Mold life prediction method based on simulation
Through simulation simulation technology combined with digital twins and finite element simulation, the industrial Internet is used to predict mold life, which solves the problem of singleness and incomplete factors of existing mold life prediction methods, and improves the accuracy of mold replacement and the stability of equipment .
Patent Information
- Application Number
- CN202411134257.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-19
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2044-08-19
AI Technical Summary
The existing mold life prediction method is relatively single, and the considerations are not comprehensive, resulting in poor accuracy, practicality and reliability of mold life prediction under different working conditions, resulting in poor mold replacement accuracy.
Using simulation-based simulation methods, we can obtain the key influencing factors of mold fatigue damage, build the mold stamping twin model through digital twin technology and finite element simulation, combine the industrial Internet to conduct positive sample retrieval and predetermined attenuation factor testing, and conduct weight fusion analysis to improve Accuracy and scientificity of mold life prediction.
Improve the accuracy and scientificity of mold life prediction, enhance the accuracy and rationality of mold replacement, reduce resource waste, and avoid shutdown losses caused by sudden mold failure, ensuring the safe and stable operation of stamping equipment .
Smart Images

Figure CN118657030B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of life prediction, and in particular to a mold life prediction method based on simulation. Background Art
[0002] With the rapid development of industrial production and manufacturing technology, the importance of molds in the manufacturing of various products has become increasingly prominent. However, with the continuous expansion of production scale and the complexity of process flow, the monitoring and management of mold life are also facing increasing challenges. The traditional mold life prediction method has problems such as poor accuracy, practicality and reliability of mold life prediction under different working conditions due to the relatively single prediction method and incomplete consideration of factors, which makes it difficult to meet the needs of modern industrial production for efficiency, accuracy and safety. Summary of the invention
[0003] The present application provides a mold life prediction method based on simulation, which is used to solve the technical problem that the existing mold life prediction method has a relatively single prediction method and does not take into account all factors, resulting in poor accuracy, practicality and reliability of mold life prediction under different working conditions, resulting in poor accuracy of mold replacement.
[0004] In view of the above problems, the present application provides a mold life prediction method based on simulation.
[0005] The present application provides a mold life prediction method based on simulation, the method comprising:
[0006] The key influencing factors of mold fatigue damage are obtained, and features of the target mold are extracted according to the key influencing factors to obtain a mold feature set; based on the digital twin, simulation is performed in a three-dimensional simulation platform according to the mold attribute information and the mold feature set to construct a mold stamping twin model; using finite element simulation technology, the mold stamping twin model is locally segmented, and stamping is simulated based on multiple local twin models to determine the predicted life of the first mold, wherein the predicted mold life is the number of cyclic loads; using the mold attribute information as a static constraint and the mold feature set as a dynamic constraint, positive sample retrieval is performed based on the industrial Internet to obtain an initial predicted mold life; performing an attenuation test on the initial predicted mold life based on a predetermined attenuation factor and a predetermined risk threshold to obtain a second predicted mold life; performing a fusion analysis on the first predicted mold life and the second predicted mold life according to a predetermined strategy, and outputting the predicted mold life of the target mold.
[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages:
[0008] This application obtains key influencing factors of mold fatigue damage, extracts features of the target mold based on the key influencing factors, and obtains a mold feature set; based on digital twins, simulation is performed in a three-dimensional simulation platform based on mold attribute information and mold feature sets to construct a mold stamping twin model; finite element simulation technology is used to locally segment the mold stamping twin model, and simulate stamping based on multiple local twin models to determine the predicted life of the first mold, wherein the predicted mold life is the number of cyclic loads; using mold attribute information as a static constraint and the mold feature set as a dynamic constraint, positive sample retrieval is performed based on the industrial Internet to obtain an initial predicted mold life; an attenuation test is performed on the initial predicted mold life based on a predetermined attenuation factor and a predetermined risk threshold to obtain a second predicted mold life; a fusion analysis is performed on the first predicted mold life and the second predicted mold life according to a predetermined strategy, and the predicted mold life of the target mold is output. The present invention solves the technical problem that the existing mold life prediction method has a relatively single prediction method and does not take into account all factors, resulting in poor accuracy, practicality and reliability of mold life prediction under different working conditions, and poor accuracy of mold replacement. By combining the key influencing factors of mold fatigue damage, digital twin technology and finite element simulation, a mold stamping twin model is constructed and analyzed, and the industrial Internet is used to perform positive sample retrieval and predetermined attenuation factor testing. Finally, through weight fusion analysis, the accuracy and scientificity of mold life prediction under different working conditions are improved, thereby improving the accuracy and rationality of mold replacement, reducing the waste of mold resources, and avoiding downtime losses caused by sudden mold failures, thereby achieving the technical effect of ensuring the safe and stable operation of stamping equipment. BRIEF DESCRIPTION OF THE DRAWINGS
[0009] In order to more clearly illustrate the technical solutions in 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 creative work.
[0010] Figure 1 A schematic diagram of a mold life prediction method based on simulation provided in an embodiment of the present application;
[0011] Figure 2 A schematic diagram of the process of performing local segmentation in the mold life prediction method based on simulation provided in an embodiment of the present application. DETAILED DESCRIPTION
[0012] The present application provides a mold life prediction method based on simulation, which is used to solve the technical problems in the existing mold life prediction methods, such as the poor accuracy, practicality and reliability of mold life prediction under different working conditions due to the relatively single prediction method and incomplete consideration of factors, resulting in poor accuracy of mold replacement. By combining the key influencing factors of mold fatigue damage, digital twin technology and finite element simulation, a mold stamping twin model is constructed and analyzed, and the industrial Internet is used to perform positive sample retrieval and predetermined attenuation factor testing. Finally, through weight fusion analysis, the accuracy and scientificity of mold life prediction under different working conditions are improved, thereby improving the accuracy and rationality of mold replacement, reducing the waste of mold resources, and avoiding downtime losses caused by sudden mold failures, thereby achieving the technical effect of ensuring the safe and stable operation of stamping equipment.
[0013] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.
[0014] It should be noted that any variations of the terms "include" and "have" are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules that are not explicitly listed or inherent to these processes, methods, products or devices.
[0015] like Figure 1 As shown, the present application provides a mold life prediction method based on simulation, the method comprising:
[0016] Step S100: Acquire key influencing factors of mold fatigue damage, perform feature extraction on a target mold according to the key influencing factors, and obtain a mold feature set.
[0017] Furthermore, the method provided in the application embodiment also includes:
[0018] The key influencing factors include mold material characteristics, mold surface characteristics, stamping load characteristics and working environment characteristics.
[0019] In the embodiments of the present application, the key influencing factors of mold fatigue damage include mold material characteristics, mold surface characteristics, stamping load characteristics and working environment characteristics. Mold material characteristics include mechanical properties such as hardness, strength, toughness and wear resistance of the material. Mold surface characteristics include surface roughness and finish. Stamping load characteristics include load size and distribution, stamping frequency and stroke speed during the stamping process. Working environment characteristics include working temperature and temperature change, etc.
[0020] The mold material characteristics are automatically acquired through databases or sensors, including hardness, strength, toughness and wear resistance, and the microstructure of the material is automatically analyzed through equipment such as scanning electron microscopes and X-ray diffractometers. The mold surface characteristics automatically measure the surface roughness and finish through equipment such as surface roughness meters and optical microscopes, and record the surface treatment process parameters. The stamping load characteristics monitor and record the load size, distribution, stamping frequency and stroke speed in real time through force sensors and pressure sensors installed on the mold, and automatically measure stress concentration and cyclic stress characteristics through strain gauges and dynamic stress analyzers. The working environment characteristics automatically monitor the ambient temperature, humidity, corrosive medium concentration and lubrication conditions through temperature sensors, humidity sensors and chemical analysis instruments. According to the above key influencing factors, the target mold is automatically feature extracted to form a mold feature set.
[0021] Step S200: Based on the digital twin, simulation is performed in a three-dimensional simulation platform according to the mold attribute information and the mold feature set to construct a mold stamping twin model.
[0022] In the embodiment of the present application, the attribute information of the mold is first obtained and integrated through the database, including data such as material type, geometric dimensions and manufacturing process. This information, as a static constraint, provides accurate physical parameters for the basic construction of the model. Next, key features are extracted from the mold feature set, which include the material properties, surface features, stamping load features and working environment features of the mold. These dynamic constraints ensure that the model can accurately reflect the changing factors in the actual use environment.
[0023] Subsequently, simulation is performed on a three-dimensional simulation platform based on the above static and dynamic constraints, and the mold attribute information and mold feature set are input into the simulation platform to generate a mold stamping twin model.
[0024] Step S300: using finite element simulation technology to locally segment the die stamping twin model, simulate stamping based on multiple local twin models, and determine the predicted life of the first die, wherein the predicted life of the die is the number of cyclic loads.
[0025] In the embodiment of the present application, the die stamping twin model is first subjected to finite element analysis, and the stress and strain distribution of each part of the die during the stamping process is analyzed through FEA. After completing the preliminary overall simulation, the die stamping twin model is locally segmented to identify and focus on analyzing the key areas of the die that are most susceptible to fatigue damage during the stamping process. These key areas are places where stress is concentrated or where stress changes greatly. Through local segmentation, multiple local twin models are generated.
[0026] Next, simulate stamping based on the local twin model. During the simulated stamping process, cyclic loads under actual working conditions are applied to simulate the stress and strain changes of the mold during multiple stamping operations. The simulation results of each local twin model will reflect the fatigue life of the area, that is, the number of cyclic loads that the area can withstand before reaching fatigue failure.
[0027] In order to calculate the fatigue life of the entire mold, the fatigue life data of each local twin model are summarized and analyzed. Specifically, fatigue life calculation methods, such as Miner's law, are applied to perform equivalent transformation and weighted cumulative superposition of the fatigue life of each local area. Through this process, the predicted life of the first mold is finally determined.
[0028] Further, such as Figure 2 As shown, in the method provided in the embodiment of the application, the die stamping twin model is locally segmented, and further includes:
[0029] Taking the mold attribute information and the mold feature set as constraints, a sample mold stamping model set and a sample mold damage frequency set are obtained, wherein the sample mold damage frequency set includes multiple local mold damage frequencies; the multiple local mold damage frequencies are screened based on a predetermined frequency threshold to determine multiple high-frequency local molds; the sample mold stamping model set is locally segmented using finite element simulation technology to obtain multiple sample mold local model sets; the multiple sample mold local model sets are supplemented according to the multiple high-frequency local molds to obtain multiple sample optimized local model sets; a mold segmentation channel is constructed based on the sample mold stamping model set and the multiple sample optimized local model sets, the mold stamping twin model is locally segmented through the mold segmentation channel, and multiple local twin models are output.
[0030] In the embodiment of the present application, the mold attribute information and the mold feature set are used as constraints to obtain a sample mold stamping model from an existing mold database or through a sensor data acquisition system. The number of damages in each local area of the sample mold during the stamping process is counted. The data is processed using data analysis tools and algorithms, such as statistical analysis software, to generate the damage frequency of each local area, forming a sample mold damage frequency set, including multiple local mold damage frequency data.
[0031] Based on a predetermined frequency threshold, the local frequencies in the sample mold damage frequency concentration are screened. A data screening algorithm, such as a threshold filtering algorithm, is used to screen out local mold damage frequencies above the predetermined frequency threshold, and multiple high-frequency local mold areas are determined.
[0032] Next, the finite element simulation technology is used to locally segment the sample mold stamping model set. The entire mold model is decomposed into multiple small units through finite element analysis software such as ANSYS, and stress and strain analysis is performed on each unit to obtain multiple sample mold local model sets. These model sets reflect the stress concentration and vulnerable areas of the mold during the stamping process. The local model set generated in the previous step is then supplemented and optimized. Specifically, the data of the high-frequency local mold area is supplemented in detail to the local model to ensure that the model of these areas has higher accuracy and authenticity. The optimization process includes adjusting the mesh density of the model, increasing the number of units in the high-stress area, and refining the key parameters to generate multiple sample optimized local model sets.
[0033] Based on the sample mold stamping model set and multiple sample optimized local model sets, the mold segmentation channel is constructed. Using parallel computing and model integration technology, local models from different sources are integrated to form a mold segmentation channel. Finally, the mold stamping twin model is refined locally segmented through the mold segmentation channel. Combining the results of all previous steps, high-precision simulation technology and segmentation algorithms are used to generate multiple local twin models.
[0034] Furthermore, in the method provided in the embodiment of the application, constructing the mold segmentation channel also includes:
[0035] A sample training set is constructed based on the sample mold stamping model set and the multiple sample optimized local model sets, and the sample training set is equally divided into M parts to obtain M training sets; supervised training and cross-validation are performed on the BP network through the M training sets to obtain M mold segmenters that meet the convergence conditions, and the mold segmentation channel is built in parallel based on the M mold segmenters, wherein the output of the mold segmentation channel is the mode of the output results of the M mold segmenters.
[0036] In an embodiment of the present application, a comprehensive sample training set is first constructed based on a sample mold stamping model set and multiple sample optimized local model sets. The sample training set contains various characteristic data of the mold and the corresponding damage conditions. Through data fusion technology, these data from different sources are integrated together to ensure the consistency and integrity of the data. Then, the sample training set is divided into M parts for cross-validation. Specifically, by using a data segmentation algorithm, such as the data partitioning method in K-fold cross-validation, the training set is divided into M parts on average to obtain M training sets.
[0037] Next, supervised training and cross-validation of the BP network are performed using M training sets. During the supervised learning process, one of the training sets is selected as the validation set each time, and the remaining M-1 training sets are used as the training set. This is repeated M times to ensure that each training set is used as a validation set once. In this way, M mold segmenters that meet the convergence conditions are obtained.
[0038] In order to improve the processing efficiency, a mold segmentation channel is built in parallel based on M mold segmenters. By using the parallel computing framework and model integration technology, the M mold segmenters are integrated into one segmentation channel. This mold segmentation channel can use M mold segmenters for prediction and analysis at the same time, thereby greatly improving the calculation speed and efficiency. In the mold segmentation channel, the final segmentation result is calculated based on the output results of the M mold segmenters. The specific method is to take the mode of the output results of the M segmenters, calculate the mode by statistical methods, and ensure that the final result is representative and consistent. This method can effectively eliminate the errors of individual segmenters and ensure the stability and reliability of the segmentation results.
[0039] In summary, the mold segmentation channel is constructed by constructing a sample training set, supervising the training of the BP network, building the mold segmentation channel in parallel, and calculating the mode of the output results.
[0040] Furthermore, in the method provided in the embodiment of the application, determining the predicted life of the first mold further includes:
[0041] The multiple local twin models are used to simulate stamping and output multiple local fatigue lives. The multiple local fatigue lives are equivalently transformed based on Miner's law to obtain multiple equivalent fatigue lives. The multiple equivalent fatigue lives are weightedly accumulated and superimposed according to a predetermined scheme to output the predicted life of the first mold.
[0042] In an embodiment of the present application, finite element analysis software is first used to perform a detailed simulated stamping analysis on each local twin model. During the simulated stamping process, the stress and strain cycles of each local area are calculated, and finally the fatigue life data of each local area is output. These data reflect the fatigue damage of the mold in different local areas.
[0043] Next, multiple local fatigue lives are equivalently transformed based on Miner's law. The basic principle is to accumulate the damage caused by each load cycle until the total damage reaches the fatigue limit of the material. Through this method, the fatigue life of each local area is converted into an equivalent fatigue life to obtain multiple equivalent fatigue lives. Then, multiple equivalent fatigue lives are weighted cumulatively superimposed according to a predetermined scheme. The purpose of weighted cumulative superposition is to comprehensively consider the fatigue life of each local area and obtain the predicted life of the entire mold. Specifically, the weight is determined according to factors such as the importance, stress level, and damage frequency of each local area. Importance refers to the degree of influence of a local area on the entire mold life; the stress level reflects the stress magnitude of the area during the stamping process; and the damage frequency is the number of times the area is damaged during the stamping process. According to the above principles, a weight factor is assigned to each local area. In the specific operation, normalization is performed according to the stress level and damage frequency so that the sum of all weight factors is equal to 1. For example, if a local area is subjected to a large stress and a high damage frequency, a higher weight is assigned; otherwise, a lower weight is assigned. The equivalent fatigue life of each local area is multiplied by its corresponding weight factor and then cumulatively superimposed. In this way, the fatigue life of each local area and its relative importance are comprehensively considered to obtain a weighted overall mold fatigue life. Finally, the results of weighted cumulative superposition are summarized and the first mold predicted life is output. The first mold predicted life is expressed in the number of cyclic loads, providing the fatigue damage of the mold under actual working conditions.
[0044] Step S400: Using the mold attribute information as a static constraint and the mold feature set as a dynamic constraint, perform positive sample retrieval based on the industrial Internet to obtain an initial mold predicted life.
[0045] Furthermore, in the method provided in the embodiment of the application, positive sample retrieval is performed based on the industrial Internet to obtain the initial predicted mold life, and the method also includes:
[0046] Taking the mold attribute information as a static constraint and the mold feature set as a dynamic constraint, sample information retrieval is performed based on the industrial Internet to obtain an initial sample data set, wherein the initial sample data includes sample mold attribute information, a sample mold feature set and a sample mold service life; taking the mold attribute information and the mold feature set as a benchmark, a comprehensive similarity calculation is performed on the sample mold attribute information and the sample mold feature set to determine the similarity coefficient of the initial sample data; the sample mold service life corresponding to the similarity coefficient that meets a predetermined similarity threshold is selected to obtain a sample mold life set, and the sample mold life set is mean calculated to determine the mean of the sample mold life; the sample mold life set is deviation calculated based on the mean of the sample mold life, the deviation calculation results are screened according to a predetermined deviation threshold to obtain a qualified mold life set, and the initial mold predicted life is obtained after the mean calculation.
[0047] In an embodiment of the present application, the attribute information and feature set of the mold are first determined. Mold attribute information refers to the fixed parameters of the mold, such as material type, size, and manufacturing process, etc., and this information is used as static constraints. The mold feature set includes dynamically changing parameters, such as surface features, usage environment, and stamping load characteristics, etc., which are used as dynamic constraints. By connecting to the industrial Internet, a large amount of relevant production and usage data can be accessed and retrieved. Perform sample information retrieval, find historical sample data similar to the current mold, and obtain an initial sample data set. The initial sample data set includes the attribute information, feature set, and service life of the sample mold.
[0048] Next, based on the mold attribute information and feature set, the attribute information and feature set of the sample mold are comprehensively calculated for similarity, and the similarity coefficient of each initial sample data is calculated using a similarity calculation algorithm, such as cosine similarity or Euclidean distance. The similarity coefficient represents the degree of match between the sample data and the current mold. Based on the calculated similarity coefficient, sample data that meets a predetermined similarity threshold is selected. The predetermined similarity threshold is a standard for screening similar samples. The selected sample data includes its service life, and these data form a sample mold life set.
[0049] Next, the sample mold life set is averaged to determine the sample mold life mean. The sample mold life mean represents the average service life of the mold under similar conditions. Then, based on the sample mold life mean, the sample mold life set is deviation calculated. Deviation calculation is to evaluate the degree of difference between each sample life and the mean. According to the predetermined deviation threshold, the deviation calculation results are screened, and samples with excessive deviation are excluded to obtain a qualified mold life set. Finally, the qualified mold life set is averaged again to obtain the initial mold predicted life.
[0050] Step S500: performing an attenuation test on the initial predicted mold life based on a predetermined attenuation factor and a predetermined risk threshold to obtain a second predicted mold life.
[0051] Furthermore, in the method provided in the embodiment of the application, obtaining the predicted life of the second mold also includes:
[0052] The initial die predicted life is decayed once based on a predetermined decay factor to obtain a decayed die life, and a die life interval is constructed according to the initial die predicted life and the decayed die life; a positive sample search is performed with the die attribute information as a static constraint, the die feature set as a dynamic constraint, and the die life interval as a time zone constraint to obtain a sample stamping workpiece set, wherein the sample stamping workpiece is a qualified workpiece or a defective workpiece; based on the sample stamping workpiece set, the proportion of defective workpieces is counted to determine the first yield rate, the first yield rate is input into a predetermined comparison table, and a first stamping risk level is output; if the first stamping risk level is less than the predetermined risk threshold, the initial die predicted life is set as the second die predicted life; if the first stamping risk level is greater than or equal to the predetermined risk threshold, a second decay is performed on the decayed die life based on a predetermined decay factor, and an iterative analysis is performed based on the secondary die life interval until a die predicted life less than the predetermined risk threshold is obtained, which is set as the second die predicted life.
[0053] In the embodiment of the present application, the initial predicted mold life is first decayed by applying a predetermined decay factor. Assuming that the initial predicted life is 10,000 stamping cycles and the step size of the decay factor is 0.1%, the mold life after one decay is 10,000 times minus 0.1%, that is, 9,990 stamping cycles. A mold life interval is constructed based on the initial predicted mold life and the once decayed mold life. For example, if the initial life is 10,000 times and it is 9,990 times after decay, the life interval is 9,990 to 10,000 times.
[0054] Next, the Industrial Internet is used to retrieve positive samples. Specifically, the mold attribute information is used as a static constraint, the mold feature set is used as a dynamic constraint, and the mold life interval is used as a time zone constraint. Through these constraints, historical data is retrieved in the Industrial Internet database to obtain a set of sample stamping workpieces. These sample stamping workpieces can be qualified workpieces or defective workpieces. Then, the proportion of defective workpieces is counted based on the sample stamping workpiece set to determine the first yield rate. The first yield rate reflects the quality of the mold within this life interval. The first yield rate is input into a predetermined comparison table, and the first stamping risk level is output. If the first stamping risk level is less than the predetermined risk threshold, indicating that the risk of the mold within this life interval is acceptable, the initial mold predicted life is set to the second mold predicted life.
[0055] If the first stamping risk level is greater than or equal to the predetermined risk threshold, a secondary decay will be performed on the once-decayed die life based on a predetermined decay factor. Each decay is in steps of 0.1%, and the predicted life value is gradually reduced. For example, the life after the first decay is 9990 times, and after the second decay is 9980 times. According to the new life interval, such as 9980 to 9990 times, positive sample retrieval and defective rate statistics are performed again. Repeat the above process, perform multiple iterative analyses, and gradually reduce the life interval until a predicted die life less than the predetermined risk threshold is obtained. When the attenuated life prediction value meets the predetermined risk threshold, the value is determined as the second die predicted life.
[0056] Step S600: performing a fusion analysis on the first mold predicted life and the second mold predicted life according to a predetermined strategy, and outputting the mold predicted life of the target mold.
[0057] In an embodiment of the present application, a fusion analysis is performed according to a predetermined strategy, which determines how to reasonably combine the two predicted life data, mainly including credible weights, data consistency, and historical experience. Credible weights refer to the weight values assigned to the first and second predicted lifespans to reflect their respective reliability. For example, the predicted lifespan of the second mold based on actual data and iterative analysis may have a higher weight. Data consistency is used to evaluate the difference between the two predicted lifespans to ensure that the fused results are logically consistent. Historical experience and statistical data help correct the prediction results to make them closer to the actual situation.
[0058] In the weight assignment stage, the weights of the first and second predicted lifespans are assigned according to a predetermined strategy. For example, assuming that the weight of the predicted lifespan of the first mold is 0.4 and the weight of the predicted lifespan of the second mold is 0.6, the sum of these two weights is 1, reflecting their relative importance in the final prediction. Subsequently, the weighted average method is used for fusion analysis, which integrates various information by assigning weights to different data. The calculation process is that if the predicted lifespan of the first mold is 9000 times and the predicted lifespan of the second mold is 9500 times, and the weights are 0.4 and 0.6 respectively, then the predicted lifespan of the target mold after weighted average is 9000 times multiplied by 0.4 plus 9500 times multiplied by 0.6, that is, 9300 times. To ensure the reliability of the results, the results of the weighted average are checked for data consistency. Deviation analysis is used to calculate the difference between the fusion result and the first and second predicted lifespans to ensure that the deviation is within a reasonable range. By calculating the difference between the fusion result and each input predicted lifespan value, potential abnormal data can be discovered and corrected. Consistency detection compares the fusion result with the actual usage data to ensure that the prediction result is consistent with the actual situation. Verification is performed using historical data and empirical rules to confirm the accuracy and consistency of the fusion results.
[0059] Finally, through the above fusion analysis and data verification, the predicted life of the target mold is determined and output.
[0060] Furthermore, in the method provided in the embodiment of the application, configuring a predetermined strategy further includes:
[0061] Collect sample data for prediction accuracy analysis to determine a first prediction accuracy and a second prediction accuracy, wherein the first prediction accuracy is a simulation prediction accuracy and the second prediction accuracy is a positive sample retrieval prediction accuracy; based on the coefficient of variation method, configure credible weights according to the first prediction accuracy and the second prediction accuracy to generate the predetermined strategy.
[0062] In an embodiment of the present application, sample data is first collected from actual production and use processes. These data include the service life and actual usage of historical molds. The collection of sample data is achieved through a database management system and a data acquisition system to ensure the accuracy and completeness of the data. Next, the first prediction accuracy and the second prediction accuracy are calculated respectively. The first prediction accuracy, that is, the simulation prediction accuracy, is the predicted life obtained by finite element analysis and simulation. The mold is simulated and tested using simulation software, and the simulation results are compared with the actual use data to calculate its prediction accuracy. The specific steps include calculating the difference between the predicted life and the actual service life of each sample, and using these difference values to calculate the mean absolute error or root mean square error to quantify the prediction accuracy.
[0063] The second prediction accuracy, namely the positive sample retrieval prediction accuracy, is the predicted life obtained through positive sample retrieval based on the Industrial Internet. By retrieving historical data in the Industrial Internet, samples similar to the current mold conditions are found, and the actual service life of these samples is compared with the predicted life to calculate their prediction accuracy. The specific operation is similar to the calculation of simulation prediction accuracy to ensure the consistency of the method.
[0064] After determining the two prediction accuracies, the credible weights are configured based on the coefficient of variation method. The coefficient of variation is an indicator to measure the degree of data variation, which is defined as the ratio of the standard deviation to the mean. The smaller the coefficient of variation, the more stable the data and the higher the prediction accuracy. Calculate the coefficient of variation of the first prediction accuracy and the second prediction accuracy, and assign weights according to the coefficient of variation. Specifically, first calculate the coefficient of variation of the two prediction accuracies, that is, the coefficient of variation of the first prediction accuracy and the coefficient of variation of the second prediction accuracy. Then, assign weights according to the coefficient of variation, and the smaller the coefficient of variation, the higher the weight of the prediction accuracy. The weight of the first prediction accuracy is the reciprocal of the coefficient of variation of the second prediction accuracy divided by the sum of the reciprocal of the coefficient of variation of the first prediction accuracy and the reciprocal of the coefficient of variation of the second prediction accuracy; the weight of the second prediction accuracy is the reciprocal of the coefficient of variation of the first prediction accuracy divided by the sum of the reciprocal of the coefficient of variation of the first prediction accuracy and the reciprocal of the coefficient of variation of the second prediction accuracy.
[0065] Through the above steps, a predetermined strategy for fusion analysis is generated, and the strategy includes credible weights assigned to the first and second predicted life spans.
[0066] In the embodiments of the present application, in summary, the embodiments of the present application have at least the following technical effects:
[0067] This application obtains key influencing factors of mold fatigue damage, extracts features of the target mold based on the key influencing factors, and obtains a mold feature set; based on digital twins, simulation is performed in a three-dimensional simulation platform based on mold attribute information and mold feature sets to construct a mold stamping twin model; finite element simulation technology is used to locally segment the mold stamping twin model, and simulate stamping based on multiple local twin models to determine the predicted life of the first mold, wherein the predicted mold life is the number of cyclic loads; using mold attribute information as a static constraint and the mold feature set as a dynamic constraint, positive sample retrieval is performed based on the industrial Internet to obtain an initial predicted mold life; an attenuation test is performed on the initial predicted mold life based on a predetermined attenuation factor and a predetermined risk threshold to obtain a second predicted mold life; a fusion analysis is performed on the first predicted mold life and the second predicted mold life according to a predetermined strategy, and the predicted mold life of the target mold is output. The present invention solves the technical problem that the existing mold life prediction method has a relatively single prediction method and does not take into account all factors, resulting in poor accuracy, practicality and reliability of mold life prediction under different working conditions, and poor accuracy of mold replacement. By combining the key influencing factors of mold fatigue damage, digital twin technology and finite element simulation, a mold stamping twin model is constructed and analyzed, and the industrial Internet is used to perform positive sample retrieval and predetermined attenuation factor testing. Finally, through weight fusion analysis, the accuracy and scientificity of mold life prediction under different working conditions are improved, thereby improving the accuracy and rationality of mold replacement, reducing the waste of mold resources, and avoiding downtime losses caused by sudden mold failures, thereby achieving the technical effect of ensuring the safe and stable operation of stamping equipment.
[0068] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. The processes depicted in the accompanying drawings do not necessarily require the specific order and continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0069] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.
[0070] This specification and drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.
Claims
1. Mold life prediction method based on simulation, It is characterized in that include: Acquire key influencing factors of mold fatigue damage, extract features of the target mold according to the key influencing factors, and obtain a mold feature set; Based on the digital twin, simulation is performed in a three-dimensional simulation platform according to the mold attribute information and the mold feature set to construct a mold stamping twin model; Using finite element simulation technology, the die stamping twin model is locally segmented, and stamping simulation is performed based on multiple local twin models to determine the predicted life of the first die, wherein the predicted life of the die is the number of cyclic loads; Taking the mold attribute information as a static constraint and the mold feature set as a dynamic constraint, performing positive sample retrieval based on the industrial Internet to obtain an initial mold predicted life; Performing an attenuation test on the initial predicted mold life based on a predetermined attenuation factor and a predetermined risk threshold to obtain a second predicted mold life; Performing a fusion analysis on the first mold predicted life and the second mold predicted life according to a predetermined strategy, and outputting a mold predicted life of a target mold; Among them, the key influencing factors include mold material characteristics, mold surface characteristics, stamping load characteristics and working environment characteristics; Configure the reservation policy, including: Collect sample data to perform prediction accuracy analysis to determine a first prediction accuracy and a second prediction accuracy, wherein the first prediction accuracy is a simulation prediction accuracy and the second prediction accuracy is a positive sample retrieval prediction accuracy; Based on the coefficient of variation method, the credible weights are configured according to the first prediction accuracy and the second prediction accuracy to generate the predetermined strategy.
2. According to the mold life prediction method based on simulation according to claim 1, It is characterized in that The die stamping twin model is locally segmented, including: Taking the mold attribute information and the mold feature set as constraints, obtaining a sample mold stamping model set and a sample mold damage frequency set, wherein the sample mold damage frequency set includes a plurality of local mold damage frequencies; Screening the plurality of local mold damage frequencies based on a predetermined frequency threshold to determine a plurality of high-frequency local molds; Using finite element simulation technology, the sample mold stamping model set is locally segmented to obtain multiple sample mold local model sets; Supplementing the plurality of sample mold local model sets according to the plurality of high-frequency local molds to obtain a plurality of sample optimized local model sets; A mold segmentation channel is constructed based on the sample mold stamping model set and the multiple sample optimized local model sets, and the mold stamping twin model is locally segmented through the mold segmentation channel to output multiple local twin models.
3. According to the mold life prediction method based on simulation according to claim 2, It is characterized in that Construct mold segmentation channels, including: Constructing a sample training set based on the sample mold stamping model set and the multiple sample optimized local model sets, dividing the sample training set into M equal parts, and obtaining M training sets; The BP network is supervisedly trained and cross-validated through the M training sets to obtain M mold segmenters that meet the convergence conditions, and the mold segmentation channel is built in parallel based on the M mold segmenters, wherein the output of the mold segmentation channel is the mode of the output results of the M mold segmenters.
4. According to the mold life prediction method based on simulation according to claim 1, It is characterized in that Determine the first die life prediction, including: Simulating stamping through the multiple local twin models to output multiple local fatigue lives; Based on Miner's rule, the multiple local fatigue lives are equivalently transformed to obtain multiple equivalent fatigue lives, and the multiple equivalent fatigue lives are weightedly accumulated and superimposed according to a predetermined scheme to output the predicted life of the first mold.
5. According to the mold life prediction method based on simulation according to claim 1, It is characterized in that Perform positive sample retrieval based on the Industrial Internet to obtain the initial mold life prediction, including: Taking the mold attribute information as a static constraint and the mold feature set as a dynamic constraint, sample information retrieval is performed based on the industrial Internet to obtain an initial sample data set, wherein the initial sample data includes sample mold attribute information, sample mold feature set and sample mold service life; Based on the mold attribute information and the mold feature set, a comprehensive similarity calculation is performed on the sample mold attribute information and the sample mold feature set to determine a similarity coefficient of the initial sample data; Selecting sample mold service lives corresponding to similarity coefficients that meet a predetermined similarity threshold to obtain a sample mold service life set, performing mean calculation on the sample mold service life set, and determining a mean of the sample mold service life; The sample mold life set is subjected to deviation calculation based on the sample mold life mean, and the deviation calculation result is screened according to a predetermined deviation threshold to obtain a qualified mold life set, and the initial mold predicted life is obtained after mean calculation.
6. According to the mold life prediction method based on simulation according to claim 5, It is characterized in that Get the predicted life of the second mold, including: Performing a decay on the initial predicted mold life based on a predetermined decay factor to obtain a decayed mold life, and constructing a mold life interval according to the initial predicted mold life and the decayed mold life; Taking the mold attribute information as a static constraint, the mold feature set as a dynamic constraint, and the one-time mold life interval as a time zone constraint, a positive sample search is performed to obtain a sample stamping workpiece set, wherein the sample stamping workpiece is a qualified workpiece or a defective workpiece; Based on the sample stamping workpiece set, the proportion of defective workpieces is counted, the first yield rate is determined, the first yield rate is input into a predetermined comparison table, and a first stamping risk level is output; If the first stamping risk level is less than the predetermined risk threshold, setting the initial die predicted life to the second die predicted life; If the first stamping risk level is greater than or equal to the predetermined risk threshold, a secondary decay is performed on the first decay die life based on a predetermined attenuation factor, and an iterative analysis is performed based on the secondary die life interval until a predicted die life less than the predetermined risk threshold is obtained, which is set as the second predicted die life.
