A mold life prediction and maintenance decision-making system based on digital twins

Abstract

This invention discloses a mold life prediction and maintenance decision-making system based on digital twins, including modules for data acquisition, data fusion, digital twin construction and updating, remaining life prediction, maintenance decision-making and optimization, and closed-loop execution and feedback. The system collects real-time operating data and production parameters of the physical mold, processes them to generate a comprehensive health status index, and constructs a dynamic digital twin. Based on this digital twin, it simulates future production plans, predicts the remaining lifespan of the mold, and, combined with production scheduling, resource inventory, and cost models, generates and executes the optimal maintenance decision plan. Through a closed-loop feedback mechanism, the system continuously updates and optimizes the digital twin using post-maintenance data, achieving dynamic optimization of mold health status assessment, lifespan prediction, and maintenance strategies. This effectively improves mold management and reduces maintenance costs and production downtime risks.

Description

Technical Field

[0001] This invention belongs to the field of intelligent manufacturing and predictive maintenance technology, and specifically discloses a mold life prediction and maintenance decision system based on digital twins. Background Technology

[0002] As a key basic process equipment in industrial production, the health of molds directly affects product quality, production efficiency and cost. Traditional mold maintenance mostly adopts the mode of regular preventive maintenance or repair after failure, which carries the risk of over-maintenance leading to waste of resources or under-maintenance leading to unplanned downtime.

[0003] In recent years, with the development of IoT and big data technologies, predictive maintenance has become a new trend in equipment maintenance. However, existing technical solutions often have the following limitations:

[0004] Insufficient data utilization: It relies heavily on single sensor data, such as temperature and pressure, and fails to effectively integrate multi-source heterogeneous data, such as production parameters and offline detection data, resulting in incomplete condition assessment.

[0005] Disconnected from business operations: The lifespan prediction results are not deeply coupled with business systems such as production planning and resource inventory, and cannot generate optimal decisions that can be directly executed, thus forming "data silos".

[0006] Lack of closed-loop optimization: The system fails to feed back the effects of maintenance to the model, and lacks the ability to learn and optimize itself.

[0007] Therefore, it is necessary to invent a mold life prediction and maintenance decision-making system based on digital twins to solve the above problems. Summary of the Invention

[0008] To overcome the aforementioned deficiencies in the prior art, this invention provides a mold life prediction and maintenance decision system based on digital twins, which effectively solves the problems mentioned in the background art.

[0009] To achieve the above objectives, the present invention provides the following technical solution: a mold life prediction and maintenance decision-making system based on digital twins, specifically comprising:

[0010] The data acquisition module is used to collect multi-source heterogeneous data of the physical mold in real time. The multi-source heterogeneous data includes operating condition data, production parameters and result data.

[0011] The data fusion module is used to perform spatiotemporal alignment, noise filtering, and feature extraction on the collected multi-source heterogeneous data, and to generate a comprehensive health status index for the mold.

[0012] The digital twin construction and update module is used to build and maintain a dynamic digital twin corresponding to the physical model. The digital twin includes a geometric model, a physical model, and a behavioral model.

[0013] The remaining life prediction module is used to predict the remaining life of the physical mold based on the current state of the digital twin and future production plans.

[0014] The maintenance decision and optimization module is used to generate the optimal maintenance decision plan based on the prediction results of the remaining useful life, combined with production scheduling, resource inventory and cost model;

[0015] The closed-loop execution and feedback module is used to distribute the optimal maintenance decision plan to the production management system and guide its execution, while feeding back the execution results data to the digital twin construction and update module.

[0016] After executing a maintenance decision, the closed-loop execution and feedback module evaluates the effectiveness of the maintenance action by recalculating the comprehensive health status index HI, and defines a maintenance effectiveness coefficient. In the formula, HI aft The HI value after implementing the maintenance decision, HI bef The HI value before implementing maintenance decisions, HI ide To determine the ideal value that the mold HI is expected to recover to after the maintenance decision is implemented, the η value is fed back to the digital twin construction and update module to evaluate and optimize the maintenance strategy model, and to adjust the generation logic of subsequent maintenance decisions accordingly.

[0017] Preferably, the operating condition data includes: clamping force, injection pressure, and mold temperature; the production parameters include: number of production cycles; and the result data includes: mold dimensional accuracy measurement data, mold surface quality data, and initial operating status data for the next production cycle.

[0018] Preferably, the mold dimensional accuracy measurement data includes: dimensional deviations of key parts of the cavity; the mold surface quality data includes: surface roughness value of the cavity, and the total area of ​​surface scratches, cracks and corrosion; the initial operating status data of the next production cycle includes: the deviation between the peak clamping force and the peak injection pressure of the first subsequent production cycle and the set value.

[0019] Preferably, the feature extraction includes: calculating the sub-coefficient C of the comprehensive health status index based on the data collected by the data acquisition module. sub The mold dimensional accuracy index P is calculated based on the mold dimensional accuracy measurement data. pre The mold surface quality index Q is calculated based on the mold surface quality data. sur The state index S is calculated based on the initial operating status data of the next production cycle. ini .

[0020] Preferably, the formula for calculating the sub-coefficients of the comprehensive health status index is as follows: In the formula, ΔF cla ΔP inj ΔT mol F represents the average deviation between the measured and set values ​​of clamping force, injection pressure, and mold temperature within a production cycle. set P set T set These are the set values ​​for clamping force, injection pressure, and mold temperature, respectively; N cyc N represents the number of production cycles that the current mold has completed. max The maximum number of parts that can be produced in the mold design; w1, w2, w3, and w4 are weighting coefficients, and w1+w2+w3+w4=1;

[0021] Preferably, the formula for calculating the mold dimensional accuracy index is: , where ΔD j Let D be the dimensional deviation of the j-th critical part of the cavity, n be the number of critical parts, and D be the dimensional deviation of the j-th critical part. tol These are the allowable values ​​for dimensional tolerances;

[0022] Preferably, the formula for calculating the mold surface quality index is: Ra cur Ra is the current surface roughness value. ini As the initial roughness value, A def A represents the total area of ​​surface scratches, cracks, or rust. tot The total surface area of ​​the cavity is w5 and w6 are weighting coefficients.

[0023] Preferably, the formula for calculating the state index is: , where ΔF cmax ΔP is the deviation between the peak clamping force and the set value. imax F represents the deviation between the peak injection pressure and the set value. set and P set These are the set values ​​for clamping force and injection pressure, respectively, with w7 and w8 being weighting coefficients.

[0024] Preferably, the formula for calculating the comprehensive health status index is: In the formula, HI is the comprehensive health status index, α, β, γ, and δ are weighting coefficients, and α+β+γ+δ=1.

[0025] Preferably, the behavioral model in the digital twin construction and update module is a hybrid-driven damage accumulation model, which is constructed by integrating a physical model and a data-driven algorithm;

[0026] The physical model is based on finite element simulation and fatigue damage theory to calculate the theoretical damage value for each production cycle;

[0027] The data-driven algorithm is based on machine learning algorithms, learns the deviation between actual collected data and theoretical damage values, and performs self-correction on the behavior model;

[0028] The digital twin construction and update module uses the Comprehensive Health Status Index (HI) as a key verification indicator of the model correction effect after each production cycle of the physical mold, and updates the damage status of the behavioral model in real time.

[0029] Preferably, the prediction process of the remaining lifetime prediction module is as follows:

[0030] Receive N future production plans from the production management system and drive the digital twin to simulate the operation of the mold under the N future production plans;

[0031] Starting with the current comprehensive health status index HI, the failure threshold HI of the mold is defined. fai ;

[0032] Calculate the predicted decrease in the comprehensive health status index ΔHI for each simulation period, and iterate until the predicted HI value reaches HI. fai ;

[0033] Remaining service life (RUL) equals reaching HI. fai The number of simulation cycles experienced.

[0034] The maintenance decision and optimization module is used to generate the optimal maintenance decision plan based on the predicted remaining useful life, combined with production scheduling, resource inventory, and cost models.

[0035] Preferably, the optimal maintenance decision scheme is generated as follows: based on the current value and predicted downward trend of the Comprehensive Health Status Index (HI), the health status is classified into levels; a multi-objective optimization function is established with the objectives of minimizing total maintenance cost, maximizing production efficiency, and maximizing equipment availability; the predicted remaining useful life (RUL) and HI level are used as core constraints, while also considering production scheduling gaps, spare parts inventory, and maintenance team resources; an optimization algorithm is used to solve the problem, generating an optimal maintenance decision scheme that includes the best maintenance timing, maintenance type, and maintenance resources.

[0036] The technical effects and advantages of this invention are as follows:

[0037] 1. By using the data fusion module to perform spatiotemporal alignment and feature extraction on multi-source heterogeneous data such as clamping force, temperature, pressure, dimensional accuracy, and surface quality, the limitations of a single data source are overcome.

[0038] 2. An innovative comprehensive health status index was defined, which integrates information from four dimensions: working conditions, precision, surface quality, and initial condition. This provides a quantitative, comprehensive, and reliable unified evaluation index for mold health status, improving the accuracy of condition assessment.

[0039] 3. The maintenance effectiveness coefficient can quantitatively evaluate the actual effect of each maintenance action and feed the results back to the digital twin for correction and optimization of the model and decision-making logic, enabling the system to have the ability to continuously learn and self-improve, and making predictions and decisions more and more accurate. Attached Figure Description

[0040] The present invention will be further described with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present invention. For those skilled in the art, other drawings can be obtained based on the following drawings without creative effort.

[0041] Figure 1 This is a schematic diagram of the module connection of the present invention.

[0042] Figure 2 This is a schematic diagram of the overall process of the present invention. Detailed Implementation

[0043] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0044] like Figure 1 As shown, the present invention provides a mold life prediction and maintenance decision system based on digital twin, including: a data acquisition module, a data fusion module, a digital twin construction and update module, a remaining life prediction module, a maintenance decision and optimization module, and a closed-loop execution and feedback module;

[0045] The following will combine Figure 2 The present invention will be described in detail as follows:

[0046] The data acquisition module is used to collect multi-source heterogeneous data of the physical mold in real time. The multi-source heterogeneous data includes operating condition data, production parameters and result data.

[0047] Furthermore, in the above technical solution, the working condition data includes: clamping force, injection pressure, and mold temperature; the production parameters include: number of production cycles; and the result data includes: mold dimensional accuracy measurement data, mold surface quality data, and initial operating status data for the next production cycle.

[0048] It should be further explained that one injection molding machine completes one mold opening and closing cycle, which is one production cycle.

[0049] Furthermore, in the above technical solution, the mold dimensional accuracy measurement data includes: dimensional deviations of key parts of the cavity; the mold surface quality data includes: surface roughness value of the cavity, and the total area of ​​surface scratches, cracks and corrosion; the initial operating status data of the next production cycle includes: the deviation between the peak clamping force and the peak injection pressure of the first subsequent production cycle and the set value.

[0050] It should be further explained that the data flow of the entire system originates from the data acquisition module. This module acquires multi-source heterogeneous data in real time, on a production cycle basis, through various sensors and systems deployed on physical molds, injection molding machines, and testing equipment. The specific implementation is as follows:

[0051] Data acquisition: Clamping force is collected by installing strain gauge force sensors on the adjusting nut or tie rod of the injection molding machine, recording the real-time change curve of the clamping force, and extracting the average deviation between the clamping force and the set clamping force; injection pressure is collected by installing piezoelectric or strain gauge pressure sensors at the front end of the injection molding machine barrel or nozzle, recording the pressure curve, and extracting the average deviation between the injection pressure and the set injection pressure; mold temperature is directly collected by drilling blind holes near the cavities of the moving and fixed molds, embedding temperature sensors, recording the temperature change curve, and extracting the deviation between the temperature and the set temperature value.

[0052] Production parameters are acquired directly from the technical controller of the injection molding machine.

[0053] Data collection: The mold was disassembled and mounted on a coordinate measuring machine (CMM) to perform high-precision measurements on key geometric features of the cavity and core, such as diameter, depth, and spacing. The measurement results were compared with the initial CAD model or design tolerances of the mold to calculate dimensional deviations. A machine vision system fixed inside the injection molding machine was used to automatically scan and photograph the cavity surface. Image processing algorithms were used to identify and calculate the total area of ​​surface scratches, cracks, and corrosion. A portable surface roughness meter was used to perform contact measurements at specific points in the cavity, with the maximum value used as the representative value.

[0054] Data collection of initial operating status for the next production cycle: After completing a maintenance, such as cleaning, polishing, or replacing parts, and restarting production, record the peak clamping force and peak injection pressure of the first subsequent cycle, compare these values ​​with the standard process setting values ​​corresponding to the product model, and calculate the deviation.

[0055] The data fusion module is used to perform spatiotemporal alignment, noise filtering, and feature extraction on the collected multi-source heterogeneous data, and to generate a comprehensive health status index for the mold.

[0056] Furthermore, in the above technical solution, the feature extraction includes: extracting and calculating the sub-coefficient C of the comprehensive health status index based on the data collected by the data acquisition module. sub The mold dimensional accuracy index P is calculated based on the mold dimensional accuracy measurement data. pre The mold surface quality index Q is calculated based on the mold surface quality data. sur The state index S is calculated based on the initial operating status data of the next production cycle. ini .

[0057] Furthermore, in the above technical solution, the calculation formula for the sub-coefficients of the comprehensive health status index is as follows: In the formula, ΔF cla ΔP inj ΔT mol F represents the average deviation between the measured and set values ​​of clamping force, injection pressure, and mold temperature within a production cycle. set P set T set These are the set values ​​for clamping force, injection pressure, and mold temperature, respectively; N cyc N represents the number of production cycles that the current mold has completed. max The maximum number of parts that can be produced in the mold design; w1, w2, w3, and w4 are weighting coefficients, and w1+w2+w3+w4=1;

[0058] The formula for calculating the mold dimensional accuracy index is as follows: , where ΔD j Let D be the dimensional deviation of the j-th critical part of the cavity, n be the number of critical parts, and D be the dimensional deviation of the j-th critical part. tol These are the allowable values ​​for dimensional tolerances;

[0059] The formula for calculating the surface quality index of the mold is as follows: Ra cur Ra is the current surface roughness value. ini As the initial roughness value, A def A represents the total area of ​​surface scratches, cracks, or rust. tot The total surface area of ​​the cavity is w5 and w6 are weighting coefficients.

[0060] The formula for calculating the state index is as follows: , where ΔF cmax ΔP is the deviation between the peak clamping force and the set value. imax F represents the deviation between the peak injection pressure and the set value. set and P set These are the set values ​​for clamping force and injection pressure, respectively, with w7 and w8 being weighting coefficients.

[0061] Furthermore, in the above technical solution, the formula for calculating the comprehensive health status index is: In the formula, HI is the comprehensive health status index, α, β, γ, and δ are weighting coefficients, and α+β+γ+δ=1.

[0062] It should be further explained that the initial values ​​of the weight coefficients w1, w2, w3, w4, w5, w6, w7, w8, α, β, γ and δ are given based on the experience of domain experts and statistical analysis of historical data. Reinforcement learning is introduced, and the weights are dynamically adjusted based on maintenance feedback, such as the value of η.

[0063] In a preferred embodiment, the system uses the following initial weight settings: w1=0.3, w2=0.3, w3=0.2, w4=0.2, w5=0.7, w6=0.3, w7=0.6, w8=0.4, α=0.4, β=0.3, γ=0.2, δ=0.1;

[0064] The digital twin construction and update module is used to build and maintain a dynamic digital twin corresponding to the physical model. The digital twin includes a geometric model, a physical model, and a behavioral model.

[0065] It should be further explained that the geometric model is used to accurately represent the static properties of the mold, such as its geometric structure, size, and cavity shape; the physical model is used to describe the physical properties, thermodynamic behavior, and mechanical response of the mold material; and the behavioral model is used to simulate the dynamic behavior of the mold in actual operation, such as performance degradation, damage accumulation, and changes in health status.

[0066] Furthermore, in the above technical solution, the behavior model in the digital twin construction and update module is a hybrid-driven damage accumulation model, which is constructed by integrating a physical model and a data-driven algorithm.

[0067] The physical model is based on finite element simulation, such as Abaqus, and fatigue damage theory, such as Miner's linear cumulative damage rule, to calculate the theoretical damage value for each production cycle.

[0068] The data-driven algorithm is based on machine learning algorithms, learns the deviation between actual collected data and theoretical damage values, and performs self-correction on the behavior model;

[0069] Furthermore, the machine learning algorithm can be a neural network algorithm, which takes multi-source heterogeneous data as input and the deviation between the actual damage value and the theoretical damage value as output, and performs supervised learning.

[0070] The digital twin construction and update module uses the comprehensive health status index HI as a key verification indicator of the model correction effect after each production cycle of the mold is completed, and updates the damage status of the behavioral model in real time.

[0071] The remaining life prediction module is used to predict the remaining life of the physical mold based on the current state of the digital twin and future production plans;

[0072] Furthermore, in the above technical solution, the prediction process of the remaining lifetime prediction module is as follows:

[0073] Receive N future production plans from the production management system and drive the digital twin to simulate the operation of the mold under the N future production plans;

[0074] Starting with the current comprehensive health status index HI, the failure threshold HI of the mold is defined. fai ;

[0075] Calculate the predicted decrease in the comprehensive health status index ΔHI for each simulation period, and iterate until the predicted HI value reaches HI. fai ;

[0076] Remaining service life (RUL) equals reaching HI. fai The number of simulation cycles experienced.

[0077] The maintenance decision and optimization module is used to generate the optimal maintenance decision plan based on the prediction results of the remaining useful life, combined with production scheduling, resource inventory and cost model;

[0078] Furthermore, in the above technical solution, the specific method for generating the optimal maintenance decision scheme is as follows: based on the current value and predicted downward trend of the Comprehensive Health Status Index (HI), the health status is divided into levels; a multi-objective optimization function is established with the objectives of minimizing total maintenance cost, maximizing production efficiency, and maximizing equipment availability; the predicted remaining useful life (RUL) and HI level are used as core constraints, while also considering production scheduling gaps, spare parts inventory, and maintenance team resources; an optimization algorithm is used to solve the problem, generating an optimal maintenance decision scheme that includes the best maintenance timing, maintenance type, and maintenance resources.

[0079] In a preferred embodiment of the present invention, the health status is classified as follows: based on the current HI value and the predicted trend, the health status of the mold is divided into four levels: good, attention, warning, and danger.

[0080] The closed-loop execution and feedback module is used to distribute the optimal maintenance decision plan to the production management system and guide its execution, while feeding back the execution results data to the digital twin construction and update module.

[0081] Furthermore, in the above technical solution, after executing the maintenance decision, the closed-loop execution and feedback module evaluates the effectiveness of the maintenance action by recalculating the comprehensive health status index HI, and defines a maintenance effectiveness coefficient. In the formula, HI aft The HI value after implementing the maintenance decision, HI bef The HI value before implementing maintenance decisions, HI ide To determine the ideal value that the mold HI is expected to recover to after the maintenance decision is implemented, the η value is fed back to the digital twin construction and update module to evaluate and optimize the maintenance strategy model, and to adjust the generation logic of subsequent maintenance decisions accordingly.

[0082] It should be further explained that the ideal value HI that the mold HI is expected to recover to after the maintenance decision is implemented. ide Based on statistical analysis of historical maintenance data, the average health status that similar molds can achieve after similar maintenance operations is calculated.

[0083] It should be further explained that the maintenance effectiveness coefficient ranges from [0,1]. The closer the value is to 1, the better the maintenance effect. If η remains low, it indicates that the current maintenance strategy is insufficient to restore the expected performance. In the future, the decision module should recommend a more thorough maintenance solution, such as upgrading components or expanding the scope of repair.

[0084] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A digital twin based mold life prediction and maintenance decision system, characterized in that, The method comprises the following steps: a data acquisition module for real-time acquisition of multi-source heterogeneous data of the physical mold, the multi-source heterogeneous data including working condition data, production parameters and result data; a data fusion module for temporal and spatial alignment, noise filtering and feature extraction of the acquired multi-source heterogeneous data, and generation of a comprehensive health status index of the mold; a digital twin construction and update module for constructing and maintaining a dynamic digital twin corresponding to the physical mold, the digital twin including a geometric model, a physical model and a behavior model; a residual life prediction module for predicting the residual service life of the physical mold based on the current state of the digital twin and future production plans; a maintenance decision and optimization module for generating an optimal maintenance decision scheme based on the prediction results of the residual service life, in combination with production scheduling, resource inventory and cost models; a closed-loop execution and feedback module for issuing the optimal maintenance decision scheme to the production management system and guiding execution, and feeding back the result data after execution to the digital twin construction and update module; The closed-loop execution and feedback module evaluates the effectiveness of the maintenance action by recalculating the comprehensive health index HI after the execution of the maintenance decision, defining a maintenance effectiveness coefficient , in the formula HI aft is the HI value after the execution of the maintenance decision, HI bef is the HI value before the execution of the maintenance decision, HI ide is the ideal value to which the mold HI is expected to recover after the execution of the maintenance decision; the value of η is fed back to the digital twin construction and update module for evaluating and optimizing the maintenance strategy model, and adjusting the generation logic of subsequent maintenance decisions accordingly; The feature extraction comprises: extracting a sub-coefficient C of a comprehensive health state index based on data collected by the data collection module sub , calculating a mold size precision index P based on mold size precision measurement data pre , calculating a mold surface quality index Q based on mold surface quality data sur , calculating a state index S based on initial operation state data of the next production cycle ini ; The calculation formula of the sub-coefficient of the comprehensive health state index is: In the formula, ΔF cla , ΔP inj , and ΔT mol respectively represent the average deviation of the measured value of the clamp force, the injection pressure, and the mold temperature in a production cycle from the set value, F set , P set , and T set respectively represent the set value of the clamp force, the injection pressure, and the mold temperature; N cyc represents the number of production cycles completed by the current mold, N max represents the maximum number of parts that can be produced by the mold design; w1, w2, w3, and w4 are weight coefficients, and w1+w2+w3+w4=1. The calculation formula of the mold size precision index is: Wherein, ΔD j is the size deviation of the jth key position of the cavity, n is the number of key positions, D tol is the size tolerance allowance value; The formula for calculating the mold surface quality index is: wherein Ra cur is the current surface roughness value, Ra ini is the initial roughness value, A def is the total area of surface scratches, cracks or rust, A tot is the total surface area of the cavity, and w5 and w6 are weight coefficients. The calculation formula of the state index is: wherein ΔF cmax is the deviation of the clamping force peak value from the set value, ΔP imax is the deviation of the injection pressure peak value from the set value, F set and P set are the set values of the clamping force and the injection pressure, respectively, and w7 and w8 are weight coefficients.

2. The digital twin based mold life prediction and maintenance decision system of claim 1, wherein: The working condition data includes clamping force, injection pressure and mold temperature; the production parameters include production cycle times; and the result data includes mold dimensional accuracy measurement data, mold surface quality data and initial operating state data of the next production cycle.

3. The digital twin based mold life prediction and maintenance decision system of claim 2, wherein: The mold dimensional accuracy measurement data includes dimensional deviation of key parts of the cavity; the mold surface quality data includes total area of roughness value, surface scratches, cracks and corrosion of the cavity surface; and the initial operating state data of the next production cycle includes deviation of clamping force peak value and injection pressure peak value from set values of the first production cycle.

4. The digital twin based mold life prediction and maintenance decision system of claim 1, wherein: The calculation formula of the comprehensive health state index is: In the formula, HI is the comprehensive health state index, and a, β, γ, and δ are weight coefficients, and a+β+γ+δ=1.

5. The digital twin based mold life prediction and maintenance decision system of claim 1, wherein: The behavior model in the digital twin construction and update module is a hybrid driven damage accumulation model, which is constructed by fusing a physical model and a data driven algorithm; The physical model calculates the theoretical damage value of each production cycle based on finite element simulation and fatigue damage theory; The data driven algorithm learns the deviation between actual collected data and theoretical damage value based on a machine learning algorithm, and performs self-correction on the behavior model; The digital twin construction and update module updates the damage state of the behavior model in real time after the physical mold completes each production cycle, taking the comprehensive health status index HI as a key verification index of model correction effect.

6. The digital twin based mold life prediction and maintenance decision system of claim 1, wherein: The prediction process of the residual life prediction module is as follows: Receive future N production plans from the production management system to drive the digital twin to simulate the running process of the mold under the future N production plans; With the current comprehensive health status index HI as the starting point, define the failure threshold HI of the mold fai ; The predicted decrease in the index of overall health status ΔHI is calculated for each simulation period, by iterative calculation until the predicted value of HI reaches the HI fai ; The remaining useful life RUL is equal to the number of simulated cycles to reach the HI fai the number of cycles experienced; The maintenance decision and optimization module generates an optimal maintenance decision scheme based on the prediction results of the residual service life, in combination with production scheduling, resource inventory and cost models.

7. The digital twin based mold life prediction and maintenance decision system of claim 1, wherein: The specific generation manner of the optimal maintenance decision scheme is as follows: according to the current value and the predicted downward trend of the comprehensive health index HI, the health state is classified; a multi-objective optimization function is established with the lowest total maintenance cost, the highest production efficiency and the maximum equipment availability as targets; the predicted remaining useful life RUL and the HI grade are taken as core constraint conditions, and the gap of production scheduling, the inventory situation of spare parts and the maintenance team resources are considered; and an optimization algorithm is used for solving, so that the optimal maintenance decision scheme including the best maintenance time, the maintenance type and the maintenance resources is generated.

Images