory network, so that the long short-term memory network model can learn the changing pattern of the wear volume of key parts when the production efficiency fluctuation signal is generated. The wear volume threshold of key parts is input into the long short-term memory network after training, and the time point when the wear volume of key parts reaches the wear volume threshold is output. The result is subtracted from the time point when the production efficiency fluctuation signal is generated to obtain the remaining wear period.

[0041] The beneficial effects of the present invention are as follows:

[0042] The present invention obtains smart manufacturing production line data and preprocesses it, and analyzes the production line operation characteristics of the preprocessed smart manufacturing production line data to evaluate whether the smart manufacturing production line is in a healthy state. The production line data includes the number of production line failures, the location of the production line failure, and the production line output. If the smart manufacturing production line is in an unhealthy state, the smart manufacturing production line in an unhealthy state will be recorded as an unhealthy production line. According to the output of the unhealthy production line in the historical period and the production time corresponding to the production line output, the production efficiency of the production line in the historical period is determined, and the fluctuation analysis of the production efficiency of the production line in multiple historical periods is performed to identify the production efficiency change pattern of the unhealthy production line in multiple historical periods and determine whether to generate a production efficiency fluctuation signal. The present invention realizes the refined health management of the smart manufacturing production line through multi-dimensional data fusion and intelligent analysis, quantitatively evaluates the health status of the production line, improves the accuracy of fault warning, quantifies production stability through the coefficient of variation, determines the implicit efficiency decline trend of the production line, traces the root cause of efficiency fluctuations, and provides data support for preventive maintenance.

[0043] This method uses Fourier transform to analyze the dominant frequency of the production efficiency fluctuation signal and the vibration frequency of the faulty part's vibration signal. The coupling between the two is then determined. If the coupling is high, the faulty part with the high coupling is extracted and designated as the key part. The wear process of the key part is simulated using a digital twin platform. A wear degradation model is constructed using a long-short-term memory network. The wear volume threshold of the key part is input into the wear degradation model. The time point at which the wear volume of the key part reaches the wear volume threshold is output, and the remaining wear period is determined. The key part is then replaced and repaired before the remaining wear period ends. This method establishes an intelligent operation and maintenance system for intelligent manufacturing production lines. Spectral correlation analysis is performed between production efficiency fluctuations and equipment vibration characteristics. Through quantitative evaluation of cross-spectral density and frequency difference ratio, the fault source causing the efficiency fluctuation can be accurately located, improving fault tracing efficiency. The wear degradation model constructed on the digital twin platform can predict the remaining life of key components in advance, transforming passive maintenance into active intervention, effectively avoiding sudden downtime, significantly reducing the risk of unplanned downtime, and improving production line operational stability, providing a full lifecycle health management solution for intelligent manufacturing systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The present invention will be further described below with reference to the accompanying drawings.

[0045] Figure 1 This is a system block diagram of an intelligent manufacturing system based on digital twins according to an embodiment of the present invention;

[0046] Figure 2 This is a flowchart of the steps of an intelligent manufacturing method based on digital twins in an embodiment of the present invention. DETAILED DESCRIPTION

[0047] In order to make the technical means, creative features, objectives and effects achieved by the present invention easier to understand, the present invention is further described below in conjunction with specific implementation methods.

[0048] Example 1, please refer to Figure 1 As shown, an intelligent manufacturing system based on digital twins according to an embodiment of the present invention includes the following modules:

[0049] Health Assessment Module: This module obtains and pre-processes historical smart manufacturing production line data. It then analyzes the operational characteristics of the pre-processed data to assess whether the smart manufacturing production line is in a healthy state. The production line data includes the number of production line failures, the location of the failures, and the output of the production line.

[0050] Over a historical period, we extracted the number and location of intelligent manufacturing line failures from system logs. We used fiber optic sensors to obtain output from these lines. We then cleaned all collected data, removing outliers and filling in missing values. To eliminate the impact of varying data dimensions and value ranges, we normalized the data using the Min-Max normalization method, mapping the processed data to a range of 0 to 1.

[0051] For any production line, count the number of production line failures in the historical period, perform subtraction processing on the number of production line failures and the standard value of failure count, and then perform ratio processing on the number of production line failures with the standard value of failure count to obtain the production line failure ratio;

[0052] Count the number of production line fault locations within a historical period, perform subtraction processing on the number of production line fault locations and the standard value of the number of production line fault locations, and then perform ratio processing on the number of production line fault locations and the standard value of the number of production line fault locations to obtain the production line fault location ratio;

[0053] It should be noted that the standard values ​​for the number of failures and the number of fault locations on a production line are set by those skilled in the art based on historical experience;

[0054] Using a weighted comprehensive scoring method, the ratio of production line failures and the ratio of production line failure locations are weighted and summed to obtain a comprehensive health score for the intelligent manufacturing production line. The weight distribution is determined by technical personnel in this field based on industry experience.

[0055] In some embodiments, the comprehensive health score is compared with the comprehensive health score threshold, and the specific comparison process is:

[0056] If the comprehensive health score is greater than or equal to the comprehensive health score threshold, it means that the smart manufacturing production line is in an unhealthy state;

[0057] If the comprehensive health score is less than the comprehensive health score threshold, it means that the smart manufacturing production line is in a healthy state;

[0058] Efficiency Fluctuation Analysis Module: If a smart manufacturing production line is in an unhealthy state, it will be recorded as an unhealthy production line. Based on the output of the unhealthy production line in the historical period and the production time corresponding to the production line output, the production efficiency of the production line in the historical period is determined. The production efficiency of the production line in multiple historical periods is analyzed for fluctuations, and the production efficiency change pattern of the unhealthy production line in multiple historical periods is identified to determine whether a production efficiency fluctuation signal is generated.

[0059] Extract the production line output of all healthy smart manufacturing production lines over multiple historical cycles;

[0060] Based on any unhealthy production line, according to the output of the unhealthy production line in the historical period, the production time corresponding to the output of the unhealthy production line in the historical period is obtained through the time timer, and the production efficiency SX of the unhealthy production line in the historical period is calculated. The specific calculation formula is:

[0061] ;

[0062] Where CL represents the output of the unhealthy production line in the historical period, Indicates the production time corresponding to the output of unhealthy production lines in the historical period;

[0063] The production line production efficiency of multiple historical cycles of the unhealthy production line is integrated into a production line production efficiency data group. The mean of all data in the production line production efficiency data group is calculated using the mean formula to obtain the mean production efficiency of the current unhealthy production line. The standard deviation of all data in the production line production efficiency data group is calculated using the standard deviation formula to obtain the standard deviation of the production efficiency of the current unhealthy production line.

[0064] The efficiency fluctuation value of the current unhealthy production line is obtained by comparing the standard deviation of the production efficiency of the current unhealthy production line with the mean production efficiency.

[0065] It should be noted that the efficiency fluctuation value is essentially the coefficient of variation in statistics, which is used to measure the degree of dispersion of data relative to the mean. The smaller the efficiency fluctuation value, the smaller the fluctuation in production efficiency of the production line over multiple historical cycles, and the higher the stability. The larger the efficiency fluctuation value, the more drastic the fluctuation in production efficiency of the production line, and the lower the stability. Even if the production line passes the health assessment (no significant faults), if the efficiency fluctuation value is too high, it indicates that the production line's production efficiency is on a downward trend.

[0066] In some embodiments, the efficiency fluctuation value of the current unhealthy production line is compared with the efficiency fluctuation threshold. The specific comparison process is:

[0067] If the efficiency fluctuation value of the current unhealthy production line is greater than or equal to the efficiency fluctuation threshold, it means that the production efficiency of the corresponding unhealthy production line has fluctuated over multiple historical periods, and a production efficiency fluctuation signal is generated;

[0068] If the efficiency fluctuation value of the current unhealthy production line is less than the efficiency fluctuation threshold, it means that the production efficiency changes of the corresponding unhealthy production line in multiple historical cycles are stable, and a production efficiency stability signal is generated;

[0069] The technical solution of this embodiment is: obtaining the smart manufacturing production line data in the historical period and preprocessing it, and analyzing the production line operation characteristics of the preprocessed smart manufacturing production line data to evaluate whether the smart manufacturing production line is in a healthy state, wherein the production line data includes the number of production line failures, the location of the production line failure, and the production line output. If the smart manufacturing production line is in an unhealthy state, the smart manufacturing production line in the unhealthy state is recorded as an unhealthy production line. According to the output of the unhealthy production line in the historical period and the production time corresponding to the production line output, the production efficiency of the production line in the historical period is determined, and the production efficiency of the production line in multiple historical periods is analyzed for fluctuations, the production efficiency change pattern of the unhealthy production line in multiple historical periods is identified, and it is determined whether a production efficiency fluctuation signal is generated. The present invention realizes the refined health management of the smart manufacturing production line through multi-dimensional data fusion and intelligent analysis, quantitatively evaluates the health status of the production line, improves the accuracy of fault warning, quantifies production stability through the coefficient of variation, determines the implicit efficiency decline trend of the production line, traces the root cause of efficiency fluctuations, and provides data support for preventive maintenance;

[0070] Example 2, please refer to Figure 1 As shown, an intelligent manufacturing system based on digital twins according to an embodiment of the present invention includes the following modules:

[0071] Coupling-related modules: If a production efficiency fluctuation signal is generated, the dominant frequency of the production efficiency fluctuation signal and the vibration frequency of the fault location in the production line when the production efficiency fluctuation signal is generated are analyzed through Fourier transform to determine the degree of coupling between the two.

[0072] Use an acceleration sensor to collect vibration signals at the fault location on the production line;

[0073] It can be understood that the sampling frequencies of the production efficiency fluctuation signal and the fault location fluctuation signal are consistent, and the sampling frequency needs to meet the Nyquist criterion;

[0074] Based on any fault location;

[0075] Signals of production efficiency fluctuations and vibration signals of the fault location Perform Fourier transform to obtain frequency domain signal and , specifically:

[0076] , ;

[0077] Where f is the frequency point, j is the imaginary unit, N is the total number of sampling points, t=0,1,…,N-1;

[0078] For frequency domain signals and , and calculate the cross-spectral density of the production efficiency fluctuation signal and the vibration signal. , the specific calculation formula is:

[0079] Where, is a frequency domain signal The conjugate of is the phase difference between the production efficiency fluctuation signal and the vibration signal at frequency f;

[0080] Obtain the frequency domain amplitude spectrum corresponding to the Fourier transform of the production efficiency fluctuation signal, and extract the frequency with the largest amplitude from it, which is the dominant frequency of the production efficiency fluctuation signal ;

[0081] Obtain the frequency domain amplitude spectrum corresponding to the vibration signal of the fault location after Fourier transform, and extract the fault characteristic frequency through time-frequency analysis of the vibration signal. Subtract the frequency characteristic frequency from the standard frequency value and take the absolute value. Then perform ratio processing with the standard frequency value to obtain the frequency error ratio.

[0082] It should be noted that the frequency standard value is set by those skilled in the art by calculating the theoretical characteristic frequency through a physical model;

[0083] If the frequency error ratio is within the frequency error ratio range, the frequency signal The corresponding frequency standard value is the vibration frequency of the fault location ;

[0084] If the frequency error ratio is not within the frequency error ratio range, the fault characteristic frequency is extracted through time-frequency analysis of the vibration signal, which is the vibration frequency of the fault location. ;

[0085] The dominant frequency of production efficiency is subtracted from the vibration frequency of the fault location, and the absolute value is taken to obtain the frequency difference. The frequency difference is then compared with the standard value of the frequency difference to obtain the frequency difference ratio.

[0086] It should be noted that the standard value of the frequency difference is set by those skilled in the art by calculating the theoretical characteristic frequency through a physical model;

[0087] According to the cross-spectral density of production efficiency fluctuation signal and vibration signal , calculate the amplitude of the cross-spectral density corresponding to the current frequency. The specific calculation formula is:

[0088] ;

[0089] Where, and denote the real and imaginary parts respectively;

[0090] The amplitude of the current frequency cross-spectral density is subtracted from the amplitude standard value, and the absolute value is taken, and then the amplitude is compared with the amplitude standard value to obtain the amplitude correlation value;

[0091] It should be noted that the amplitude standard value is set by those skilled in the art based on historical experience;

[0092] Perform weighted summation of the amplitude correlation value and the frequency difference ratio to obtain a coupling correlation value;

[0093] It can be understood that the amplitude correlation value reflects the relative size of the cross-spectral density amplitude. The smaller the amplitude correlation value, the stronger the amplitude correlation between the production efficiency fluctuation signal and the vibration signal at the current frequency. The frequency difference ratio reflects the frequency consistency. The smaller the frequency difference ratio, the smaller the difference between the dominant frequency of production efficiency and the vibration frequency of the fault location, indicating that the production efficiency fluctuation is directly caused by the vibration of the fault location.

[0094] In some embodiments, the coupling-related value is compared with the coupling-related threshold. The specific comparison process is:

[0095] If the coupling correlation value is less than the coupling correlation threshold, it means that the dominant frequency of the production efficiency fluctuation is highly coupled with the vibration frequency of the current fault location;

[0096] If the coupling correlation value is greater than or equal to the coupling correlation threshold, it means that the dominant frequency of the production efficiency fluctuation is poorly coupled with the vibration frequency of the current fault location;

[0097] The purpose of analyzing the degree of coupling between the dominant frequency of production efficiency fluctuations and the vibration frequency of the fault location is to:

[0098] Function 1: Through frequency correlation analysis, a dynamic mapping relationship is established between production efficiency and equipment operating status. When a fault occurs in a certain part, its vibration frequency will affect the overall operating status through mechanical conduction, causing fluctuations in production efficiency. By comparing whether the two frequencies match, it is helpful to determine the specific fault location causing the production efficiency fluctuation, avoiding the inefficiency of traditional "trial and error" troubleshooting.

[0099] Function 2: By analyzing the coupling degree between the dominant frequency of production efficiency fluctuations and the vibration frequency of the fault location, it is helpful to identify early equipment failures before production efficiency drops significantly, turning reactive maintenance into proactive maintenance and reducing the risk of sudden downtime;

[0100] Wear analysis module: If the coupling degree is high, the faulty parts with high coupling degree are extracted and recorded as key parts. The wear process of key parts is simulated through the digital twin platform. A wear degradation model is constructed using a long-short-term memory network. The wear volume threshold of the key parts is input into the wear degradation model. The time point when the wear volume of the key parts reaches the wear volume threshold is output, and the remaining wear period is determined. The key parts are replaced and repaired before the remaining wear period ends.

[0101] According to Archard wear theory, the specific formula for calculating the wear volume of key parts is:

[0102] ; Where V is the wear volume of the key parts, K is the wear coefficient set according to the material properties of the key parts, is the sliding friction force calculated by finite element analysis, H is the material hardness, and L is the sliding distance;

[0103] On the digital twin platform, simulate the operating conditions of key parts in the intelligent manufacturing production line when the production efficiency fluctuation signal is generated. Preset the simulation operation period, divide the simulation operation period into several simulation time points with equal time intervals, run the simulation model, and output the wear volume of key parts at the simulation time point;

[0104] Compare the wear of key parts output at the simulation time point with the actual wear of key parts in the historical period, and make the simulation model output as close to the real value as possible by minimizing the mean square error;

[0105] A long short-term memory (LSTM) network was used to construct a wear degradation model. The wear volume of key parts over multiple historical cycles was used as a training set for the LSTM network, which contains two hidden layers, each with 128 neurons. Through training and optimization of network parameters, the LSTM model was able to learn the changing patterns of wear volume in key parts when production efficiency fluctuation signals were generated. The wear volume threshold for key parts was input into the trained LSTM network, and the time point at which the wear volume of key parts reached the threshold was output. This time point was subtracted from the time point when the production efficiency fluctuation signal was generated to obtain the remaining wear period. Key parts could then be replaced and repaired before the end of the remaining wear period to avoid production line shutdowns due to sudden failures.

[0106] The technical solution of this embodiment is: if a production efficiency fluctuation signal is generated, the production line fault part in the production line when the production efficiency fluctuation signal is generated is analyzed by Fourier transform, and the dominant frequency when the production efficiency fluctuation signal is generated and the vibration frequency when the fault part vibration signal is generated are analyzed, and the degree of coupling between the two is judged. If the coupling degree is high, the fault part with a high coupling degree is extracted and recorded as the key part. The wear process of the key part is simulated through the digital twin platform, and a long short-term memory network is used to construct a wear degradation model. The wear volume threshold of the key part is input into the wear degradation model, and the time point when the wear volume of the key part reaches the wear volume threshold is output to confirm The remaining wear period is determined, and key parts are replaced and repaired before the remaining wear period ends. This invention builds an intelligent operation and maintenance system for intelligent manufacturing production lines, performs spectral correlation analysis on production efficiency fluctuations and equipment vibration characteristics, and through quantitative evaluation of cross-spectral density and frequency difference ratio, can accurately locate the fault source causing efficiency fluctuations, thereby improving the efficiency of fault tracing. The wear degradation model is constructed on the digital twin platform, which can predict the remaining life of key components in advance, transform passive maintenance into active intervention, effectively avoid sudden downtime, significantly reduce the risk of unplanned downtime, improve the stability of production line operation, and provide a full life cycle health management solution for intelligent manufacturing systems.

[0107] Example 3, please refer to Figure 2 As shown, an intelligent manufacturing system based on digital twins according to an embodiment of the present invention includes the following steps:

[0108] Step 1: Obtain and preprocess the historical data of smart manufacturing production lines. Analyze the operational characteristics of the preprocessed data to assess whether the smart manufacturing production lines are in a healthy state. The production line data includes the number of production line failures, the location of the failures, and the output of the production line.

[0109] Step 2: If the smart manufacturing production line is in an unhealthy state, record it as an unhealthy production line. Determine the production efficiency of the production line in the historical period based on the output of the unhealthy production line in combination with the production time corresponding to the output of the production line. Perform a fluctuation analysis on the production efficiency of the production line in multiple historical periods to identify the change pattern of the production efficiency of the unhealthy production line in multiple historical periods and determine whether to generate a production efficiency fluctuation signal.

[0110] Step 3: If a production efficiency fluctuation signal is generated, the fault location in the production line at the time the production efficiency fluctuation signal is generated is analyzed using Fourier transform to determine the dominant frequency of the production efficiency fluctuation signal and the vibration frequency of the fault location vibration signal, and to determine the degree of coupling between the two.

[0111] Step 4: If the coupling degree is high, extract the faulty parts with a high coupling degree and record them as key parts. Use the digital twin platform to simulate the wear process of the key parts, use the long short-term memory network to build a wear degradation model, input the wear volume threshold of the key parts into the wear degradation model, output the time point when the wear volume of the key parts reaches the wear volume threshold, determine the remaining wear period, and replace and repair the key parts before the end of the remaining wear period.

[0112] The basic principles, main features, and advantages of the present invention are shown and described above. Those skilled in the art should understand that the present invention is not limited to the foregoing embodiments. The foregoing embodiments and descriptions are merely illustrative of the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and modifications are intended to fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent manufacturing system based on digital twins, characterized by: Specifically including the following modules: Health Assessment Module: This module obtains historical data on smart manufacturing production lines and analyzes their operational characteristics to assess whether they are in a healthy state. This data includes the number of production line failures, the location of the failures, and the output of the production line. Efficiency Fluctuation Analysis Module: If the system is in an unhealthy state, the production efficiency of the production line in the historical period is determined based on the output of the unhealthy production line in the historical period and the production time corresponding to the output of the production line. The system also performs a fluctuation analysis on the production efficiency of the production line in multiple historical periods to determine whether a production efficiency fluctuation signal should be generated. Coupling-related modules: If a production efficiency fluctuation signal is generated, the corresponding production line fault location in the unhealthy production line is extracted. The dominant frequency when the production efficiency fluctuation signal is generated and the vibration frequency when the vibration signal of the fault location are generated are analyzed through Fourier transform to determine the degree of coupling. The specific process of judging the degree of coupling is as follows: Extract the fault part of the production line when the production efficiency fluctuation signal is generated, collect the vibration signal of the fault part of the production line, and analyze the production efficiency fluctuation signal. and vibration signals of faulty parts of the production line Perform Fourier transform and analyze to obtain amplitude correlation value and frequency difference ratio; Perform weighted summation of the amplitude correlation value and the frequency difference ratio to obtain a coupling correlation value; If the coupling correlation value is less than the coupling correlation threshold, it means that the dominant frequency when the production efficiency fluctuation signal is generated is highly coupled with the vibration frequency when the vibration signal of the faulty part of the current production line is generated; Wear analysis module: If the degree of coupling is high, the digital twin platform is used to simulate the wear process of the fault part with a high degree of coupling, and a long-short-term memory network is used to construct a wear degradation model. The wear volume threshold of the key parts is input into the wear degradation model to determine the remaining wear period.

2. The digital twin-based intelligent manufacturing system according to claim 1, characterized in that: The specific process of evaluating whether the intelligent manufacturing production line is in a healthy state is as follows: Based on any production line, analyze the intelligent manufacturing production line data within the historical period to obtain the production line failure rate ratio and production line failure location ratio; The comprehensive health score is obtained by weighted summing the ratio of production line failure times and the ratio of production line failure locations. If the comprehensive health score is greater than or equal to the comprehensive health score threshold, it means that the smart manufacturing production line is in an unhealthy state.

3. The digital twin-based intelligent manufacturing system according to claim 2, characterized in that: The process of obtaining the production line failure ratio is as follows: The number of production line failures in the historical period is counted, and the difference between the number and the standard value of the number of failures is processed, and then the ratio is processed with the standard value of the number of failures to obtain the production line failure ratio.

4. The digital twin-based intelligent manufacturing system according to claim 2, characterized in that: The process of obtaining the failure location ratio of the production line is as follows: The number of production line fault locations within the historical period is counted, and the difference between the number and the standard value of the number of production line fault locations is processed. The difference is then compared with the standard value of the number of production line fault locations to obtain the production line fault location ratio.

5. The digital twin-based intelligent manufacturing system according to claim 2, characterized in that: The specific process of determining whether to generate a production efficiency fluctuation signal is as follows: The intelligent manufacturing production line in an unhealthy state is recorded as an unhealthy production line. The mean and standard deviation formulas are used to calculate the production efficiency mean and production efficiency standard deviation of the current unhealthy production line. The efficiency fluctuation value of the current unhealthy production line is obtained by comparing the standard deviation of the production efficiency of the current unhealthy production line with the mean production efficiency. If the efficiency fluctuation value of the current unhealthy production line is greater than or equal to the efficiency fluctuation threshold, a production efficiency fluctuation signal is generated.

6. The digital twin-based intelligent manufacturing system according to claim 1, characterized in that: The process of obtaining the amplitude-related value is as follows: Signals of production efficiency fluctuations and vibration signals of faulty parts of the production line Perform Fourier transform to obtain frequency domain signal and , for frequency domain signals and The cross-spectral density of the production efficiency fluctuation signal and the vibration signal is calculated by multiplying the conjugate of ; Calculate the amplitude of the cross-spectral density corresponding to the current frequency based on the cross-spectral density of the production efficiency fluctuation signal and the vibration signal; The amplitude of the current frequency cross-spectral density is subtracted from the amplitude standard value, and the absolute value is taken. The difference is then compared with the amplitude standard value to obtain the amplitude correlation value.

7. The digital twin-based intelligent manufacturing system according to claim 1, characterized in that: The process of obtaining the frequency difference ratio is as follows: Obtain the frequency domain amplitude spectrum of the production efficiency fluctuation signal after Fourier transformation, and extract the frequency with the largest amplitude from it, which is the dominant frequency of the production efficiency fluctuation signal; Obtain the frequency domain amplitude spectrum of the vibration signal at the fault location after Fourier transformation to determine the vibration frequency of the fault location; The dominant frequency of production efficiency is subtracted from the vibration frequency of the fault location, and the absolute value is taken to obtain the frequency difference. The frequency difference is then compared with the standard value of the frequency difference to obtain the frequency difference ratio.

8. The digital twin-based intelligent manufacturing system according to claim 7, characterized in that: The process of obtaining the vibration frequency of the fault location is as follows: The frequency domain amplitude spectrum of the vibration signal at the fault location is obtained by Fourier transform. The fault characteristic frequency is extracted through time-frequency analysis of the vibration signal. The frequency characteristic frequency is subtracted from the standard frequency value and the absolute value is taken. The frequency error ratio is then obtained by comparing the frequency characteristic frequency with the standard frequency value. If the frequency error ratio is within the frequency error ratio range, the frequency standard value corresponding to the frequency signal is the vibration frequency of the fault location; otherwise, the fault characteristic frequency is extracted through time-frequency analysis of the vibration signal, which is the vibration frequency of the fault location.

9. The digital twin-based intelligent manufacturing system according to claim 1, characterized in that: The process of obtaining the remaining wear period is as follows: According to Archard wear theory, the fault location with high coupling degree is recorded as the key location, and the wear volume of the key location is calculated; The wear volume of key parts in multiple historical periods is used as a training set to train the long short-term memory network, so that the long short-term memory network model can learn the changing pattern of the wear volume of key parts when the production efficiency fluctuation signal is generated. The wear volume threshold of key parts is input into the long short-term memory network after training, and the time point when the wear volume of key parts reaches the wear volume threshold is output. The result is subtracted from the time point when the production efficiency fluctuation signal is generated to obtain the remaining wear period.

Images