Turning machining center production scheduling management system

Abstract

The present application relates to intelligent manufacturing and production scheduling technical field, specifically for turning and milling machining center production scheduling management system;Contain data acquisition, particle mapping, potential field construction, evolution calculation and scheduling generation module;System obtains order and resource state, maps the order into virtual particle flow with spin attribute, and converts the resource into gravitational potential well;Its core is based on the principle of free energy minimization, driving particles in multidimensional space-time potential field to carry out gradient evolution;When the total energy of the system converges, the particle steady-state coordinates are locked to generate the schedule;The present application converts discrete scheduling into physical field energy minimization problem, uses gradient guidance instead of blind search, and solves the problem that static algorithm is difficult to balance calculation depth and response speed.

Description

Technical Field

[0001] This invention relates to the field of intelligent manufacturing and production scheduling technology, specifically to a production scheduling management system for milling and turning machining centers. Background Technology

[0002] With the widespread application of mill-turn machining technology, the manufacturing industry is increasingly demanding multi-variety, small-batch production models. This production model leads to extremely high complexity in workshop task scheduling, especially in scenarios involving multiple processes, tool switching, and equipment load balancing.

[0003] Currently, production scheduling mainly relies on manual experience or traditional heuristic algorithms for management. Schedulers typically formulate work plans based on static order data and equipment status using basic rules, and make manual adjustments when anomalies occur. However, traditional methods have significant limitations when dealing with high-dimensional constraints. On the one hand, static algorithms struggle to balance computational depth and response speed, often getting bogged down in blindly searching the solution space, resulting in low scheduling efficiency. On the other hand, when faced with dynamic disturbances such as emergency order insertions or equipment failures, existing methods lack adaptive adjustment capabilities. Frequent tool changes and equipment idleness increase production costs, making it difficult to balance global resource optimization with real-time response requirements.

[0004] Therefore, how to achieve efficient calculation and dynamic adaptive optimization of production scheduling in complex workshop environments has become an urgent problem to be solved in this field. Summary of the Invention

[0005] To solve the above-mentioned technical problems, the present invention provides a production scheduling management system for milling and turning machining centers. Specifically, the technical solution of the present invention includes:

[0006] The data acquisition module is used to acquire the set of production orders to be scheduled and the current resource status vector of the machining center. The set of production orders contains multiple sets of process task data, including: process feature sequence, standard machining time, delivery deadline, and tooling requirement characteristics. The resource status vector includes: equipment load rate, remaining tool life, and current idle time window. The particle mapping module is used to map the set of production orders into a virtual particle flow based on the process feature sequence, the standard machining time, and the tooling requirement characteristics. Each task particle in the virtual particle flow is assigned a spin attribute corresponding to the tooling requirement characteristics and a mass attribute corresponding to the standard machining time. Potential energy field construction. The module is used to construct a multidimensional spatiotemporal potential energy field based on the equipment load rate and the current idle time window, wherein the available resource capacity of the processing center is mapped to a gravitational potential energy well in the multidimensional spatiotemporal potential energy field; the evolution calculation module is used to drive the virtual particle flow to perform gradient evolution motion in the multidimensional spatiotemporal potential energy field based on the principle of minimizing free energy, wherein the evolution motion includes: calculating the interaction force between particles based on the spin property, and calculating the field gravity on the particles based on the gravitational potential energy well, until the total energy of the system converges; the scheduling generation module is used to lock the steady-state coordinates of the virtual particle flow in the multidimensional spatiotemporal potential energy field in response to the convergence of the total energy of the system, and decode the steady-state coordinates into production scheduling instructions.

[0007] Preferably, based on the process feature sequence, the standard machining time, and the tooling requirement features, mapping the production order set into a virtual particle flow includes: invoking the tooling requirement features and the standard machining time; defining process tasks with the same tooling requirement features as particles with the same spin direction; defining the standard machining time as the particle's inertial mass; defining the difference between the delivery deadline and the current time as the particle's thermal driving force; and generating the initial state vector of the virtual particle flow based on the spin direction, the inertial mass, and the thermal driving force.

[0008] Preferably, constructing a multidimensional spatiotemporal potential energy field based on the device load rate and the current idle time window includes: calling the device load rate and the current idle time window; defining the region where the device load rate is higher than a preset load threshold as a high potential energy repulsion zone; defining the time period of the current idle time window as a low potential energy attraction zone; and generating the terrain gradient matrix of the multidimensional spatiotemporal potential energy field based on the distribution of the high potential energy repulsion zone and the low potential energy attraction zone.

[0009] Preferably, calculating the interaction force between particles based on the spin property includes: identifying the spin properties of adjacent particles in the virtual particle stream; if the spin properties of adjacent particles are the same, generating a virtual attraction between particles to promote the spatiotemporal clustering of task particles with the same tooling requirements; if the spin properties of adjacent particles are different, generating a virtual repulsion between particles to characterize the damping cost generated by the tool changing operation.

[0010] Preferably, the system further includes a dynamic order insertion response module, used for: monitoring whether any new emergency order insertion tasks have entered the production order set; in response to detecting a new emergency order insertion task, mapping the emergency order insertion task as a high-energy perturbation particle; injecting the high-energy perturbation particle into the current multidimensional spatiotemporal potential energy field to trigger a redistribution of the local field strength; and based on the redistributed local field strength, performing local relaxation evolution of the virtual particle flow until an energy balance state is reached again.

[0011] Preferably, based on the principle of minimizing free energy, the virtual particle flow is driven to undergo gradient evolution motion in the multidimensional spatiotemporal potential energy field, including: defining a total system energy function, which is composed of a weighted sum of delay penalty potential energy, replacement cost potential energy, and equipment idle potential energy; calculating the resultant force vector acting on each task particle within a virtual time step; and updating the position coordinates of each task particle according to the resultant force vector, so that the value of the total system energy function approaches a minimum.

[0012] Preferably, the scheduling generation module is specifically used to: extract the time dimension component and the equipment dimension component from the steady-state coordinates; map the equipment dimension component to the target machine tool ID; map the time dimension component to the process start time and end time; and generate a production scheduling instruction containing material delivery time points and G-code call sequences based on the target machine tool ID, the process start time, and the end time.

[0013] Preferably, the system further includes a bottleneck visualization module, used to: acquire pressure distribution data of the multidimensional spatiotemporal potential energy field after it has evolved and stabilized; identify regions in the pressure distribution data where the potential energy density is higher than a preset warning value; mark the regions as resource bottleneck nodes; and generate equipment supplementation suggestions or process diversion strategies for the resource bottleneck nodes.

[0014] Compared with the prior art, the present invention has the following beneficial effects:

[0015] 1. This system transforms the complex discrete scheduling problem into an energy minimization problem in a physical field by constructing a multidimensional spatiotemporal potential energy field and mapping production orders to virtual particle flows. Utilizing the gradient guidance effect of the potential energy field, the system can guide task particles to automatically avoid high-load regions and slide into idle time windows, avoiding the blind search in the solution space of traditional heuristic algorithms. This solution method based on a physical model, using the principle of free energy minimization, can complete evolution calculations in milliseconds, effectively solving the problem mentioned in the background technology that static algorithms struggle to balance computational depth and response speed, and significantly improving scheduling efficiency.

[0016] 2. This system innovatively introduces spin properties to characterize the tooling requirements of process tasks and optimizes the scheduling sequence by calculating the interaction forces between particles. The system enables task particles with the same tooling requirements to generate virtual attraction, thereby automatically clustering them in space and time. At the same time, it generates virtual repulsion between tasks with different tooling requirements to reserve time for tool changeover. This mechanism does not require pre-setting complex hard rules and can spontaneously generate grouped technical features during the evolution process, reducing frequent tool change operations and equipment idle time, and solving the problem of increased production costs caused by frequent tool changes in the background technology.

[0017] 3. This system establishes a dynamic task insertion response mechanism, mapping emergency task insertion to the injection of perturbation particles with high thermal driving force and high energy density into the potential energy field. Through the definition of thermal driving force, tasks nearing their deadline can gain the ability to cross the high potential energy repulsion zone and prioritize resource acquisition. At the same time, by utilizing local field strength redistribution and local relaxation evolution strategies, the system can fine-tune only the affected region while accepting new tasks, rather than rearranging the entire plan. This self-healing mechanism, similar to physical elasticity, effectively solves the problem in the background technology that existing methods lack adaptive adjustment capabilities in the face of dynamic perturbations and are difficult to balance global optimization and real-time response.

[0018] 4. This system automatically balances equipment load by mapping equipment load rate to repulsive potential energy and idle time window to gravitational potential energy well, utilizing the peak-shaving and valley-filling effect of the physical field to prevent resource congestion and improve utilization. In addition, by analyzing the pressure distribution data after the evolution and stabilization of the multidimensional spatiotemporal potential energy field, the system can accurately identify areas with excessively high potential energy density, thereby intuitively locating resource bottleneck nodes. This not only optimizes the current production resource allocation but also provides quantitative data support for equipment replenishment or process diversion, overcoming the shortcomings of traditional scheduling that rely on manual experience and are difficult to handle high-dimensional constraints. Attached Figure Description

[0019] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0020] Figure 1 This is a structural diagram of the system of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0022] Example 1:

[0023] Please see Figure 1 The production scheduling management system for milling and turning machining centers includes: a data acquisition module, used to acquire the set of production orders to be scheduled and the current resource status vector of the machining center. The production order set contains multiple sets of process task data, including: process feature sequence, standard machining time, delivery deadline, and tool requirement characteristics; the resource status vector includes: equipment load rate, remaining tool life, and current idle time window; and a particle mapping module, used to map the production order set into a virtual particle flow based on the process feature sequence, standard machining time, and tool requirement characteristics. Each task particle in the virtual particle flow is assigned a spin attribute corresponding to the tool requirement characteristics and a spin attribute corresponding to the standard machining time. The system comprises the following modules: a quality attribute module; a potential energy field construction module, used to construct a multi-dimensional spatiotemporal potential energy field based on equipment load rate and current idle time window, wherein the available resource capacity of the processing center is mapped as a gravitational potential energy well in the multi-dimensional spatiotemporal potential energy field; an evolution calculation module, used to drive the virtual particle flow to perform gradient evolution motion in the multi-dimensional spatiotemporal potential energy field based on the principle of minimizing free energy, the evolution motion including: calculating the interaction force between particles based on spin attribute, and calculating the field gravity on the particles based on gravitational potential energy well, until the total energy of the system converges; and a scheduling generation module, used to lock the steady-state coordinates of the virtual particle flow in the multi-dimensional spatiotemporal potential energy field in response to the convergence of the total energy of the system, and decode the steady-state coordinates into production scheduling instructions.

[0024] This embodiment details the core architecture design of the production scheduling management system for a milling and turning machining center. This design aims to solve the deadlock problem of computational depth and response speed in traditional scheduling algorithms. The system uses a data acquisition module as the sensing front end to acquire real-time full-scale status data of the physical workshop. The set of production orders to be scheduled serves as the input source. Each order is parsed into multiple sets of process task data. The process feature sequence refers to the specific chain of operation steps required to process the part. The standard processing time is derived from historical MES data statistics or CAM software simulation. The delivery deadline is the hard constraint time point for the order. The tool requirement feature refers to the tool ID and cutting parameters required to complete the process. The current resource status vector of the machining center is updated in real-time as an environmental constraint, including equipment load rate, remaining tool life, and current idle time window.

[0025] The particle mapping module acts as the system's encoder, transforming discrete production tasks into a continuous medium capable of physical field calculations. The system does not treat tasks as static data blocks but rather as entities moving in spacetime, assigning spin and mass attributes to each task particle in the virtual particle stream, corresponding to tool requirements and standard machining time, respectively. The potential energy field construction module acts as an environment generator, constructing a multi-dimensional spatiotemporal potential energy field based on equipment load rate and the current idle time window. It maps the available resource capacity of the machining center to gravitational potential energy wells; the stronger the resource capacity, the deeper the gravitational well, and the longer the idle time, the wider the gravitational well. Based on this, the evolution calculation module acts as a solution engine, driving the virtual particle stream to undergo gradient evolution based on the principle of free energy minimization. It calculates the interaction force between particles based on spin attributes and the field gravity experienced by the particles based on the gravitational potential energy wells, until the total system energy converges. The scheduling generation module acts as a decoder, responding to the convergence of the total system energy by locking the steady-state coordinates of the virtual particle stream in the multi-dimensional spatiotemporal potential energy field—that is, the spatiotemporal position where the particles finally stop—and decoding these steady-state coordinates into executable production scheduling instructions.

[0026] This embodiment transforms the discrete scheduling problem into an energy minimization problem in a physical field. The system uses the guiding effect of gradients to quickly find the optimal solution, avoiding the blind search in the solution space of traditional algorithms. This physical model-based architecture not only significantly improves scheduling efficiency, but also achieves adaptive matching of complex workshop constraints through the physical mapping of gravitational potential energy wells and particle properties, ensuring the unity of global resource optimization and millisecond-level dynamic response.

[0027] Example 2:

[0028] Based on the process feature sequence, standard machining time, and tooling requirement characteristics, the production order set is mapped to a virtual particle flow, including: calling up tooling requirement characteristics and standard machining time; defining process tasks with the same tooling requirement characteristics as particles with the same spin direction; defining the standard machining time as the particle's inertial mass; defining the difference between the delivery deadline and the current time as the particle's thermal driving force; and generating the initial state vector of the virtual particle flow based on the spin direction, inertial mass, and thermal driving force.

[0029] This embodiment further elaborates on the specific implementation logic of the particle mapping module, namely, how to transform industrial data into physical parameters; the system retrieves the tool requirement characteristics and standard machining time of the process from the database; the system executes the spin attribute definition step, defining process tasks with the same tool requirement characteristics as particles with the same spin direction; the spin direction originates from the vectorized mapping of tool requirement characteristics, specifically constructed using one-hot encoding, with a physical meaning of multi-dimensional vector. ,in, This represents the total number of available tools in the workshop. In the vector, only the index position corresponding to the tool ID is set to 1, and the rest are set to 0; if the task... and tasks If the same cutting tools are required, then their spin vector dot product... ,otherwise ;

[0030] Prior to this, the system performs a mass attribute definition step, defining the ratio of the standard processing time to the system's reference time slice, such as 1 minute, as the dimensionless inertial mass of the particle. Inertial mass The source is the normalization of standard processing time, and its physical meaning is the ability of a particle to resist changes in its state of motion; it is a dimensionless scalar quantity. According to Newton's second law... ,quality Larger particles, i.e., longer tasks, are subject to the same insertion perturbation force. At that time, its acceleration The smaller the value; based on this, the system executes the thermal driving force definition step, setting the delivery deadline. With current time The difference is defined as the thermal driving force of the particle; to quantify the urgency and avoid mathematical singularities, this embodiment introduces a thermal driving urgency parameter. The calculation formula is:

[0031]

[0032] in, The source is the reciprocal transformation of the difference in deadlines, and its physical meaning is the first... The virtual urgency potential energy coefficient of each task particle, in Kelvin analog quantity; The source is a system preset, and its physical meaning is a virtual Boltzmann constant; The source is a preset numerical stability constant, for example The term "hour" in physics refers to the minimum buffer period for action, used to prevent the denominator from being zero; among which, It is a proportionality constant that adjusts the weight of the influence of time urgency on potential energy, and its dimension is energy-time. In this embodiment, it is preferably set to The larger this value, the stronger the driving force of the deadline on the particles. The determination is based on ensuring that... At that time, the potential energy generated is sufficient to overcome the average background noise of the system;

[0033] To accommodate the continuous gradient calculations in the subsequent dynamic equations, the urgency of the above reciprocal form is preferably converted into soft-constrained potential energy through an exponential mapping when constructing the potential energy field. To avoid numerical singularities and ensure gradient smoothness, a parameter is introduced here. As a time-sensitivity attenuation factor, its dimension is time. Typical value This parameter controls the rate at which the urgency level increases as time approaches. The larger the value, the steeper the thermal driving force obtained by tasks nearing their delivery date, thus physically simulating the business logic of prioritizing urgent tasks.

[0034] This exponential mapping mechanism is used to prevent the system from crashing due to the generation of infinite heat when the current time approaches the deadline infinitely; among other things... The source is real-time calculation, and the physical meaning is the remaining delivery time; based on the above parameters, the initial state vector of the virtual particle flow is generated, including position, momentum and spin value;

[0035] This embodiment introduces the physical definitions of inertial mass and thermal driving force, giving static task data dynamic behavioral characteristics; long-duration tasks have high stability due to their large mass and are not easily interrupted, ensuring the rigidity of the main job plan; while tasks nearing their deadline gain great thermal driving force due to rising temperature, enabling them to cross high potential energy repulsion zones and seize busy resources, thus spontaneously ensuring on-time delivery of orders at the physical level.

[0036] Example 3:

[0037] Based on the device load rate and the current idle time window, a multidimensional spatiotemporal potential energy field is constructed, including: calling the device load rate and the current idle time window; defining the area where the device load rate is higher than the preset load threshold as the high potential energy repulsion zone; defining the time period of the current idle time window as the low potential energy attraction zone; and generating the terrain gradient matrix of the multidimensional spatiotemporal potential energy field based on the distribution of the high potential energy repulsion zone and the low potential energy attraction zone.

[0038] This embodiment further elaborates on the terrain gradient generation logic in the potential energy field construction module; the system calls the equipment load rate of each machine tool. and current idle time window The system executes the high potential energy repulsion zone definition step, defining the area where the device load rate is higher than a preset load threshold, such as 85%, as the high potential energy repulsion zone; the system executes the low potential energy attraction zone definition step, defining the current idle time window as the low potential energy attraction zone.

[0039] To achieve code-level field strength calculation, this step constructs a scalar function for the multidimensional spatiotemporal potential energy field. ,in, From a device perspective, it should be noted that, in order to support gradient calculations in continuous fields, the system pre-maps discrete machine tool IDs to continuous virtual topological coordinates. The specific mapping steps are as follows:

[0040] Constructing the device feature matrix Each row represents the function vector of a device. Before constructing the matrix, the system performs device function clustering, grouping devices with the same processing capabilities into one group. Within each group, principal component analysis is used to transform the high-dimensional feature matrix. Dimensionality reduced to one-dimensional scalar space, the first principal component is extracted as the virtual topological coordinates of the device. ; and through formula Add inter-group offsets, where... This is the group number index to which the device belongs; specifically, The value of strictly follows the formula ,in, The maximum feature span within a single device group, for example, the value. ,but Set as This ensures that different device groups form a clear deep potential well isolation in virtual space;

[0041] in, The span should be greater than 1.5 times the maximum coordinate span within the equipment group to ensure that equipment groups with different functions can... The axes provide significant coordinate isolation, preventing particles from accidentally entering non-functionally matched regions during evolution; for example, mapping lathe groups to intervals. Milling machine group mapped to interval and in the interval Constructing a high potential energy barrier; when task particles need to flow from the turning process to the milling process, although gradient guidance is hindered, the thermal fluctuation term in the embodiment helps to mitigate this. Particles have a certain probability of tunneling through this potential energy barrier, thus achieving global optimization across device groups; it should be noted that, in order to ensure gradient... In terms of physical validity, the system stipulates that local gradients are calculated only within the functional range of devices in the same group. When a particle is in the potential energy barrier region between two device groups, its motion is mainly driven by thermal fluctuation terms to make random jumps, rather than relying on a fixed gradient descent, thereby avoiding the particle from erroneously sliding to the coordinates of a device with mismatched functions due to dimensional compression.

[0042] Coordinates of all devices Normalize it to make it distributed in Within a continuous interval, this ensures that devices with similar functions are adjacent in the potential energy field, thus making the gradient... This has physical guiding significance; to prevent gradient-guided particles from sliding towards the coordinates of nearby devices that lack processing capabilities, the system further performs a functional barrier construction step: defining a set of non-functionally matched regions. For any device coordinates This means that the machine tool corresponding to this coordinate does not have the technological capability required for the current process, coupled with the potential energy of a high-steep soft barrier.

[0043]

[0044] in, This represents the barrier penalty coefficient. Values This ensures that a near-rigid body-like rebound force is generated at the boundary. For the power of the barrier steepness, even numbers are preferred. or This is to ensure the non-negativity of the potential energy function and the sharp increase of the derivative at the boundary; These are the boundary coordinates of the non-functional matching area of ​​the device;

[0045] This potential energy barrier ensures the topographic gradient. Instead of an infinite amount that would cause numerical overflow, a huge reverse repulsive force is generated at the functional boundary, thus constraining the particles to evolve only within the functionally matched device group.

[0046] For the time dimension, the specific formula is as follows:

[0047]

[0048] The first term is the repulsive potential energy based on the Sigmoid function. For equipment At any moment Real-time load rate, The preset load threshold is set to 0.85. The steepness coefficient is set to 10.0. The first term is the repulsion coefficient; the second term is the attractive potential energy well based on the Gaussian function. For equipment The set of all idle time windows on the screen. For time windows The central moment, The variance parameter is positively correlated with the width of the time window. The above coefficient is the attraction coefficient; and It can be obtained through training on historical production data, or set by the user according to the scheduling strategy; the specific parameter settings and calculation methods are as follows: Dimensions: and All of these are parameters in the dimension of energy, with typical setpoints as follows: and To ensure that the load repulsive force is numerically slightly greater than the idle attractive force, thus preventing overload; load rate function Calculation: Trilinear interpolation is used, based on discrete equipment status snapshot data, with time granularity. In continuous spacetime coordinates Calculate smooth load values ​​to ensure gradient. Continuously differentiable; variance parameter Follow the formula ,in, For idle time windows End time, For idle time windows The start time, This is the window width coefficient, with values ​​ranging from [value missing]. This ensures that the coverage of the attraction potential trap is positively correlated with the actual idle time window width and is center-aligned;

[0049] Based on the above scalar field The terrain gradient matrix of the multidimensional spatiotemporal potential energy field is generated by using the discrete difference operator. Each element in the matrix It indicates the direction of the force on the particle in the spacetime grid, enabling the particle to automatically avoid the high-load area, i.e. high potential energy, and slide to the empty window, i.e. low potential energy.

[0050] This embodiment achieves a continuous field description of workshop resource status by constructing a hybrid potential energy field function containing Sigmoid repulsion and Gaussian attraction terms; the peak effect in the high-load area automatically diverts tasks to prevent resource congestion, while the valley effect in the low-load area automatically absorbs tasks, improving equipment utilization; this field construction method based on a clear mathematical model ensures the convergence of evolutionary calculations and the clarity of physical meaning.

[0051] Example 4:

[0052] The interaction force between particles is calculated based on spin properties, including: identifying the spin properties of adjacent particles in a virtual particle flow; if adjacent particles have the same spin properties, a virtual attraction force is generated between the particles to promote the spatiotemporal clustering of task particles with the same tooling requirements; if adjacent particles have different spin properties, a virtual repulsion force is generated between the particles to characterize the damping cost generated by the tool changing operation.

[0053] This embodiment further elaborates on the calculation logic of inter-particle interaction forces in the evolutionary computation module; in the virtual spacetime, the system identifies adjacent particles in the virtual particle stream, that is, tasks that are processed sequentially on the same device; that is, in the current evolutionary virtual spacetime, the Euclidean distance to the target particle is less than the preset cutoff radius. The system determines the spin properties of adjacent particles; in response to adjacent particles having the same spin property (i.e., using the same tool), the system generates a virtual attraction force between the particles. ; The source is calculated by the inverse ratio of the spin vector dot product to the square of the distance. Its physical meaning is that the formula for the force that causes similar tasks to cluster together is:

[0054]

[0055] in, To prevent the distance softening constant from having a denominator of zero, a value is taken as follows: ; It is the attractive coupling constant; This is the weighted Euclidean distance of a particle in spacetime, which is the rigorous mathematical expression of the aforementioned time distance.

[0056] In response to the different spin properties of adjacent particles, the system generates a virtual repulsive force between the particles. In addition, the system also performs a timing constraint calculation step: identifying adjacent particle pairs that belong to the same production order and have sequential process relationships. ,in, For the preceding process, For the subsequent process; if the time coordinate of the subsequent process Time coordinates smaller than those of the preceding process ,Right now This generates a strong unidirectional repulsive force between the two particles. :

[0057]

[0058] in, For time series weights, To penalize steepness, this force ensures that the processes are arranged strictly in the time sequence of the process route; The spin coupling strength coefficient is given by a value of [value missing]. This is used to define the tightness of clustering similar tasks; Value , Value This set of highly weighted parameters ensures that the process sequence constraints have physically hard spherical characteristics, making... The reversed state is almost impossible to exist stably during evolution; all the above parameters have been dimensionless to match the total energy level of the system.

[0059] The source is spin difference, and its physical meaning is to characterize the damping cost generated by the tool change operation; the repulsive force will push away the time distance between the two tasks, and this distance physically corresponds to the actual tool change time.

[0060] This embodiment simulates the spin interaction between particles, automatically generating grouped technical features in the scheduling results; virtual attraction promotes continuous machining of the same tool task, while virtual repulsion automatically reserves tool change time. This mechanism reduces changeover cost (SetupCost) without explicit programming, achieving synergistic optimization of process clustering and machining efficiency.

[0061] Example 5:

[0062] The system also includes a dynamic order insertion response module, which is used to: monitor whether there are any new emergency order insertion tasks entering the production order set; in response to the detection of new emergency order insertion tasks, map the emergency order insertion tasks as high-energy perturbation particles; inject the high-energy perturbation particles into the current multidimensional spatiotemporal potential energy field, triggering the redistribution of local field strength; and based on the redistributed local field strength, perform local relaxation evolution of the virtual particle flow until the energy balance state is reached again.

[0063] This embodiment further elaborates on the system's dynamic response mechanism for emergency order insertion, focusing on how to achieve flexible adjustments to the existing schedule through physical field equations; the system monitors in real time whether new emergency order insertion tasks are added to the production order set; in response to detecting an order insertion, the system maps it to a high-energy perturbation particle; this particle has extremely high energy density and thermal driving force, specifically defined as: setting the mass of the perturbation particle. ,in, The average mass of particles in the current field. This is a priority coefficient, ranging from [5.0, 10.0], and is given a fixed and unchangeable spatiotemporal coordinate constraint, i.e., the forced execution time of the insertion. It should be noted that this high-energy perturbation particle is set as a static field source in subsequent evolution calculations, and its velocity vector... It is always forcibly set to 0, meaning that its spatiotemporal coordinates are not updated with evolution, but only continuously release a repulsive potential energy field into the surrounding space;

[0064] The system injects high-energy perturbation particles into the current multidimensional spacetime potential energy field, triggering a redistribution of the local field strength. This redistribution does not rewrite the entire potential energy field matrix, but rather achieves it in the form of a superposition field, with the system constructing a perturbation potential energy function. :

[0065]

[0066] in, This is the height of the high-energy repulsion barrier. This is the target location for the insertion task. The radius of influence of the disturbance; here, The value is determined based on the current average potential energy of the system. times, that is This is to ensure that the barrier formed by the insert task is high enough to repel all the surrounding regular task particles. The value is determined based on the characteristic span of the equipment group, and is usually set to . This limits the disturbance to a localized area within the relevant equipment group, without affecting the entire plant schedule; Superimposed on the original multidimensional spacetime potential energy field ;

[0067] Based on the redistributed local field strength, the system performs local relaxation evolution of the virtual particle flow; the system utilizes spatial grid hash indexing to quickly locate the insertion position at a distance exceeding [a certain value]. The far-end particles are only released within the perturbed region; the perturbed particles are in the new gradient field. Under its influence, you will feel the effect from The strong repulsive force causes them to automatically slide into the surrounding low-potential idle time windows; the evolution process continues until the resultant force modulus of all particles in the disturbed region reaches its maximum. Less than the convergence threshold , For example, the dimensionless force threshold relative to the average particle mass. If the number of evolution iterations reaches the preset maximum threshold, such as 1000 times, the system will reach an energy balance state or metastable state again. If the system fails to converge after reaching the maximum threshold, it will automatically lock the configuration with the lowest energy and output a warning.

[0068] This embodiment utilizes the physical perturbation characteristics of high-energy particles to achieve a millisecond-level response to emergency orders; by superimposing a Gaussian repulsive potential energy field and limiting the relaxation range, the system minimizes disruption to the established plan while accepting new tasks, demonstrating the production plan's elastic self-healing capability in the face of sudden demands.

[0069] Example 6:

[0070] Based on the principle of minimizing free energy, the virtual particle flow is driven to undergo gradient evolution in a multidimensional spatiotemporal potential energy field. This includes: defining the total system energy function, which is composed of a weighted sum of the delay penalty potential energy, the replacement cost potential energy, and the equipment idle potential energy; calculating the resultant force vector of each task particle within the virtual time step; and updating the position coordinates of each task particle according to the resultant force vector so that the value of the total system energy function approaches the minimum value.

[0071] This embodiment further elaborates on the specific mathematical implementation of the free energy minimization principle in the evolutionary calculation module; in order to quantify the merits of the scheduling scheme, the system's total energy function... The mathematical expression is constructed as follows:

[0072]

[0073] Specifically, it can be elaborated as follows:

[0074]

[0075] in, This is the set of all task pairs with process sequence constraints. This is the weight of idle cost per unit time, with the dimension [energy / time]. It is used to convert the time dimension obtained by integration into the energy dimension to ensure that the physical units of each item in the formula are consistent.

[0076] The first item is the potential energy of delayed penalty. An exponential soft-constraint potential function is adopted. For the first The standard processing time for each task particle is used to ensure that the gradient is continuous and non-zero throughout the entire spatiotemporal domain; where For steepness factor, for example, a value of 0.5, when A weak repulsive force is provided to maintain gradient orientation when approaching or exceeding [a certain value]. It provides a repulsive force for exponential growth; This refers to the delivery deadline; further definition is provided below: For the first Standard processing time for each task particle;

[0077] The second item is the potential energy of replacement cost. A barrier is built based on the One-Hot encoding feature. When and Different (i.e.) When the preceding term takes effect, it generates high potential energy to push the particles away, thus allowing time for tool replacement; when and Same (i.e.) When the latter takes effect, it reduces system energy to promote task clustering; The weighted Euclidean distance of a particle in spacetime is specifically defined as:

[0078]

[0079] in, , These are used to divide the device dimensions. With the time dimension Mapped to a dimensionless unit interval, The total span of the equipment coordinates. To determine the scheduling time span, ensure that distance calculations have physical meaning; It is a neighborhood set dynamically determined based on real-time location;

[0080] The third item is the idle potential energy of the equipment. It refers to the equipment occupancy rate. The integral of the complement of; where, It is a binary indicator function if and only if any device in the system at time t = 0. The value is 1 when the task is occupied, and 0 otherwise.

[0081] This is the upper limit of the total duration of the scheduling cycle; where, To prevent numerical stability constants with a denominator of zero, a value of [value missing] is taken. ; The steepness factor for the delayed penalty, in units of ; These are all preset dimensionless weighting coefficients used to convert different physical dimensions, such as time and distance, into energy units. These weighting coefficients can be optimized through training with historical data or set by the user according to the current production strategy. To enable those skilled in the art to implement this, the specific definitions and recommended values ​​for each weighting coefficient are provided below: (Delay Weight): Value This is used to convert the time-dimension delay penalty into an energy value of the same order of magnitude as the spatial potential energy. (Tool Repulsion Weight): Value This characterizes the energy loss corresponding to a single tool change operation. (Similarity Attraction Weight): Value The value is slightly smaller than the repulsion weight to avoid excessive clustering that leads to long waiting times. (Idle Weight): Energy Time for Value Acquisition As a background penalty term; in addition, the distance normalization parameter and The specific calculation method is as follows: , ,in, The total number of available devices. The average distribution width of the device coordinates. This represents the latest delivery deadline in the order set, thus mapping physical distances of different dimensions to... Dimensionless values ​​within an interval;

[0082] Prior to this, the system needs to define the temperature evolution function for simulated annealing. To control the convergence speed of the system; specifically, an exponential cooling scheme is adopted:

[0083]

[0084] in, The initial virtual temperature is set to a value. ; The cooling coefficient is given by the following values: ; The number of time steps for temperature updates; this function ensures that the system experiences high thermal fluctuations in the early stages of evolution to escape local minima, while the temperature approaches zero in the later stages to converge to a steady state; in the virtual time step... Inside, calculate the resultant force vector acting on each task particle. To transform the static thermal-driven properties defined in Example 2 into driving terms in the dynamic equations, and to ensure the convergence of the algorithm by conforming to the fluctuation dissipation theorem in physics, the formula for calculating the resultant force vector is modified as follows to strictly conform to the fluctuation dissipation theorem in statistical mechanics:

[0085]

[0086] at this time, The first item In terms of dynamics, the thermal driving force of Example 2 is specifically implemented. Although the mathematical form is modified from the reciprocal mapping to the exponential mapping to optimize convergence, its physical essence is the same, that is, its magnitude increases exponentially as the deadline approaches, automatically pushing the particles to accelerate in the negative direction of the time axis. The field coupling coefficient is... The terrain gradient matrix is ​​the force that guides particles to slide towards areas with low load and open windows. This is the viscous drag term; the last term is the random fluctuation force term; according to the fluctuation dissipation theorem, the amplitude of the random force must be equal to the damping coefficient. and temperature It is directly proportional to the square root, that is This ensures that the system eventually converges to a Boltzmann distribution; among which, The virtual Boltzmann constant is used to adjust the intensity of random forces; the key physical parameters in the above dynamic equations are defined as follows: (Field coupling coefficient) is dimensionless, and its value ranges from... Used to convert terrain gradients Directly mapped to force; (Viscous damping coefficient) Dimensions are mass-time value range This parameter determines the system's energy dissipation rate. If the value is too small, the particle oscillations will not converge. Excessive size can lead to evolutionary stagnation; Value (In a dimensionless system), this is used to define the reference energy level for thermal fluctuations; this system constructs a system that satisfies the following conditions during discretization. The noise generator ensures that the energy distribution over long-term evolution follows a Gibbs distribution. ; This is a Gaussian white noise vector with a mean of 0 and a variance of 1.

[0087] Update the position coordinates of each task particle based on the resultant force vector. Includes time coordinates and device coordinates It should be noted that, in order to eliminate the dimensional differences between different physical quantities, the system pre-calculates all time variables. and device coordinates Dimensionless normalization was performed, making... It becomes a pure scalar; specifically, the Velocity-Verlet integration algorithm is used:

[0088]

[0089] in, The dimensionless inertial mass is defined in Example 2. For dimensionless virtual resultant force, To establish a virtual evolution step size, ensuring dimensional consistency on both sides of the equation; The value of must satisfy the numerical stability condition, and the specific calculation formula is as follows: Under typical operating conditions, set (Virtual time unit) to ensure integral stability in high curvature potential energy field regions and prevent particle escape;

[0090] Through iterative updates, the value of the total energy function of the system is continuously brought closer to the minimum value while satisfying the physical field constraints. The final particle distribution obtained is the scheduling scheme with the lowest overall cost.

[0091] This embodiment defines a Hamiltonian with explicit mathematical constraints and explicitly incorporates the potential energy field gravity and thermal driving force into the dynamic equation, transforming the complex multi-objective optimization problem of production scheduling into a global optimization problem of non-convex functions. The system can not only automatically weigh the costs of delays, model changes, and idle time, but also guide particles to automatically avoid congestion areas using field gradients and ensure urgent orders using thermal driving force, thus ensuring high-quality scheduling results under complex constraints.

[0092] Example 7:

[0093] The scheduling generation module is specifically used to: extract the time dimension component and the equipment dimension component from the steady-state coordinates; map the equipment dimension component to the target machine tool ID; map the time dimension component to the process start time and end time; and generate production scheduling instructions containing material delivery time points and G-code call sequences based on the target machine tool ID, process start time, and end time.

[0094] This embodiment further illustrates the decoding logic of the scheduling generation module, namely, how to discretize the continuous physical field steady-state deconstruction into industrial control commands; when the system energy converges, the system extracts steady-state coordinates; these coordinates are multi-dimensional vectors. ,in, and All are consecutive floating-point numbers;

[0095] The system executes the device mapping step, mapping the device dimension components to specific target machine tool IDs; specifically, the nearest neighbor adsorption algorithm is used for discretization. It should be noted that the device virtual topology coordinate mapping table generated in Example 3 needs to be called here. ,in, Indicates device The corresponding coordinate values ​​in a continuous potential energy field; the calculation formula is:

[0096]

[0097] in, This is a set of integer indices for available machine tools; indexes are used here. This is to avoid the physical constants in Example 2 and quality A symbol definition conflict occurs; if multiple particles adsorb to the same machine tool ID and the time overlaps, the conflict is determined by the particle's final energy value. The system sorts devices by energy level, prioritizing those with lower energy (higher compatibility) and relocating them to the next best neighbor ID. Devices with higher energy are then automatically moved to the next best neighbor ID. Modified to the suboptimal solution And apply a random micro-perturbation. This triggers a new round of local gradient descent;

[0098] The system performs a time mapping step, mapping the time dimension components to specific process start and end times; the calculation formula is:

[0099] ;

[0100]

[0101] in, The minimum scheduling time granularity, such as 1 minute; The time dimension component in the extracted steady-state coordinates;

[0102] Based on the above determination and time window The system generates production scheduling instructions that include material delivery times and G-code call sequences; the system uses... That is, the lead time, which generates AGV delivery tasks for the trigger point, and uses Using the index key, retrieve the corresponding post-processor from the DNC database and generate a program in the form of...<Time:10:00,Machine:MC-01,Program:O1234.NC> The instruction package;

[0103] This embodiment realizes the lossless transformation from the steady-state solution of the abstract physical field to specific industrial control instructions; through a clear discretization algorithm and conflict resolution mechanism, the system not only determines the processing schedule, but also generates the calling instructions for logistics and CNC programs simultaneously, realizing closed-loop control from the algorithm layer to the execution layer and ensuring the executability of the scheduling scheme.

[0104] Example 8:

[0105] The system also includes a bottleneck visualization module, which is used to: acquire pressure distribution data of the multidimensional spatiotemporal potential energy field after it has evolved and stabilized; identify regions in the pressure distribution data where the potential energy density is higher than a preset warning value; mark the regions as resource bottleneck nodes; and generate equipment supplementation suggestions or process diversion strategies for resource bottleneck nodes.

[0106] This embodiment further illustrates the system's bottleneck visualization function; the system acquires pressure distribution data of the multidimensional spatiotemporal potential energy field after its evolution and stabilization; the pressure distribution data comes from field calculation results, and its physical meaning is a scalar field of particle density within a unit spatiotemporal region; the system identifies regions in the pressure distribution data where the potential energy density is higher than a preset warning value; high pressure means that a large number of particles are trying to squeeze into the equipment during this time period, but cannot enter due to potential energy repulsion; the system marks the region as a resource bottleneck node and generates suggestions; for equipment addition suggestions, if a certain type of machine tool is under high pressure for a long time, it is recommended to purchase a new machine; for process diversion strategies, it is recommended to adjust the process route and divert some processes to alternative equipment in low-pressure areas;

[0107] This embodiment uses the concept of pressure in a physical field to intuitively display production bottlenecks. Compared with traditional report analysis, it can more dynamically reflect the deep contradictions between resource supply and demand. By identifying high-pressure areas, managers can accurately locate the limiting factors of the system and formulate scientific capacity expansion or process adjustment strategies accordingly, thus realizing data-driven decision support.

[0108] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A production scheduling management system for milling and turning machining centers, characterized in that, include: The data acquisition module is used to acquire the set of production orders to be scheduled and the current resource status vector of the machining center. The set of production orders contains multiple sets of process task data, including: process feature sequence, standard machining time, delivery deadline, and tooling requirement characteristics. The resource status vector includes: equipment load rate, remaining tool life, and current idle time window. The particle mapping module is used to map the set of production orders into a virtual particle flow based on the process feature sequence, the standard machining time, and the tooling requirement characteristics. Each task particle in the virtual particle flow is assigned a spin attribute corresponding to the tooling requirement characteristics and a mass attribute corresponding to the standard machining time. Potential energy field construction. The module is used to construct a multidimensional spatiotemporal potential energy field based on the equipment load rate and the current idle time window, wherein the available resource capacity of the processing center is mapped to a gravitational potential energy well in the multidimensional spatiotemporal potential energy field; the evolution calculation module is used to drive the virtual particle flow to perform gradient evolution motion in the multidimensional spatiotemporal potential energy field based on the principle of minimizing free energy, wherein the evolution motion includes: calculating the interaction force between particles based on the spin property, and calculating the field gravity on the particles based on the gravitational potential energy well, until the total energy of the system converges; the scheduling generation module is used to lock the steady-state coordinates of the virtual particle flow in the multidimensional spatiotemporal potential energy field in response to the convergence of the total energy of the system, and decode the steady-state coordinates into production scheduling instructions.

2. The production scheduling management system for turning and milling machining centers according to claim 1, characterized in that, Based on the process feature sequence, the standard machining time, and the tooling requirement features, the production order set is mapped to a virtual particle flow, including: invoking the tooling requirement features and the standard machining time; defining process tasks with the same tooling requirement features as particles with the same spin direction; defining the standard machining time as the particle's inertial mass; defining the difference between the delivery deadline and the current time as the particle's thermal driving force; and generating the initial state vector of the virtual particle flow based on the spin direction, the inertial mass, and the thermal driving force.

3. The production scheduling management system for turning and milling machining centers according to claim 1, characterized in that, Based on the device load rate and the current idle time window, a multidimensional spatiotemporal potential energy field is constructed, including: calling the device load rate and the current idle time window; defining the region where the device load rate is higher than a preset load threshold as a high potential energy repulsion zone; defining the time period of the current idle time window as a low potential energy attraction zone; and generating the terrain gradient matrix of the multidimensional spatiotemporal potential energy field based on the distribution of the high potential energy repulsion zone and the low potential energy attraction zone.

4. The production scheduling management system for turning and milling machining centers according to claim 1, characterized in that, Calculating the interaction force between particles based on the spin property includes: identifying the spin properties of adjacent particles in the virtual particle flow; if adjacent particles have the same spin property, generating a virtual attraction force between particles to promote the spatiotemporal clustering of task particles with the same tooling requirements; if adjacent particles have different spin properties, generating a virtual repulsion force between particles to characterize the damping cost generated by the tool changing operation.

5. The production scheduling management system for turning and milling machining centers according to claim 1, characterized in that, The system also includes a dynamic order insertion response module, used to: monitor whether any new emergency order insertion tasks have entered the production order set; in response to the detection of a new emergency order insertion task, map the emergency order insertion task as a high-energy perturbation particle; inject the high-energy perturbation particle into the current multidimensional spatiotemporal potential energy field, triggering a redistribution of the local field strength; and based on the redistributed local field strength, perform the local relaxation evolution of the virtual particle flow until an energy balance state is reached again.

6. The production scheduling management system for turning and milling machining centers according to claim 1, characterized in that, Based on the principle of minimizing free energy, the virtual particle flow is driven to undergo gradient evolution in the multidimensional spatiotemporal potential energy field, including: defining a total system energy function, which is composed of a weighted sum of delay penalty potential energy, replacement cost potential energy, and equipment idle potential energy; calculating the resultant force vector of each task particle within a virtual time step; and updating the position coordinates of each task particle according to the resultant force vector so that the value of the total system energy function approaches its minimum value.

7. The production scheduling management system for turning and milling machining centers according to claim 6, characterized in that, The scheduling generation module is specifically used for: extracting the time dimension component and the equipment dimension component from the steady-state coordinates; mapping the equipment dimension component to the target machine tool ID; mapping the time dimension component to the process start time and end time; and generating a production scheduling instruction containing material delivery time points and G-code call sequences based on the target machine tool ID, the process start time, and the end time.

8. The production scheduling management system for turning and milling machining centers according to claim 1, characterized in that, The system also includes a bottleneck visualization module, used to: acquire pressure distribution data of the multidimensional spatiotemporal potential energy field after it has evolved and stabilized; identify regions in the pressure distribution data where the potential energy density is higher than a preset warning value; mark the regions as resource bottleneck nodes; and generate equipment supplementation suggestions or process diversion strategies for the resource bottleneck nodes.

Images