A twin model-based sensitivity analysis and particle swarm algorithm machine tool feeding system service performance maintenance method

Abstract

This invention discloses a method for maintaining the service performance of machine tool feed systems based on sensitivity analysis using a twin model and a particle swarm optimization algorithm. The overall method encompasses a five-layer framework for digital twin modeling, electromechanical wear mechanism fusion and encapsulation, time-varying element sensitivity analysis, and multi-objective maintenance decision optimization. The modeling part employs an object-oriented five-layer architecture, deeply encapsulating physical mechanisms such as electromechanical coupling wear evolution based on spatial location distribution into a quasi-physical model layer, achieving accurate description and real-time updates of implicit time-varying elements such as localized guideway wear. The decision-making part uses the Sobol sensitivity analysis method to quantify the impact weight of each element on service performance, constructs a full-effect exponential-driven adaptive tightening mechanism for early warning thresholds, and combines it with a particle swarm optimization algorithm (PI-PSO) that introduces degradation inertia and physical boundary penalties to iteratively calculate the optimal maintenance scheme with the goal of minimizing maintenance costs, downtime losses, and performance degradation. This method not only solves the problem that general models are difficult to describe the underlying electromechanical-dynamic coupling mechanism but also achieves deep integration of optimization algorithms and physical degradation laws, significantly improving the service reliability and economic benefits of machine tool feed systems.

Description

Technical Field

[0001] This invention belongs to the field of digital twin technology, specifically relating to a method for maintaining the service performance of machine tool guideway systems based on sensitivity analysis of twin models and particle swarm optimization algorithm. Background Technology

[0002] As the manufacturing industry moves towards higher precision, the reliability of machine tool feed systems, including components such as guideways and lead screws, directly determines the machining quality. However, during long-term operation, the feed system is affected by time-varying factors such as alternating loads and wear, leading to a deterioration in accuracy.

[0003] Current maintenance strategies are mainly divided into two categories: one is preventive maintenance based on fixed cycles, which is prone to over-maintenance or under-maintenance; the other is heuristic optimization algorithms such as conventional standard particle swarm optimization, which mostly seek multi-objective optimization from a purely mathematical level, rarely taking into account the unidirectional irreversibility of mechanical degradation, and are prone to getting stuck in local optima with maintenance time lag.

[0004] Digital twin technology provides a new approach to status monitoring through virtual-real mapping. In its prior patent CN117910237 B, the applicant proposed an object-oriented twin model construction method, which constructs a five-layer modeling framework applicable from MEML to INML, solving the problems of standardization and reusability in digital twin modeling.

[0005] However, the aforementioned prior patents mainly focus on general modeling methodologies. When it comes to precision machine tool feed systems, the general framework fails to specifically define how to electromechanically map physical parameters such as underlying contact pressure, making it difficult to accurately describe the implicit wear evolution process under the coupling of multiple factors. At the same time, the existing system lacks a method to quantify the sensitivity of numerous time-varying factors, resulting in a lack of dynamic adaptive mechanism for setting warning thresholds. In addition, it does not include multi-objective optimization decision-making logic that integrates physical degradation states.

[0006] Therefore, based on the above five-layer modeling framework, this invention proposes a feed system maintenance method that integrates electromechanical wear mechanism, sensitivity analysis and particle swarm optimization algorithm. Summary of the Invention

[0007] Purpose of the Invention: This invention aims to address the technical problems existing in the maintenance strategies for CNC machine tool feed systems, such as the difficulty in accurately quantifying hidden wear, weak adaptive capability of static early warning thresholds, and insufficient coupling between general optimization algorithms and physical realities. This invention integrates the physical degradation mechanism of the feed system into the underlying iterative logic of a digital twin model and a multi-objective optimization algorithm, providing a machine tool feed system service performance maintenance method based on sensitivity analysis of a twin model and particle swarm optimization.

[0008] To achieve the above objectives, the present invention adopts the following technical solution: a method for maintaining the service performance of a machine tool feed system based on sensitivity analysis of a twin model and particle swarm optimization algorithm, comprising the following steps:

[0009] (1) A five-layer framework modeling method based on object-oriented features is used to perform digital twin modeling of the precision machine tool feed system;

[0010] (2) Perform time-varying element analysis on each element of the state attribute at each level in the digital twin model of the precision machine tool feed system. Based on the division results, perform data acquisition, data mapping, and data preprocessing at the instance model level, confirm the update timing, and realize the real-time update of the digital twin model.

[0011] (3) To model the requirements of production behavior in each level of the digital twin model of the precision machine tool feed system, and to encapsulate and open the mechanism model;

[0012] (4) Based on the output results of the mechanism model and the service performance requirements of the machine tool feed system, the sensitivity analysis method is applied to analyze the impact of each time-varying factor on the service performance of the machine tool feed system and to confirm the warning threshold of each time-varying factor.

[0013] (5) Apply the particle swarm optimization algorithm to minimize maintenance costs, downtime losses and performance degradation penalties. With constraints such as machine tool feed system positioning error - workpiece tolerance and ideal efficiency of machine tool processing workpiece - actual processing efficiency, iteratively calculate "machine tool feed system time-varying element change threshold - maintenance timing" and finally output the optimal maintenance plan.

[0014] Furthermore, in step (1), the five-layer modeling framework based on object-oriented features is used to perform digital twin modeling of the precision machine tool feed system. The specific implementation is as follows:

[0015] (11) The object-oriented feature five-layer framework modeling method adopts the five-layer architecture in the authorized object-oriented twin model construction method (CN 117910237 B). The five-layer modeling framework is as follows: meta-model layer (MEML), abstract model layer (ABML), logical model layer (LOML), quasi-physical model layer (PPML), and instance model layer (INML). Among them, the object-oriented feature supports the encapsulation and inheritance of modules between the five-layer framework, and realizes the polymorphic difference compatibility between different types of guide rails (such as ball / roller) and ball screws through the overloaded behavior interface.

[0016] (12) The object-oriented five-layer framework modeling method described above performs digital twin modeling of the precision machine tool feed system. This includes applying the modeling method to the precision machine tool feed system and defining specific attributes and behaviors at each level: In the metamodel layer (MEML), the general metadata of the feed system model is defined, including unique identifiers, version numbers, inheritance methods, and data format standards; the abstract model layer (ABML) inherits from the metamodel layer, defines the feed system classification and general attributes, including guide rail type, installation direction (horizontal / vertical), rated load parameters, and defines general production behavior interfaces such as "trigger wear alarm" and "dynamic early warning threshold adaptive adjustment command"; the logical model layer (LOML) inherits from the abstract model layer information and defines functional logic. The dynamic decision-making mechanism includes defining positioning accuracy, repeatability error status attributes, and production behavior logic such as dynamic lubrication adjustment; the quasi-physical model layer (PPML) inherits from the logical model layer and serves as the core mechanism layer, defining high-fidelity physical properties and simulation interfaces, including defining material properties such as the elastic modulus and Poisson's ratio of the guide rail steel, and encapsulating and integrating production behaviors such as electromechanical coupling wear simulation modules based on spatial position distribution and wear mechanism life prediction algorithms based on Archard theory; the instance model layer (INML) inherits from the quasi-physical model layer, realizing the instantiation of specific models and feed systems, defining instantiation attributes such as specific feed system geometric information and actual working parameters, and defining production behaviors including real-time data-driven interfaces.

[0017] Furthermore, in step (2), time-varying element analysis is performed on each element of the state attribute at each level in the digital twin model of the precision machine tool feed system. Based on the division results, data acquisition, data mapping, and data preprocessing are performed at the instance model layer, and the update timing is confirmed to realize the real-time update of the digital twin model. The specific implementation is as follows:

[0018] (21) The analysis of time-varying elements of each level of state attributes in the digital twin model of the precision machine tool feed system includes element classification and dynamic modeling: Element classification: According to the dynamic characteristics of the influence of elements on system performance, the state attributes are divided into time-varying elements such as guide rail wear, real-time current fluctuation of servo motor, and vibration amplitude, and non-time-varying elements such as guide rail flatness at the factory and lead screw original lead error; Dynamic modeling: A dynamic evolution model is established for time-varying elements that cannot be directly obtained by sensors. For the key implicit time-varying element of guide rail wear, the electromechanical coupling wear evolution model based on spatial position distribution encapsulated in the quasi-physical model layer (PPML) is called for prediction.

[0019] The electromechanical coupling wear evolution model obtains real-time contact pressure through electromechanical dynamics mapping. and relative sliding speed :

[0020] in, This refers to the real-time current of the servo motor. The torque constant of the motor. For the lead screw, For transmission efficiency, For the quality of the workbench, For the guide rail tilt angle, For the estimated cutting force, Effective contact area;

[0021] The relative sliding velocity is obtained by combining the position differential feedback from the grating ruler. Construct a spatial wear matrix The total wear is discretized to specific coordinate segments of the guide rail:

[0022] in The wear coefficient of the material. The Dirac function represents the function used only at the current motion coordinate position. This leads to localized wear and tear accumulation.

[0023] (22) The data acquisition, data mapping, and data preprocessing performed in the instance model layer according to the partitioning results include sensor deployment confirmation, OPC UA protocol mapping, and signal processing: data acquisition, for time-varying elements that can be directly acquired (such as vibration and motor current), confirm the sensor model and acquisition frequency; for elements that cannot be directly acquired, confirm the acquisition method of their input variables; data mapping, bind the physical side sensor data to the state attribute address of the instance model layer (INML) through the OPC UA protocol; data preprocessing, use the Kalman filter algorithm or similar signal processing algorithm to eliminate sensor noise, and extract the trend feature values ​​of time-varying elements through the sliding window algorithm.

[0024] (23) The confirmation of update timing enables real-time updates of the digital twin model, including confirming the triggering method, update calculation range, and virtual-physical consistency verification of the digital twin model update: the triggering method is set to a combination of threshold triggering and event triggering; the update calculation range is only for the affected local parameters to reduce the load; the consistency is verified by calculating the virtual-physical data difference to ensure that the model reflects the physical entity status in real time.

[0025] The advantage of this step compared to existing technologies is that it overcomes the limitation of existing digital twin models that rely heavily on surface sensor data to infer the underlying state. By introducing an electromechanical coupling wear evolution model based on spatial location distribution, the servo motor signal is mapped to contact pressure and sliding speed, which are difficult to measure directly. This effectively improves the perception accuracy and fidelity of the underlying implicit time-varying elements (such as the actual wear state) of the machine tool feed system.

[0026] Furthermore, in step (3), the requirements for production behavior in each level of the digital twin model of the precision machine tool feed system are modeled using a mechanistic model, and the mechanistic model is encapsulated and its interface is opened. The specific implementation is as follows:

[0027] (31) The requirements for production behavior in each level of the digital twin model of the precision machine tool feed system are modeled by mechanism model, including multi-level behavior modeling; dynamic lubrication decision logic is established in the logic layer (LOML); electromechanical-dynamic coupled friction simulation model is established in the quasi-physical layer (PPML) to solve the real-time contact pressure and wear rate values ​​under different working conditions; and an integrated real-time control interface is established in the instance layer (INML) to issue the final control command.

[0028] (32) The encapsulation and interface opening of the implementation mechanism model include: encapsulating the complex differential equation solution process (such as the Arcard dynamic evolution solution of electromechanical mapping) inside the model, hiding the implementation details to prevent direct external modification; opening standardized input and output interfaces to the outside; setting interface access permissions to allow only specific optimization algorithm modules to call the computing service.

[0029] The advantage of this approach compared to existing technologies lies in overcoming the deficiency of insufficient coupling depth between physical mechanisms and data-driven modules in traditional digital twin models. By deeply encapsulating the complex electromechanical-dynamic friction evolution solution process at the quasi-physical layer and exposing standardized data interfaces, it ensures the stability and independence of the underlying mechanism calculation process while providing efficient and standardized computational support for upper-layer sensitivity analysis and multi-objective optimization algorithms.

[0030] Furthermore, in step (4), based on the output results of the mechanism model and the service performance requirements of the machine tool feed system, the sensitivity analysis method is applied to analyze the impact of each time-varying factor on the service performance of the machine tool feed system, and the warning threshold of each time-varying factor is confirmed. The specific implementation is as follows:

[0031] (41) Based on the output results of the mechanism model and the service performance requirements of the machine tool feed system, the sensitivity analysis method is applied to analyze the impact of each time-varying element on the service performance of the machine tool feed system. This includes using global sensitivity analysis methods such as Sobol sequence sampling to sample all time-varying elements; calculating the main effect index and total effect index of each time-varying element on the service performance of the machine tool (such as positioning error) to quantify the independent influence and interactive influence weight of each element; and selecting key time-varying elements (such as guideway wear and other elements that have a fatal impact on accuracy) based on the weight ranking results.

[0032] (42) The confirmation of the early warning thresholds for each time-varying element includes setting the initial threshold value, dynamically correcting the threshold, and weighing multiple objectives; based on national standards and historical maintenance data, the early warning thresholds for time-varying elements are initially set; the early warning thresholds are dynamically corrected using a threshold dynamic decay function based on the Sobol total effect index, and the specific calculation logic is as follows:

[0033] set up The first number calculated by the Sobol algorithm Each time-varying element in The total effect index at any given time defines the dynamic early warning threshold for this element. for:

[0034] in, This is the initial factory-permitted maximum threshold. For safety margin adjustment factor, Design the lifespan of the feed system; through the aforementioned threshold dynamic decay function, the sensitivity is improved. For higher critical elements, the allowable warning threshold tightens exponentially with operating time, and an economic weighting indicator is introduced into the threshold setting to avoid over-maintenance.

[0035] The advantage of this step compared to existing technologies is that it changes the rigid mode of traditional equipment maintenance that relies excessively on static experience thresholds. This step innovatively establishes a dynamic threshold decay mechanism based on the Sobol full-effect index, enabling the early warning system to adaptively adjust the early warning margin according to the sensitivity of the impact of various physical elements on system accuracy at different life cycle stages, thereby effectively reducing the risk of missed or false alarms caused by a single static threshold.

[0036] Furthermore, in step (5), the particle swarm optimization algorithm is applied to minimize maintenance costs, downtime losses, and performance degradation penalties as optimization objectives. Constraints include the machine tool feed system positioning error minus workpiece tolerance and the machine tool's ideal machining efficiency minus actual machining efficiency. The algorithm iteratively calculates the "machine tool feed system time-varying element change threshold - maintenance timing," ultimately outputting the optimal maintenance plan. The specific implementation is as follows:

[0037] (51) The application of the particle swarm algorithm specifically includes particle swarm algorithm parameters and initialization, fitness function design, iterative optimization process and output optimal maintenance scheme.

[0038] (52) The particle swarm algorithm parameters and initialization include particle encoding and algorithm parameter settings: particle encoding, defining particle vectors. ,in Indicates the first Early warning thresholds for key time-varying elements, Indicates when maintenance is needed; algorithm parameter settings include population size, maximum number of iterations, inertia weight, and learning factor.

[0039] (53) The fitness function design described above constructs a multi-objective optimization fitness function. :

[0040] in, To maintain costs, For downtime losses, This is the performance degradation penalty factor. The weighting coefficient is the performance degradation penalty coefficient. A nonlinear tangent penalty function reflecting the physical tolerance margin is adopted:

[0041] in, The penalty adjustment coefficient, To predict positioning error, This represents the workpiece tolerance limit.

[0042] (54) The iterative optimization process initializes the particle swarm and randomly generates particle positions and velocities. For each particle, a digital twin model is called to predict the performance index of its corresponding maintenance scheme, and the fitness value is calculated by combining the mechanistic model output. In the iterative optimization process, a particle velocity update formula driven by physical degradation is adopted:

[0043] in, The degenerate inertia coefficient, The real-time physical wear rate is calculated by the digital twin model. The direction vector has the same dimension as the particle vector; by incorporating this physical degradation slope as a bias force, the particle swarm is used to find the optimal maintenance time. The iteration direction is forced to be guided in the forward direction, updating the individual optimal (pbest) and the global optimal (gbest) until the maximum number of iterations or the convergence condition is reached.

[0044] (55) The output of the optimal maintenance scheme includes extracting the non-dominated solution set from the final particle swarm, forming the Pareto front, weighing the user's preferences, outputting the optimal scheme containing the best combination of thresholds and maintenance time points, and pushing it to the MES system for decision-making.

[0045] The advantage of this step compared to existing technologies is that it overcomes the limitation of traditional heuristic optimization algorithms, which seek optimization from a purely mathematical level and are prone to detachment from physical reality. This step introduces a nonlinear tangent penalty function reflecting physical tolerance constraints into the iterative optimization process, and incorporates the real-time physical wear rate output by the model as a degradation bias term into the particle velocity update equation. This achieves effective coupling between the algorithm's iterative step size and the physical degradation law, significantly improving the optimization convergence efficiency and the reliability of the decision results.

[0046] Compared with the prior art, the advantages of the present invention are as follows:

[0047] (1) A deep mechanism calculation model of physical entity and twin mapping was established: In the quasi-physical model layer, an electromechanical coupling wear model based on spatial position distribution was constructed. The servo current and grating ruler data were mapped to contact pressure and sliding speed through electromechanical dynamic equations, which made up for the shortcomings of conventional prediction models that are detached from the unique physical structure of machine tools and improved the perception accuracy of the underlying hidden wear.

[0048] (2) A sensitivity-driven adaptive adjustment mechanism for the early warning threshold was constructed: This invention links the Sobol total effect index with the early warning threshold and constructs a dynamic threshold decay function. This enables the system to adaptively adjust the threshold margin based on the physical factors that have the greatest impact on accuracy, effectively reducing the risk of missed or false alarms caused by traditional static thresholds.

[0049] (3) A mechanism-driven particle swarm optimization algorithm that integrates physical degradation states is proposed: This invention constructs a nonlinear tangent penalty function that reflects tolerance constraints to reduce the probability of exceeding tolerances, and introduces real-time physical wear rate as a bias term in the particle velocity update equation. This method couples the algorithm iteration step size with the physical degradation slope, effectively overcoming the limitation of traditional algorithms that are prone to getting trapped in local optima, and improving the rationality and safety of decision-making.

[0050] (4) A technical solution combining physical mechanisms, data dimensionality reduction, and heuristic search has been formed: In this invention, multi-source latent wear is obtained using the electromechanical mapping method; the computational complexity caused by multiple coupling factors is addressed by using Sobol dynamic threshold adjustment for dimensionality reduction and constraint; the aforementioned high-fidelity data further supports the particle swarm optimization algorithm, which integrates physical constraints and degradation inertia, for solving the problem. Each step supports the others, overcoming the limitations of applying a single algorithm in complex coupled systems and improving the scientific nature and effectiveness of the maintenance strategy. Attached Figure Description

[0051] Figure 1 This is a diagram illustrating the overall technical roadmap of the method of the present invention.

[0052] Figure 2 A five-layer digital twin modeling architecture diagram for the feed system.

[0053] Figure 3 Update the logical graph for data mapping and implicit elements (wear and tear).

[0054] Figure 4 The flowchart shows the maintenance decision algorithm based on the total effect index and PI-PSO.

[0055] Figure 5 Example diagram of multi-objective optimization results (Pareto front illustration). Detailed Implementation

[0056] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way.

[0057] Combination Figure 1 The overall technical roadmap shown, taking the X-axis feed system of a certain type of precision milling machine as an example, includes the following steps in its service performance maintenance method:

[0058] Example: Maintenance decision for the X-axis feed system of a precision milling machine

[0059] Step (1): Digital twin modeling based on a five-layer framework

[0060] Combination Figure 2 The diagram shows a five-layer digital twin modeling architecture for the feed system. A five-layer framework modeling method based on object-oriented features is used to digitally twin the X-axis feed system of this precision milling machine. General attributes and production behavior interfaces are defined in the Meta-Model Layer (MEML) and Abstract Model Layer (ABML). The Positioning Accuracy Status Attributes and Dynamic Lubrication Decision Logic of the X-axis are defined in the Logical Model Layer (LOML). High-fidelity physical attributes are encapsulated in the Physicophysical Model Layer (PPML), such as setting the elastic modulus of the guide rail steel to 210 GPa and the Poisson's ratio to 0.3, and integrating an electromechanical coupling wear simulation module based on spatial position distribution. Finally, the specific geometric information of the X-axis (such as a nominal lead screw diameter of 40 mm and a lead of 10 mm) is bound in the Instantiated Model Layer (INML) to complete the instantiation.

[0061] Step (2): Time-varying element analysis and real-time model update

[0062] Combination Figure 3 The data mapping and implicit element update logic diagram shown divides the X-axis system state attributes into non-time-varying elements such as guide rail flatness at the factory, and time-varying elements such as guide rail wear and servo drive current data. For the core implicit time-varying element "guide rail wear," a spatial discrete wear matrix model encapsulated in the PPML layer is used for real-time calculation.

[0063] First, the real-time contact pressure is obtained inversely through electromechanical mapping. Set the real-time current of the X-axis servo motor. The sampled value is 5.2 A, and the motor torque constant is... The lead of the lead screw is 1.2 N·m / A. The transmission efficiency is 10 mm (0.01 m). The value is 0.9, indicating the mass of the workbench. 500 kg, horizontally mounted guide rail (tilt angle) Predict the cutting force based on the machining formula. The value is 1500 N. Substituting this into the electromechanical equations:

[0064] As the effective contact area constant, this parameter can be obtained by consulting the machine tool guideway manufacturer's design manual or calculated in advance from the ball geometry parameters based on Hertz contact theory and input into the twin model static parameter library as a static parameter.

[0065] Subsequently, the relative sliding speed is obtained by combining the position differential feedback from the grating ruler. Construct a spatial wear matrix:

[0066] Setting the wear coefficient of the guide rail material To facilitate efficient processing by computer algorithms, a spatial wear matrix is ​​used. In this algorithm, it is discretized into an array containing several coordinate grids, and the Dirac function is... The physical meaning is: only when the real-time position of the grating ruler is... Only when the device is within a specific coordinate grid is the wear amount of the micro-element generated within the current time step accumulated into the array element corresponding to that grid, thereby achieving accurate recording of the local wear distribution at different positions of the long-stroke guide rail.

[0067] Step (3): Mechanism modeling, encapsulation, and interface exposure

[0068] The solution process for the aforementioned electromechanical coupling differential equations is deeply encapsulated within the PPML layer of the twin model, shielding the underlying solution details and only exposing standardized interfaces to external multi-objective optimization algorithm modules, such as real-time input of servo current. With coordinates Real-time output of physical wear rate .

[0069] Step (4): Sensitivity analysis and early warning threshold confirmation

[0070] Combination Figure 4The first part of the "Global Sensitivity Analysis" process uses the Sobol sequence sampling method to calculate the total effect index of each element. It is assumed that during the machine tool's operation to... In the hourly phase, the total effect index of "guide rail wear" was calculated. This indicates that wear at this point is extremely sensitive to accuracy degradation.

[0071] Dynamically tightening the early warning threshold using the total effect index : Set the initial factory-permissible maximum wear threshold mm, safety margin adjustment factor X-axis design life h. Substitute into the threshold dynamic decay function:

[0072] The dynamic early warning threshold for this element can be calculated at this time. The tolerance has been adaptively tightened from 0.05 mm to approximately 0.043 mm. Based on this, the system achieves early warning driven by physical sensitivity, avoiding missed errors caused by static setpoints.

[0073] Step (5): Physical degradation-driven multi-objective particle swarm optimization decision

[0074] Combination Figure 4 The second half of the "PI-PSO iteration" process applies a mechanism-driven particle swarm optimization algorithm to find the optimal maintenance solution: setting tolerance limits for machined workpieces. mm, penalty adjustment coefficient When the maintenance scheme corresponding to a certain particle predicts the positioning error... When mm (height closes to the tolerance limit), substitute the nonlinear tangent boundary penalty function:

[0075] This results in a very large performance degradation penalty (approximately 6313), which is incorporated into the fitness function. This forces the algorithm to quickly eliminate substandard solutions that could lead to defective products.

[0076] In the particle velocity optimization update, an inertial weight is set. And read the real-time physical wear rate output by the twin interface. mm / h, assuming degradation inertia coefficient Substituting into the degradation bias rate update formula for the fusion physical wear rate:

[0077] The "degradation inertia" imposed by the wear rate bias term forces the particle iteration direction to be guided towards earlier maintenance times. Ultimately, the algorithm output contains the optimal combination of thresholds and maintenance time points (e.g., at...). The Pareto front non-dominated solution set (e.g., for guide rail lubrication and nut preload adjustment) is obtained by performing hourly guide rail lubrication and nut preload adjustment. Figure 5 (as shown), and then sent to the MES system for execution.

[0078] It should be noted that the above embodiments are not intended to limit the scope of protection of the present invention. Equivalent transformations or substitutions made based on the above technical solutions all fall within the scope of protection of the claims of the present invention.

Claims

1. A method for maintaining the service performance of a machine tool feed system based on sensitivity analysis using a twin model and particle swarm optimization algorithm, characterized in that, The method includes the following steps: (1) A five-layer framework modeling method based on object-oriented features is used to perform digital twin modeling of the precision machine tool feed system; (2) Perform time-varying element analysis on each element of the state attribute at each level in the digital twin model of the precision machine tool feed system. Based on the division results, perform data acquisition, data mapping, and data preprocessing at the instance model level, confirm the update timing, and realize the real-time update of the digital twin model. (3) To model the requirements of production behavior in each level of the digital twin model of the precision machine tool feed system, and to encapsulate and open the mechanism model; (4) Based on the output results of the mechanism model and the service performance requirements of the machine tool feed system, the sensitivity analysis method is applied to analyze the impact of each time-varying factor on the service performance of the machine tool feed system and to confirm the warning threshold of each time-varying factor. (5) Apply the particle swarm optimization algorithm to minimize maintenance costs, downtime losses and performance degradation penalties as optimization objectives. Use machine tool feed system positioning error - workpiece tolerance, ideal efficiency of machine tool processing workpiece - actual processing efficiency as constraints to iteratively calculate "machine tool feed system time-varying element change threshold - maintenance timing" and finally output the optimal maintenance plan.

2. The method for maintaining the service performance of a machine tool feed system based on sensitivity analysis using a twin model and particle swarm optimization algorithm according to claim 1, characterized in that, The specific implementation of step (1) is as follows: (11) The object-oriented feature-based five-layer framework modeling method adopts a five-layer architecture, with the five-layer modeling framework consisting of the meta-model layer (MEML), abstract model layer (ABML), logical model layer (LOML), quasi-physical model layer (PPML), and instance model layer (INML). Among them, the object-oriented feature supports the encapsulation and inheritance of modules between the five-layer framework, and realizes the polymorphic difference compatibility between different types of guide rails and ball screws through overloaded behavior interfaces. (12) The object-oriented five-layer framework modeling method is used to perform digital twin modeling of the precision machine tool feed system. This includes applying the modeling method to the precision machine tool feed system and defining specific attributes and behaviors at each level: In the meta-model layer (MEML), the general metadata of the feed system model is defined, including unique identifiers, version numbers, inheritance methods, and data format standards; the abstract model layer (ABML) inherits from the meta-model layer, defines the feed system classification and general attributes, including guide rail type, installation direction (horizontal / vertical), rated load parameters, and defines general production behavior interfaces such as "trigger wear alarm" and "dynamic early warning threshold adaptive adjustment command"; the logical model layer (LOML) inherits from the abstract model layer information, defines functional logic and dynamic decision-making mechanisms, including defining positioning accuracy, repeatability error status attributes, and production behavior logic such as dynamic lubrication adjustment; the quasi-physical model layer (PPML) The inherited logical model layer, serving as the core mechanism layer, defines high-fidelity physical properties and simulation interfaces, including defining material properties such as the elastic modulus and Poisson's ratio of the guide rail steel, as well as encapsulating and integrating electromechanical coupling wear simulation modules based on spatial location distribution and wear mechanism life prediction algorithms based on Arcard theory, among other production behaviors. The instance model layer (INML) inherits from the quasi-physical model layer, realizing the instantiation of specific models and feed systems, defining instantiation attributes such as specific feed system geometric information and actual working parameters, and defining production behaviors including real-time data-driven interfaces.

3. The method for maintaining the service performance of a machine tool feed system based on sensitivity analysis using a twin model and particle swarm optimization algorithm according to claim 1, characterized in that, Step (2) is implemented as follows: (21) Perform time-varying element analysis on each element of the state attribute at each level in the digital twin model of the precision machine tool feed system, including element classification and dynamic modeling: element classification, according to the dynamic characteristics of the influence of the elements on the system performance, divide the state attributes into time-varying elements and non-time-varying elements; dynamic modeling, establish a dynamic evolution model for time-varying elements that cannot be directly obtained by sensors, and for the key implicit time-varying element of guide rail wear, call the electromechanical coupling wear evolution model based on spatial position distribution encapsulated in the quasi-physical model layer (PPML) for prediction; The electromechanical coupling wear evolution model obtains real-time contact pressure through electromechanical dynamics mapping. and relative sliding speed : in, This refers to the real-time current of the servo motor. The torque constant of the motor. For the lead screw, For transmission efficiency, For the quality of the workbench, For the guide rail tilt angle, For the estimated cutting force, To determine the effective contact area; the relative sliding speed is obtained by combining the position differential feedback from the grating ruler. Construct a spatial wear matrix The total wear is discretized to specific coordinate segments of the guide rail: in The wear coefficient of the material. The Dirac function represents the function used only at the current motion coordinate position. Localized wear and tear buildup occurs. (22) Based on the partitioning results, data acquisition, data mapping, and data preprocessing are performed at the instance model layer, including sensor deployment confirmation, OPC UA protocol mapping, and signal processing: Data acquisition: For time-varying elements that can be directly acquired, the sensor model and acquisition frequency are confirmed; for elements that cannot be directly acquired, the acquisition method of their input variables is confirmed; Data mapping: The physical side sensor data is bound to the state attribute address of the instance model layer (INML) through the OPC UA protocol; Data preprocessing: Kalman filtering algorithm or similar signal processing algorithm is used to eliminate sensor noise, and the trend feature value of time-varying elements is extracted through the sliding window algorithm. (23) Confirm the timing of the update and realize the real-time update of the digital twin model, including confirming the triggering method, update calculation scope and virtual-physical consistency verification of the digital twin model update: the triggering method is set to a combination of threshold triggering and event triggering; the update calculation scope is only for the affected local parameters to reduce the load; the consistency is verified by calculating the virtual-physical data difference to ensure that the model reflects the physical entity status in real time.

4. The method for maintaining the service performance of a machine tool feed system based on sensitivity analysis using a twin model and particle swarm optimization algorithm according to claim 1, characterized in that, The specific implementation of step (3) is as follows: To meet the production behavior requirements at each level of the digital twin model of a precision machine tool feed system, mechanistic modeling is performed, including multi-level behavioral modeling. Dynamic lubrication decision logic is established at the logic layer (LOML); an electromechanical-dynamic coupled friction simulation model is established at the physical model layer (PPML) to solve for real-time contact pressure and wear rate values ​​under different operating conditions; and an integrated real-time control interface is established at the instance layer (INML) to issue final control commands. (32) Implement the encapsulation and interface opening of the mechanism model, including: encapsulating the complex differential equation solution process inside the model and hiding the implementation details to prevent direct external modification; opening up standardized input and output interfaces to the outside; setting interface access permissions to allow only specific optimization algorithm modules to call the computing service.

5. The method for maintaining the service performance of a machine tool feed system based on sensitivity analysis using a twin model and particle swarm optimization algorithm according to claim 1, characterized in that, Step (4) is implemented as follows: (41) Based on the output results of the mechanism model and the service performance requirements of the machine tool feed system, the sensitivity analysis method is applied to analyze the impact of each time-varying element on the service performance of the machine tool feed system, including the use of global sensitivity analysis methods such as Sobol sequence sampling method to sample all time-varying elements; Calculate the main effect index and total effect index of each time-varying factor on the machine tool service performance, quantify the independent and interactive influence weights of each factor, and select key time-varying factors based on the weight ranking results. (42) Confirm the early warning thresholds for each time-varying element, including the initial threshold setting, dynamic correction threshold, and multi-objective trade-off; based on national standards and historical maintenance data, initially set the early warning thresholds for time-varying elements; The warning threshold is dynamically adjusted using a threshold dynamic decay function based on the Sobol total effect index. The specific calculation logic is as follows: Let... The first number calculated by the Sobol algorithm Each time-varying element in The total effect index at any given time defines the dynamic early warning threshold for this element. for: in, This is the initial factory-permitted maximum threshold. For safety margin adjustment factor, Design the lifespan of the feed system; through the aforementioned threshold dynamic decay function, the sensitivity is improved. For higher critical elements, the allowable warning threshold tightens exponentially with operating time, and an economic weighting indicator is introduced into the threshold setting to avoid over-maintenance.

6. The method for maintaining the service performance of a machine tool feed system based on sensitivity analysis using a twin model and particle swarm optimization algorithm according to claim 1, characterized in that, Step (5) is implemented as follows: (51) The application of the particle swarm optimization algorithm specifically includes particle swarm optimization algorithm parameters and initialization, fitness function design, iterative optimization process, and output optimal maintenance scheme. (52) Particle swarm optimization algorithm parameters and initialization, including particle encoding and algorithm parameter settings: Particle encoding, defining particle vectors ,in Indicates the first Early warning thresholds for key time-varying elements, Indicates when maintenance is needed; algorithm parameter settings include population size, maximum number of iterations, inertia weight, and learning factor. (53) Fitness function design, constructing a multi-objective optimization fitness function : in, To maintain costs, For downtime losses, This is the performance degradation penalty factor. The weighting coefficient is the performance degradation penalty coefficient. A nonlinear tangent penalty function reflecting the physical tolerance margin is adopted: in, The penalty adjustment coefficient, To predict positioning error, For the workpiece tolerance limit, (54) Iterative optimization process, including initializing the particle swarm and randomly generating particle positions and velocities, calling the digital twin model for each particle to predict the performance index of its corresponding maintenance scheme, calculating the fitness value by combining the output of the mechanistic model, updating the individual optimal (pbest) and the global optimal (gbest) of the population, until the maximum number of iterations or the convergence condition is reached. (54) Iterative optimization process: Initialize the particle swarm and randomly generate particle positions and velocities. For each particle, call the digital twin model to predict the performance index of its corresponding maintenance scheme, and combine the mechanistic model to output the calculated fitness value. In the iterative optimization process, the particle velocity update formula driven by physical degradation is adopted: in, The degenerate inertia coefficient, The real-time physical wear rate is calculated by the digital twin model, where 1 represents the direction vector with the same dimension as the particle vector. By incorporating this physical degradation slope as a bias force, the particle swarm searches for the optimal maintenance time. The iteration direction is forcibly guided in the forward direction, updating the individual optimal (pbest) and the population global optimal (gbest) until the maximum number of iterations or the convergence condition is reached. (55) Output the optimal maintenance scheme, including extracting the non-dominated solution set from the final particle swarm, forming the Pareto front, weighing the user's preferences, outputting the optimal scheme containing the best combination of thresholds and maintenance time points, and pushing it to the MES system for decision-making.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the machine tool feed system service performance maintenance method based on twin model sensitivity analysis and particle swarm algorithm as described in any one of claims 1 to 6.

8. A computer-readable storage medium storing computer instructions thereon, characterized in that, When executed by the processor, the computer instructions implement the machine tool feed system service performance maintenance method based on twin model sensitivity analysis and particle swarm algorithm as described in any one of claims 1-6.

Images