Methods and Systems for Optimizing Cost Operations in CNC Shops

By using IoT data collection and fusion technology, the CNC workshop operation plan was optimized, which solved the problem of cost and efficiency imbalance in multi-variety, small-batch processing, and achieved dynamic cost control and improved equipment utilization.

CN123113752APending Publication Date: 2026-11-17FUJIAN KEYE CNC TECH CO LTD +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611363543.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-04
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing CNC workshop operation planning systems struggle to achieve a precise balance between cost and efficiency when faced with complex machining scenarios involving multiple product types and small batches. They are unable to perceive machine tool load and tool wear status in real time, leading to premature tool failure or waste, high equipment energy consumption, and increased operating costs.

Method used

By collecting multi-source heterogeneous data through the Internet of Things, performing timestamp alignment and feature fusion, extracting spindle load and tool wear characteristics, constructing an optimization solution space, generating an initial flexible job scheduling scheme, and performing adaptive rescheduling under dynamic disturbance events to optimize the matching relationship between processes, machine tools, and tools.

Benefits of technology

It enables dynamic control of CNC workshop operating costs, improves equipment utilization and order delivery capabilities, reduces overall manufacturing costs, and enhances the ability to make flexible decisions in the face of uncertainties.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123113752A_ABST
    Figure CN123113752A_ABST
Patent Text Reader

Abstract

This invention provides a method and system for optimizing the operation and management of flexible work planning in CNC workshops, relating to the field of production scheduling optimization technology. The method includes: performing timestamp alignment and feature fusion on multi-source heterogeneous data to obtain a micro-level process-level production dataset; based on the micro-level process-level production dataset, extracting the spindle load timing characteristics and tool wear rate characteristics of specific processes on specific machine tools, calculating the corresponding dynamic energy consumption cost, tool wear cost, and failure risk cost, forming micro-level dynamic machining cost data with process, machine tool, tool matching, and tool feed trajectory parameters as independent variables. This invention reduces the deviation between planned costs and actual execution costs, improving the controllability and real-time response of flexible work planning in dynamic machining environments.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of production scheduling optimization technology, and in particular to a method and system for optimizing the operation and cost of flexible work plans in index-controlled workshops. Background Technology

[0002] In flexible operation management in CNC workshops, most existing solutions rely on the Production Execution System (MES) combined with preset process parameters to formulate work plans. This approach can maintain basic operational order when dealing with routine and stable production tasks, but it may be difficult to achieve a refined balance between cost and efficiency when facing complex processing scenarios with multiple varieties and small batches.

[0003] Current work planning processes mostly rely on historical average data or theoretical standard values ​​to set machining parameters, such as fixed spindle speed, feed rate, and estimated tool life. However, in actual machining, the spindle load, tool wear, and process execution time are significantly affected by various dynamic factors, such as material batch differences, real-time equipment operating conditions, and ambient temperature and humidity. Existing management systems lack the ability to collect and integrate real-time data at this micro level, making it difficult to accurately capture the actual cost changes of specific processes executed on specific machine tools. For example, when machining a batch of aerospace structural parts, the local hardness of different workpiece materials may have slight differences, resulting in inconsistent wear rates of the same tool when machining different workpieces. If scheduling is still based on a preset fixed tool life, premature tool failure may lead to workpiece scrapping, or tools may be replaced prematurely before reaching their limit, resulting in waste. At the same time, because the dynamic changes in spindle load cannot be perceived in real time, it is difficult to adjust the machining sequence or equipment allocation according to actual energy consumption and failure risk. This can easily lead to high-energy-consuming or high-risk processes being concentrated in certain time periods or on certain equipment, increasing overall operating costs and the probability of unexpected downtime. Summary of the Invention

[0004] This invention provides a method and system for optimizing the operation and management of flexible work planning costs in CNC workshops, reducing the deviation between planned costs and actual execution costs, and improving the controllability and real-time response of flexible work planning in dynamic processing environments.

[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:

[0006] Firstly, a method for optimizing the operation and cost management of flexible work planning in CNC workshops, the method comprising:

[0007] Step 1: Collect spindle running status data, feed speed data, process execution status data, RFID-read tool remaining life data, and order delivery date data of heterogeneous CNC machine tools through the workshop Internet of Things to obtain multi-source heterogeneous data;

[0008] Step 2: Perform timestamp alignment and feature fusion on the multi-source heterogeneous data to obtain a micro-process-level production dataset;

[0009] Step 3: Based on the micro-process-level production dataset, extract the spindle load timing characteristics and tool wear rate characteristics of a specific process on a specific machine tool, calculate the corresponding dynamic energy consumption cost, tool wear cost and failure risk cost, and form micro-dynamic machining cost data with process, machine tool, tool matching and tool feed trajectory parameters as independent variables.

[0010] Step 4: Based on the micro-dynamic processing cost data, construct an optimization solution space with the objective function of minimizing the overall operating cost and the constraints of the process sequence and equipment capacity. Encode the matching relationship between different processes, machine tools and cutting tools and the tool feed trajectory parameters into population individuals. Calculate the fitness value of each population individual in the optimization solution space, and iteratively update the population through selection, crossover and mutation operations until the preset convergence condition is met to obtain the initial flexible operation scheduling scheme.

[0011] Step 5: Based on the initial flexible operation scheduling scheme, monitor dynamic disturbance events in the processing process in real time. When a dynamic disturbance event is triggered, recalculate the micro-dynamic processing cost data of the remaining unexecuted processes based on the real-time updated micro-process-level production dataset, and perform adaptive rescheduling calculation on the initial flexible operation scheduling scheme to obtain the updated flexible operation scheduling scheme, thereby controlling the processing flow of physical entities in the workshop.

[0012] Secondly, the CNC workshop flexible operation planning cost optimization management system includes:

[0013] The data acquisition module is used to collect spindle operating status data, feed speed data, process execution status data, RFID-read tool remaining life data, and order delivery date data of heterogeneous CNC machine tools through the workshop Internet of Things, so as to obtain multi-source heterogeneous data.

[0014] The fusion module is used to perform timestamp alignment and feature fusion on multi-source heterogeneous data to obtain micro-level production datasets.

[0015] The calculation module is used to extract the spindle load timing characteristics and tool wear rate characteristics of a specific process on a specific machine tool based on the micro-process-level production dataset, and calculate the corresponding dynamic energy consumption cost, tool wear cost and failure risk cost, forming micro-dynamic machining cost data with process, machine tool, tool matching and tool feed trajectory parameters as independent variables.

[0016] The optimization module is used to construct an optimization solution space based on micro-dynamic processing cost data, with the objective function of minimizing the overall operating cost and the constraints of the process sequence and equipment capacity. It encodes the matching relationship between different processes, machine tools and cutting tools and the tool feed trajectory parameters as population individuals, calculates the fitness value of each population individual in the optimization solution space, and iteratively updates the population through selection, crossover and mutation operations until the preset convergence condition is met, and obtains the initial flexible operation scheduling scheme.

[0017] The response module is used to monitor dynamic disturbance events in the processing process in real time according to the initial flexible operation scheduling scheme. When a dynamic disturbance event is triggered, it recalculates the micro-dynamic processing cost data of the remaining unexecuted processes based on the real-time updated micro-process-level production dataset, and performs adaptive rescheduling calculation on the initial flexible operation scheduling scheme to obtain the updated flexible operation scheduling scheme, thereby controlling the processing flow of physical entities in the workshop.

[0018] Thirdly, a computing device includes:

[0019] One or more processors;

[0020] A storage device for storing one or more programs that, when executed by one or more processors, cause the one or more processors to implement the method.

[0021] Fourthly, a computer-readable storage medium storing a program that, when executed by a processor, implements the method.

[0022] The above-described solution of the present invention has at least the following beneficial effects:

[0023] An optimization solution space oriented towards minimizing overall operating costs is constructed using process, machine tool, tool matching, and feed trajectory parameters as independent variables. An initial flexible job scheduling scheme is generated through population encoding and genetic iteration, which takes into account multiple constraints of process sequence and equipment capacity, and avoids global imbalance caused by local optimization of a single equipment or process. An adaptive rescheduling mechanism driven by dynamic disturbance events is introduced. When a sudden disturbance occurs during the processing, the dynamic cost of the remaining unexecuted processes can be quickly recalculated based on the real-time updated micro dataset and the scheduling scheme can be adjusted. This enables dynamic control of the entire CNC workshop's operating costs, synergistic improvement of equipment utilization and order delivery capabilities, reduces overall manufacturing costs, and enhances the workshop's flexible decision-making ability to cope with uncertain disturbances. Attached Figure Description

[0024] Figure 1This is a flowchart illustrating the CNC workshop flexible operation planning cost optimization management method provided in an embodiment of the present invention.

[0025] Figure 2 This is a schematic diagram of the CNC workshop flexible operation planning cost optimization management system provided in an embodiment of the present invention.

[0026] Figure 3 This is a schematic diagram of the initial flexible job scheduling scheme provided in an embodiment of the present invention. Detailed Implementation

[0027] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0028] like Figure 1 As shown, embodiments of the present invention propose a method for optimizing the operation and cost management of flexible work planning in CNC workshops. The method includes the following steps:

[0029] Step 1: Collect spindle running status data, feed speed data, process execution status data, RFID-read tool remaining life data, and order delivery date data of heterogeneous CNC machine tools through the workshop Internet of Things to obtain multi-source heterogeneous data;

[0030] Step 2: Perform timestamp alignment and feature fusion on the multi-source heterogeneous data to obtain a micro-process-level production dataset;

[0031] Step 3: Based on the micro-process-level production dataset, extract the spindle load timing characteristics and tool wear rate characteristics of a specific process on a specific machine tool, calculate the corresponding dynamic energy consumption cost, tool wear cost and failure risk cost, and form micro-dynamic machining cost data with process, machine tool, tool matching and tool feed trajectory parameters as independent variables.

[0032] Step 4: Based on the micro-dynamic processing cost data, construct an optimization solution space with the objective function of minimizing the overall operating cost and the constraints of the process sequence and equipment capacity. Encode the matching relationship between different processes, machine tools and cutting tools and the tool feed trajectory parameters into population individuals. Calculate the fitness value of each population individual in the optimization solution space, and iteratively update the population through selection, crossover and mutation operations until the preset convergence condition is met to obtain the initial flexible operation scheduling scheme.

[0033] Step 5: Based on the initial flexible operation scheduling scheme, monitor dynamic disturbance events in the processing process in real time. When a dynamic disturbance event is triggered, recalculate the micro-dynamic processing cost data of the remaining unexecuted processes based on the real-time updated micro-process-level production dataset, and perform adaptive rescheduling calculation on the initial flexible operation scheduling scheme to obtain the updated flexible operation scheduling scheme, thereby controlling the processing flow of physical entities in the workshop.

[0034] In this embodiment of the invention, an optimization solution space oriented towards minimizing overall operating costs is constructed using process, machine tool, tool matching, and feed trajectory parameters as independent variables. An initial flexible job scheduling scheme is generated through population encoding and genetic iteration, which takes into account multiple constraints of process sequence and equipment capacity, and avoids global imbalance caused by local optimization of a single equipment or process. An adaptive rescheduling mechanism driven by dynamic disturbance events is introduced. When a sudden disturbance occurs during the processing, the dynamic cost of the remaining unexecuted processes can be quickly recalculated based on the real-time updated micro dataset, and the scheduling scheme can be adjusted. This achieves dynamic control of the entire process operating cost of the CNC workshop, synergistic improvement of equipment utilization and order delivery capability, reduces overall manufacturing costs, and improves the workshop's flexible decision-making ability to cope with uncertain disturbances.

[0035] In a preferred embodiment of the present invention, step 1 above may include:

[0036] In this embodiment of the invention, corresponding data acquisition adaptation schemes are adopted for heterogeneous CNC machine tools of different brands, models, and control systems within the workshop. For machine tools with standard industrial communication interfaces, Ethernet communication and fieldbus communication are used to connect to the CNC system. For older heterogeneous machine tools without standard external communication interfaces, physical signal acquisition is achieved through external industrial sensors and data acquisition modules. At the same time, high-frequency RFID reading terminals are deployed at the tool transfer station, tool storage station, and machine tool tool magazine station in the workshop. The networking binding, frequency band debugging, and tag identification calibration of all RFID devices are completed. The clock of all equipment acquisition terminals, sensor terminals, and RFID reading terminals is synchronized across the network to ensure that the timestamp reference of all acquisition nodes is consistent, eliminating the problem of timing offset between different hardware terminals and laying a unified time foundation for subsequent timing alignment processing of multi-source data.

[0037] Based on the completed and calibrated workshop IoT system, data is collected from each heterogeneous CNC machine tool throughout its entire machining cycle. Complete spindle operating status data is continuously acquired. Throughout each process, from start-up, initial cutting, intermittent idling to completion, all operating parameters are collected in real-time, including spindle speed, actual load power, load rate, operating current, vibration amplitude, start / stop status, and abnormal warning status. The dynamic changes in spindle operation under different machining materials, cutting depths, and machining periods are continuously recorded, forming a continuous, complete, and uninterrupted sequence of raw spindle operating status data. Relying on the machine tool servo system's acquisition channels and IoT acquisition terminals, real-time feed parameters such as actual feed speed, instantaneous feed rate, feed acceleration, and actual feed displacement versus theoretical feed deviation for each of the X, Y, and Z axes are captured. The entire process of tool-to-workpiece feed motion is recorded, preserving the true feed parameters at every machining moment.

[0038] The Internet of Things (IoT) system continuously monitors the process trigger signals and processing status jump signals of each machine tool, capturing the planned start time, actual start time, processing duration, pause time, resumption time, end time, and abnormal interruption status of each process. It fully records the entire process execution status changes of each production process from start to completion, forming process execution status time sequence data that can truly reflect the actual flow of processes in the workshop. Each tool put into processing in the workshop is equipped with a unique identification RFID electronic tag in advance. The tag is embedded with basic information such as tool model, tool material, rated total life, tool cutting edge parameters, and historical processing records. During the entire life cycle of tool warehousing, warehousing, machine clamping, process processing, disassembly and re-inspection, and scrapping and replacement, the RFID reading terminals deployed at each workstation continuously sense and read the tool tag information. Combined with the machine tool processing time, the tool's consumed life is accumulated in real time, dynamically updated and collected, and the current remaining usable life data of the tool is collected. This achieves accurate collection of the remaining life data of a single tool and a single process, and completely retains the entire process data of tool wear and decay.

[0039] Through the upper-level business data interface of the workshop IoT, it connects with the workshop production execution system and order management system to fully collect the order delivery date data of all in-process production orders in the workshop. Specifically, this includes all delivery date-related parameters such as the planned production time of the order, the required completion time of the order, the planned processing nodes of each process, the order delivery deadline, the order delay penalty rules, and the order priority constraints, forming complete order production constraint data. After completing the real-time collection of the above-mentioned machine tool spindle running status data, machine tool feed speed data, process execution status data, RFID tool remaining life data, and order delivery date data, all production perception data and business order data from different sources, with different structures, different sampling frequencies, and different data dimensions are aggregated and collected. Since the various types of data come from the equipment bottom control layer, the field perception terminal layer, and the upper-level business management layer, the data heterogeneity is significant, and finally, multi-source heterogeneous data is obtained through integration.

[0040] In a preferred embodiment of the present invention, step 2 above may include:

[0041] In this embodiment of the invention, step 220 involves using the actual start time of the process as a unified time reference, performing sliding window interpolation and sampling rate resampling on the time series labels in the multi-source heterogeneous data to obtain a time-aligned data sequence. Specifically, this includes using the actual start time of a single process's independent processing task as the unique unified time reference across the entire domain. Specifically, the absolute time start node corresponding to the process is locked through a joint determination method using state signals. The specific implementation process is as follows: For each processing task to be processed in the workshop, three types of hard trigger signals are continuously monitored: the CNC system cutting enable signal, the spindle servo start signal, and the coordinate axis feed motion valid signal. The trigger signal is applied if and only if... When the CNC system issues a machining instruction, the spindle switches from standby zero speed to working speed, and the coordinate axes generate effective feed displacement, all three signals simultaneously change from an invalid state to an effective state. This indicates that the current process has entered the real physical cutting start state. The physical moment of the synchronous change of these multiple signals is captured as the absolute time start node. This absolute time start node is fixed and recorded using a unified clock timestamp across the entire equipment network. This fixed high-precision absolute timestamp is defined as the time zero point of all associated data collected for this process, so that the spindle running data, feed rate data, process status data, tool life data, and order delivery date data corresponding to the same process are all mapped to the same time origin.

[0042] Based on the absolute time zero point of a unified process, sliding window interpolation is performed on all time-series data carrying time-series tags in multi-source heterogeneous data. The duration parameter of the time-series sliding window is quantified and fixed. The fixed time-series sliding window duration is determined by coupling calculation based on the minimum machining time threshold of the CNC workshop and the highest sampling frequency period of the equipment. First, the shortest effective cutting machining time of a single process in all processes in the workshop is counted. At the same time, the minimum sampling period corresponding to the highest sampling frequency of the workshop's IoT acquisition system is obtained. One-tenth of the shortest process machining time is used as the basic window duration constraint. At the same time, the window duration is ensured to be an integer multiple of the minimum sampling period to avoid data omission and duplicate calculation caused by the mismatch between the window duration and the sampling period. Finally, the time-series sliding window parameter with a fixed duration is determined. Once the window duration is set, it remains constant during the processing of a single batch of processes and does not change dynamically with the process type or equipment model.

[0043] After the window parameters are fixed, a fixed-duration sliding window is slid frame by frame along the global continuous time axis with a fixed, equal-length step size. The sliding step size is set to half the duration of the sliding window, achieving semi-overlapping coverage of adjacent window time intervals. This ensures continuous transition of time series data, without interval omissions or abrupt segmentation. Each time the window slides to a new time interval, all original sampled time points within the current window are traversed and verified point by point. Each time node is checked for invalid data states such as missing data, sampling anomalies, numerical jumps, empty sampling points, and over-range anomalies. All abnormal and missing time points are marked. For time nodes determined to be missing or abnormal, the original abnormal and invalid values ​​are removed, and linear interpolation is used for numerical completion. This ensures the continuity, smoothness, and physical accuracy of the entire time series parameter change curve, eliminating breakpoints, jumps, and null value issues. The linear interpolation calculation formula is as follows: ,in This represents the production parameter values ​​corresponding to the unknown time-series nodes that need to be completed; This represents the target time-series node where data needs to be completed. This represents the nearest valid sampling time before the target time sequence node; This represents the nearest valid sampling time after the target time sequence node; Representative moment The corresponding valid sampling parameter values; Representative moment The corresponding valid sampling parameter values ​​are calculated and filled in point by point for all missing data in all windows using the above formula. All abnormal jump points are smoothed and corrected to eliminate time-series breakpoints, blank sampling intervals, single-point abnormal missing data and numerical mutation problems in the original data, resulting in a preliminary regularized data sequence.

[0044] After completing the sliding window interpolation and obtaining the continuous time-series data sequence, the continuous time-series data sequence is further processed by global sampling rate resampling. The original multi-source heterogeneous data has significant differences in sampling frequency. Among them, the machine tool spindle running status data and feed speed data are high-frequency dense sampling at the millisecond level, and tens to hundreds of sampling points can be output per second; the RFID tool remaining life data is low-to-medium frequency sampling at the second level, and sampling is only triggered when the tool status changes or the process starts or stops; the order delivery date data is static time-period level data, with fixed values ​​and no dynamic sampling nodes throughout the entire process cycle. The sampling density and sampling triggering mechanism of the three types of data are different, and it is impossible to achieve a one-to-one correspondence between time-series nodes. Therefore, it is necessary to perform global unified resampling. A fixed-frequency equal resampling algorithm is used to achieve global time-series density unification. The specific execution logic is as follows: the absolute start time of the process is used as the starting endpoint and the absolute completion time of the process is used as the ending endpoint to construct a complete time-series time axis of the process, and the workshop is set. A unified standard sampling frequency is used as the target resampling frequency. This target frequency is between the highest sampling frequency of the equipment and the low-frequency sampling frequency of the tool, balancing data accuracy and data regularization efficiency. For the original high-frequency sampled spindle data and feed rate data, an equal-interval frame-sampling downsampling strategy is adopted. According to the time interval of the target sampling frequency, effective time-series sampling points are uniformly extracted, discarding redundant and repetitive high-frequency sampling points and retaining core data points with significant time-series change characteristics, thus avoiding the accumulation of redundant high-frequency data. For the original low-frequency sampled tool remaining life data and event-type process status data, a time-series subdivision upsampling strategy is adopted. Between the original sparse effective sampling points, time-series nodes are uniformly added according to the unified target sampling interval, and a smooth transition assignment is completed by combining the effective values ​​before and after, ensuring that each newly added time-series node has a valid parameter value. For static order delivery date data, equal-value filling and extension are performed in the entire time-series interval so that the order constraint parameters cover the entire processing time-series interval of the process.

[0045] Through the resampling processing logic of downsampling, upsampling, and equivalent extension, the spindle operation data, feed rate data, process execution status data, tool remaining life data, and order delivery date data associated with the same process are all standardized to the same fixed sampling frequency and unified time interval. This achieves a strict one-to-one correspondence between the time sampling nodes of all dimensions of data. Each unified time point can simultaneously match and obtain spindle parameters, feed parameters, tool parameters, process status parameters, and order time parameters. After completing the unified locking of the time zero point, sliding window interpolation completion, and full-domain sampling rate resampling, the effective data segment within the complete machining time window is extracted based on the absolute start time and absolute completion time of the process. Redundant standby data of the equipment before the physical start of the process and redundant idling data of the equipment after the physical completion of the process are removed, and finally, a standardized data sequence is output.

[0046] Step 221 involves performing cross-domain feature mapping on the time-aligned data sequence to establish a multi-dimensional topological association between spindle operation fluctuations, tool consumption trajectories, and order delivery nodes. Feature dimensions representing the resource interaction state of a single process are extracted to obtain a process-level association feature matrix. Specifically, this includes: performing domain-specific feature extraction and quantization calculations on the time-aligned standardized data sequence, completing the full-dimensional feature quantization solution for the equipment operation data domain, tool consumption data domain, and order delivery data domain respectively; in the equipment operation data domain, performing time-series fluctuation analysis on the full-time spindle operation data, statistically analyzing the instantaneous fluctuation amplitude of the spindle load, the dynamic offset of the spindle speed, the duration of the spindle operation load, and the frequency of spindle vibration fluctuation at each time-series node, completing the full-dimensional quantization solution for the machine tool processing load stability. The instantaneous fluctuation amplitude of the spindle load is calculated using the deviation between the real-time load and the average load, with the specific calculation formula as follows: In the formula For the first Instantaneous fluctuation amplitude of spindle load corresponding to each timing node; For the first Real-time spindle load values ​​collected by each timing node; This is the arithmetic average of the spindle load within the window during the complete machining of this process; the dynamic offset of the spindle speed is calculated using the deviation between the real-time speed and the process reference speed, and the specific calculation formula is as follows: In the formula For the first Dynamic offset of spindle speed corresponding to each timing node; For the first Real-time spindle speed of each timing node; The standard reference speed is preset for the process document of this process; based on the fluctuation amplitude and speed offset calculated above, combined with the continuous running time of the load at each time node and the frequency of vibration fluctuation, a set of equipment status characteristics that can quantify the real-time processing load stability of the machine tool is summarized, which fully covers the dynamic operation fluctuation law of the spindle throughout the entire process.

[0047] In the tool wear data domain, time-series differential calculations are performed on the continuous data of remaining tool life after time-series alignment. The tool life decay value, real-time tool wear rate, and tool wear fluctuation deviation are statistically analyzed for each machining time period to obtain the continuous consumption trajectory of the tool as the machining time sequence changes. Specifically, the quantitative calculation process involves first calculating the tool life decay value for a single time period, and then solving for the difference in lifespan between consecutive time-series nodes. In the formula For the first Tool life decay value for each machining period; This represents the remaining tool life value at the start of the current time period; The remaining tool life is the value at the end of the current time period; secondly, the real-time tool wear rate is calculated, and the rate is normalized by combining the processing time of the time period. In the formula For the first Real-time tool wear rate during each machining period; For the first The duration of each machining period is calculated; finally, the tool wear fluctuation deviation is calculated to characterize the degree of dispersion of the wear rate fluctuation in different periods. In the formula For the first Tool wear fluctuation deviation during each machining period; This represents the average tool wear rate over the entire process time. By quantifying and calculating each time period, the continuous consumption trajectory of tool life as the machining sequence decreases is obtained, and a complete set of tool dynamic wear characteristics is formed.

[0048] In the order delivery data domain, combining the actual processing time interval of the process with the planned delivery time interval of the order, the time deviation between the current processing progress and the planned progress of the process, the remaining delivery time margin of the order, and the urgency of delivery are calculated for each time node, thus completing the quantification of order constraint characteristics. The formula for calculating the process progress time deviation is as follows: In the formula This refers to the time deviation in the process progress. This represents the current cumulative processing time for the current process. This refers to the planned standard processing time under the corresponding processing schedule; the formula for calculating the remaining delivery margin of an order is as follows: In the formula This represents the remaining delivery time margin for the order. Set a final delivery deadline for the order; This represents the current processing time; the urgency of delivery is calculated using a margin ratio normalization method, as shown in the formula below. In the formula The higher the value, the greater the pressure to deliver; The order plan starts production at the designated time point. The above calculations yield the full-time node order constraint quantification parameters, forming a standardized set of order delivery constraint features.

[0049] This paper constructs a multi-dimensional topological association architecture that couples spindle operation fluctuations, tool consumption trajectories, and order delivery nodes. The architecture's construction logic and implementation process are fully disclosed. This multi-dimensional topological association architecture adopts a node-edge topology design. Topological nodes are defined as three types of entity nodes: machine tool equipment units, tool resource units, and order task units. Topological edges are defined as the resource occupancy and state transmission relationships of a single process within a continuous machining sequence. Each independent process serves as a topological constraint carrier, and the continuous time sequence of process processing is used as the constraint condition for topological association. The propagation law of spindle load fluctuations over time is bound to the machine tool equipment node, and the tool life decay trajectory with machining time is bound to the tool resource node. The time constraints and schedule deviations of order delivery are bound to the order. Task nodes; through topological association mapping, three types of core linkage relationships are established. The first type is the spindle load disturbance propagation relationship, which characterizes how instantaneous overload and load fluctuations of the machine tool spindle continuously affect the cutting state and energy consumption state of the current process, realizing the linkage binding between equipment operation fluctuations and time-series machining states. The second type is the tool wear process relationship, which characterizes the dynamic change coupling relationship of tool wear rate under different spindle load conditions, realizing the dynamic association matching between equipment conditions and tool wear state. The third type is the delivery urgency relationship, which characterizes the constraint impact of process machining progress deviation on the overall order delivery node, realizing the linkage coupling between process machining progress and order delivery constraints, and finally completing the real-time linkage topological association construction of equipment conditions, tool wear, and order constraints.

[0050] Based on the established multidimensional topological association architecture, all feature dimensions that can fully characterize the resource interaction state of a single process are quantitatively extracted. These include multidimensional quantitative indicators such as machine tool load disturbance amplitude features, tool time-series wear increment features, process progress deviation features, order delivery margin deviation features, and equipment tool matching stability features. To eliminate the differences in the dimensions and numerical magnitudes of different feature parameters, a uniform extreme value normalization dimensionless scaling process is performed on all extracted topological association features. After the dimensionless scaling process is completed, principal component analysis is used to compress data to remove multicollinearity interference between features. By calculating the individual variance contribution rate and cumulative variance contribution rate of each feature dimension, redundant features are screened and core effective features are retained.

[0051] The preset threshold for the cumulative contribution rate of the core principal components is comprehensively calibrated based on the data distribution characteristics of multi-dimensional working conditions in CNC flexible machining workshops, the standards for retaining effective information of industrial features, and the specifications for intelligent manufacturing feature dimensionality reduction engineering. It has clear engineering implementation basis and is not a randomly set parameter. The numerical range of this preset threshold is set to 85%–95%. The calibration basis for this range is that a cumulative contribution rate below 85% will result in the loss of a large amount of effective working condition feature information, leading to insufficient data representation accuracy for subsequent cost accounting and scheduling optimization; a cumulative contribution rate above 95% will excessively retain slightly redundant noise features, failing to achieve feature dimensionality reduction and deduplication effects, and significantly increasing the computational cost of subsequent matrix operations. For flexible machining scenarios involving precision small parts, multiple processes, and multiple disturbances, a high threshold range of 90%–95% is selected to retain subtle differences in working conditions to the greatest extent possible. For conventional batch CNC machining scenarios, a standard threshold range of 85%–90% is selected to balance feature effectiveness and data conciseness. By selecting core principal components with a cumulative contribution rate not lower than the above preset thresholds, effective features with core representation capabilities are retained, while invalid features with weak correlation, high overlap, and redundant information are eliminated. A two-dimensional structured data matrix is ​​constructed, with rows corresponding to processes and columns corresponding to resource status features, using the unique process code as the matrix row index and the various standardized topological association core features corresponding to machine tool and cutting tool combinations as the matrix column index.

[0052] Step 222: Based on the process-level association feature matrix, feature vectorization is performed according to the matching combination relationship of machine tool-tool-order to extract continuous numerical representations reflecting the comprehensive state of the process processing environment, resulting in a process-level fusion feature set. Specifically, this includes: performing a full traversal of all in-process processes in the workshop to complete the binding and locking of single-process resource matching relationships; for each independent process in the processing state in the workshop, through the association ledger of the production execution system and equipment acquisition terminal, matching and locking the unique processing machine tool equipment code, tool assembly combination code, and associated production order code corresponding to the current process, establishing a unique, one-to-one correspondence between process-machine tool-tool-order. A single four-element binding ledger solidifies the production resource matching and combination relationship specific to each process. Based on the solidified machine tool-tool-order ternary matching and combination relationship, the row and column positioning, full feature screening, and refined feature value extraction of the process-level association feature matrix are completed. The specific execution process and extraction dimensions are as follows: The corresponding row data in the association feature matrix is ​​located based on the unique process code; the corresponding column data in the matrix is ​​located based on the joint index of machine tool number, tool combination number, and order number; all row and column cross feature data corresponding to the current process resource matching combination are extracted; and the specific values ​​of the five core working condition quantitative features are fully extracted. The detailed extraction rules and physical contents of the five core features are detailed below. The definitions are as follows: First, the spindle load fluctuation characteristic values, which fully extract all standardized quantitative values ​​of the instantaneous load fluctuation amplitude, load fluctuation variance, dynamic speed offset, and continuous load fluctuation range of the machine tool matched to the current process under the entire machining sequence, reflecting the load stability and fluctuation disturbance degree of the machine tool spindle during the cutting process of this process; Second, the tool wear rate characteristic values, which fully extract all standardized values ​​of the tool's time-period life decay, real-time wear rate, and wear fluctuation deviation corresponding to the tool matched to the current process, characterizing the real-time wear decay state of the tool under the machining conditions of this machine tool and this process; Third, the process progress deviation characteristic values, which fully extract the current process... The continuous quantitative values ​​of the deviation between the actual processing time and the planned time of all time-series nodes, the deviation of the process progress completion, and the time lag offset reflect the degree of deviation between the actual process flow progress and the standard planned progress; fourth, the order delivery date constraint characteristic values, which fully extract the standardized quantitative values ​​corresponding to the remaining delivery date margin, delivery date urgency, and progress lag penalty coefficient of the order to which the current process belongs, bearing the time constraints and delivery pressure characteristics of order delivery; fifth, the equipment operation stability characteristic values, which fully extract the continuous quantitative values ​​of the machine tool spindle vibration frequency, feed operation smoothness, and abnormal fluctuation coefficient of equipment operating conditions, reflecting the operational stability status of the equipment throughout the entire process of process processing.

[0053] Through the above row and column positioning and directional filtering extraction, each process can obtain a set of multi-dimensional discrete original feature values ​​that match its own processing resources, processing conditions, and order constraints. The multi-dimensional discrete feature values ​​obtained through filtering and extraction are then subjected to unified precision standardization processing and fixed-rule ordered vectorization concatenation. Specifically, the calculation and concatenation process involves unified precision standardization correction of the multi-dimensional discrete features. Although the five core features have been dimensionless processed, issues such as inconsistent decimal precision and slight deviations in numerical ranges still exist. Therefore, a unified precision truncation and normalization algorithm is used to perform secondary calibration on all discrete feature values, unifying the calculation precision and numerical range of all feature values. The specific calculation formula is as follows: In the formula These are the feature values ​​after precision standardization and correction; To filter and extract the original discrete feature values; To ensure consistent decimal places, all processes use uniformly fixed values, guaranteeing consistent feature precision. Feature dimensions are ordered, with a fixed priority. All corrected discrete features are systematically categorized and arranged according to a fixed hierarchy: equipment status feature group, tool wear feature group, process progress feature group, order constraint feature group, and equipment stability feature group. This prevents vector dimension misalignment caused by disordered feature dimension ordering. A progressive dimension vector concatenation operation maps the ordered single-dimensional scalar features sequentially to a higher-dimensional vector space, completing feature fusion and concatenation. The concatenation operation formula is as follows: In the formula This is a high-dimensional feature vector after fusion of single processes; Quantify the values ​​of various core features after standardization, correction, and orderly sorting; The total number of dimensions for a single process is fixed, and the number of dimensions is uniform across all processes.

[0054] Through the above calculations, each unique machine tool-tool-order matching combination can generate a unique high-dimensional feature vector. This vector integrates all comprehensive information on equipment operating condition fluctuations, tool wear and decay, process progress deviations, order delivery constraints, and equipment operation stability. It can reflect the true comprehensive state of the current process processing environment and does not have problems such as missing features, dimensional disorder, or invalid values. The above screening, correction, and vector splicing operations are performed on all in-process processes in the workshop one by one to obtain the independent high-dimensional feature vectors corresponding to each process. All feature vectors are uniformly collected, dimensionally verified, aligned, and ordered. A small number of erroneous vectors with abnormal dimensions and out-of-bounds values ​​are removed to ensure that the fused feature vectors generated by all processes in the workshop have the same number of dimensions, one-to-one correspondence of feature indicators, and unified calculation methods. All compliant and standardized high-dimensional feature vectors are summarized and collected to form a complete standardized process-level fused feature set.

[0055] Step 223 involves dimensionality reduction, aggregation, and structure reorganization of the process-level fusion feature set, eliminating redundant dimensions and retaining core processing state parameters to form a micro-level process-level production dataset that includes process time sequence fragments and related features. Specifically, this includes: performing principal component analysis on the high-dimensional process-level fusion feature set for dimensionality reduction and aggregation. The specific implementation steps and calculation formulas are as follows: standardizing the high-dimensional feature dataset, assuming the process-level fusion feature set contains... Each process sample contains Dimensional original fusion features, construct the original feature matrix To eliminate the impact of differences in the magnitude of feature values ​​across dimensions on dimensionality reduction accuracy, standardization calculations are performed on all feature values ​​within the matrix. The calculation formula is as follows: In the formula For the first Sample of the process step Standardized values ​​of 3D features; These are the original feature values; For the first The mean of the dimensional feature across all process samples; For the first The standard deviation of each feature dimension in the full process sample is calculated; a feature covariance matrix is ​​constructed, and based on the standardized feature matrix, the covariance between each feature dimension is calculated to characterize the degree of information overlap and linear correlation between dimensions. Covariance matrix The formula for calculating one-dimensional covariance is: In the formula For the first Dimensional features and the first Covariance of dimensional features; , These are the sample means of the standardized features corresponding to the respective dimensions; the eigenvalues ​​and eigenvectors are solved, and the covariance matrix is... Perform orthogonal decomposition to obtain all eigenvalues. Using the corresponding orthogonal eigenvectors, sort all eigenvalues ​​in descending order from largest to smallest. The size of the eigenvalue directly corresponds to the information carrying capacity of a single-dimensional feature. The larger the eigenvalue, the stronger the ability of that dimension to represent the processing state of the process.

[0056] The calculation of single-dimensional variance contribution rate and cumulative contribution rate involves calculating the variance contribution rate of each single feature dimension, representing the proportion of effective information in that dimension. The calculation formula is as follows: In the formula For the first The contribution rate of one-dimensional variance of each feature dimension; For the first The eigenvalues ​​corresponding to each dimension; based on the contribution rate of each dimension, the values ​​are calculated by sequentially adding the values ​​from the previous dimension. The cumulative variance contribution rate of the dimensional feature is calculated using the following formula: In the formula For the front The cumulative variance contribution rate of the core features is calculated. Core principal component dimensions are selected based on a preset threshold. Considering the signal-to-noise ratio characteristics of CNC workshop production data, cost accounting accuracy requirements, and scheduling algorithm efficiency, the basis and fixed range for the preset threshold of the cumulative contribution rate are determined. The threshold is chosen to maximize the retention of effective production information such as equipment operating conditions, tool wear, and order constraints, while fully filtering out minor noise and redundant interference, balancing data integrity and model computation efficiency. The preset cumulative variance contribution rate threshold is 95%, with a reasonable range of 92%-98%. The contribution rates of the sorted feature dimensions are then summed sequentially. When the cumulative variance contribution rate... When the time comes, stop dimension filtering and add the currently accumulated top... The high-dimensional features are determined as the core principal component feature dimensions, while the secondary dimensions with extremely low contribution rates, weak representational capabilities, and mainly consisting of sampling noise and redundant information are discarded. Finally, the high-dimensional features are reduced and aggregated through principal component projection transformation, and the resulting low-dimensional core feature dataset is output.

[0057] For the feature dimensions after principal component analysis, a secondary screening is conducted based on the production business logic and cost accounting needs of the CNC workshop. Invalid and redundant dimensions are eliminated, specifically including duplicate feature dimensions caused by overlapping multi-dimensional information, invalid zero-value dimensions with no fluctuation within the entire process sample, dimensions causing numerical mutations due to sampling anomalies, and meaningless dimensions unrelated to process energy consumption, tool wear, equipment failure, and order delivery dates. Simultaneously, core machining state parameters with direct production and economic accounting value are selectively retained. These include the mean and variance of spindle load fluctuations, real-time tool wear rate and wear increment parameters, actual machining sequence and time deviation parameters, remaining order delivery margin and urgency parameters, and equipment operating condition stability coefficient and vibration fluctuation parameters. All retained parameters can directly participate in subsequent dynamic energy consumption calculations. The quantitative calculation of costs such as tool wear and tear and failure risk costs ensures the practicality and effectiveness of the data. Taking the continuous processing time sequence of a single process as the smallest basic data unit, the core feature data after dimensionality reduction and filtering is comprehensively structured, bound, and encapsulated. The unique code identifier of the process, the actual processing start and end time sequence interval of the process, the three-element matching number of machine tool-tool-order, and the multi-dimensional core correlation feature parameters after dimensionality reduction and purification are bound one-to-one. The original discrete, vector-like, and high-dimensional scattered data without fixed fields are transformed into standardized structured data units with time order, independent processes, regular fields, and corresponding parameters. The structured encapsulation of all processes in the workshop is completed one by one. All process structured data units are collected, sorted, deduplicated, and verified to form a standardized micro-process-level production dataset.

[0058] By unifying the process timing benchmark, multi-source data alignment is achieved, eliminating asynchronous deviations in data acquisition from heterogeneous equipment and various types of sensing terminals. By constructing a multi-dimensional topological association between machine tools, cutting tools, and orders, deep coupling and association of three types of heterogeneous information—equipment operation fluctuations, cutting tool dynamic decay, and order delivery constraints—is realized.

[0059] In a preferred embodiment of the present invention, step 3 above may include:

[0060] In this embodiment of the invention, step 330 involves extracting the spindle load fluctuation variance sequence, speed decay slope sequence, and tool wear increment sequence of a specific process on a specific machine tool from the micro-process-level production dataset, and performing multi-dimensional feature splicing and quantification calculations to obtain a process-level machining state feature group. Specifically, this includes performing targeted filtering of working condition data and time interval locking operations. For the specific process to be calculated, using the unique process code and unique matching number of the machine tool equipment fixed in the micro-process-level production dataset, the entire continuous time sequence data segment of the process on the dedicated specific machine tool from the physical cutting start time to the physical cutting completion time is located and completely extracted. Invalid data during process standby periods, equipment idling periods, process pause periods, and workpiece changeover periods are eliminated, while other processes are thoroughly screened out. Irrelevant feature data for other machine tools and cutting tools are used to ensure that all raw data involved in the calculation are real and valid production data under the current process, machine tool, and batch processing conditions. After locking the valid processing time sequence data segment, the spindle load fluctuation variance sequence is calculated and constructed for each time sequence period. Using the fixed sampling interval uniformly normalized in step 220 as the standard time sequence unit, the complete processing time window of the process is evenly divided into several continuous and non-overlapping time sequence sub-intervals of equal duration. Each time sequence sub-interval contains a fixed number of sampling points. For each time sequence sub-interval, the real-time spindle load values ​​of all sampling points in the interval are captured. The degree of fluctuation dispersion of the spindle load in this time period is quantified by variance calculation to characterize the stability state of the machine tool cutting load. The formula for calculating the spindle load fluctuation variance in a single time period is as follows: ,in This represents the variance of the spindle load fluctuation corresponding to the current time series sub-interval; This represents the total number of valid sampling points contained within the current time series sub-interval; Represents the first time series sub-interval within the current time series. Real-time spindle load values ​​obtained from sampling points; This represents the arithmetic mean of the spindle load values ​​of all sampling points within the current time series sub-interval.

[0061] The entire machining process is traversed sequentially through all time-series sub-intervals. The load fluctuation variance value for each time period is calculated segment by segment. All variance values ​​are arranged in an orderly manner according to the time sequence, forming a continuous, uninterrupted spindle load fluctuation variance sequence that is synchronized with the machining process and covers the entire machining cycle. This sequence fully reflects the dynamic fluctuation and change law of the machine tool spindle load throughout the machining process. Based on the same set of locked, valid time-series data segments, a speed decay slope sequence is calculated and constructed for each time-series interval. Using two consecutive adjacent standard time-series sampling nodes as the smallest calculation unit, the stable spindle speed value of the previous time-series node and the real-time spindle speed value of the next time-series node are captured. Combined with a fixed time-series interval, the dynamic change slope of the spindle speed is calculated, characterizing the speed decay trend and magnitude of the spindle during continuous cutting. The formula for calculating the spindle speed decay slope in a single interval is as follows: ,in This represents the spindle speed decay slope corresponding to the current adjacent time interval; Represents the current time sequence node The corresponding real-time machining speed of the spindle; Represents the next adjacent time sequence node The corresponding real-time machining speed of the spindle; It represents a fixed time interval between two adjacent time nodes, which remains consistent throughout the entire process.

[0062] The process continuously traverses all adjacent time intervals within the entire machining cycle, calculating the speed decay slope value for each interval. All slope values ​​are then arranged sequentially according to their time order, forming a continuous and complete speed decay slope sequence. This allows the capture of subtle dynamic changes such as speed decay, performance degradation, and operating condition deviations during long-term spindle operation under heavy load. Based on the tool life data bound to the process, a tool wear increment sequence is calculated and constructed for each time period. Utilizing the tool remaining life parameters centrally stored in the micro-process-level production data, and using standard time intervals as statistical units, the remaining tool life values ​​at the start and end of each time period are captured. The difference between these values ​​is used to obtain the actual tool wear decay for a single time period, reflecting the real-time wear level of the tool under the current machine tool load conditions. The formula for calculating the tool wear increment for a single time period is as follows: ,in Representing the The tool wear increment corresponding to each time-series machining period; This represents the remaining tool life value corresponding to the start time of the current machining period. This represents the remaining tool life value corresponding to the end of the current processing period. The wear increment is calculated for each time period of the entire process cycle, and the increment values ​​of all time periods are summarized in an orderly manner to form a continuous time-series tool wear increment sequence. Since the aforementioned three types of feature sequences are all calculated based on the same time-series benchmark, the same sampling frequency, and the same process processing window, they have consistent time-series dimensions and data lengths, and there are no time-series misalignments or dimensional differences. According to the fixed hierarchical order of equipment load fluctuation characteristics, spindle speed decay characteristics, and tool wear loss characteristics, the three types of one-dimensional time-series sequences are structured and dimensionally spliced ​​to form an integrated multi-dimensional time-series feature set. At the same time, all feature values ​​after splicing are uniformly quantified and calibrated to unify the accuracy of the entire field value retention, unify the feature representation caliber, and eliminate dimensional magnitude differences, ultimately generating a process-level processing status feature group.

[0063] Step 331: Based on the process-level machining state feature group, and combined with the preset machine tool energy efficiency benchmark curve, perform energy consumption integral mapping to obtain the energy consumption conversion value, and perform life depreciation calculation through the preset tool life depreciation accounting rules to obtain the tool wear conversion value. Combine the energy consumption conversion value and the tool wear conversion value to obtain the basic direct cost data. Specifically, this includes: the preset machine tool energy efficiency benchmark curve is a dedicated spindle load-instantaneous power nonlinear correspondence curve generated mathematically after machine-by-machine, categorized, full-condition, and standardized experimental calibration of all models, all power, and all structural forms of heterogeneous CNC machine tools in the workshop. It avoids the coarse setting of using a unified energy efficiency curve for all equipment in the factory, and adapts to the differences in equipment body, servo drive, mechanical transmission loss, and no-load energy consumption of different machine tools. The specific calibration implementation process is to conduct a full-coverage gradient load energy consumption calibration experiment for each independent CNC machine tool in the workshop under the standard experimental conditions of completed equipment factory parameter verification, qualified equipment maintenance status, no mechanical faults, no servo abnormalities, and stable circuit power supply. The system rigorously covers all actual production operating conditions across six major categories: idle / standby, low-speed light-load cutting, standard load cutting, half-load enhanced cutting, full-load extreme cutting, and short-term overload cutting. Under each condition, the CNC system reads the spindle load rate parameter in real time, while a high-precision power acquisition terminal synchronously collects the instantaneous power consumption data of the entire machine. Load parameters and power parameters are aligned and bound according to a unified timing benchmark, resulting in a massive dataset of raw operating condition-energy consumption samples for each machine tool. According to the data, after collecting full-condition sample data, the original sample data is cleaned and screened to remove abnormal points such as power sudden change at the moment of equipment start-up and shutdown, abnormal points of power supply fluctuation interference, invalid sample points of no-load tool switching, and sample points of unstable transition during working condition switching. Only the effective load-power paired samples under steady-state continuous cutting conditions are retained. Based on the cleaned high-precision effective sample set, the nonlinear least squares fitting method is used to continuously fit the correlation between the spindle load of a single machine tool and the power consumption of the whole machine, and finally generate a dedicated machine tool energy efficiency benchmark curve.

[0064] This energy efficiency benchmark curve strictly uses the real-time load value of the machine tool spindle as the sole independent variable and the real-time power consumption of the entire machine per unit time as the sole dependent variable. It can depict the energy consumption variation pattern of a single machine tool across the entire load range from no-load to full-load, realistically restoring the differentiated energy consumption characteristics of different equipment under different load fluctuation states. Simultaneously, the dedicated energy efficiency benchmark curve for each machine tool is permanently stored as a set of function parameters, establishing a one-to-one correspondence between machine tool equipment codes and energy efficiency curve parameters. During actual process accounting, the corresponding dedicated energy efficiency benchmark curve is automatically retrieved based on the machine tool number bound to the current process. Based on the full-time spindle load fluctuation data in the process-level machining state feature group, a continuous time-domain integral mapping method is used to accumulate and calculate the real-time power consumption within the complete machining cycle of the process, obtaining the converted value of the total power consumption for a single process. The integral calculation formula for process power consumption is as follows: ,in This represents the converted value of electrical energy consumption during the complete processing cycle of the current process. The starting moment of the physical cutting process; The end point of the physical cutting process; Represents continuous time series variables The corresponding real-time load value of the spindle; This represents the result obtained through real-time mapping of the machine tool energy efficiency benchmark curve. The instantaneous power consumption of the equipment at any given moment; this integral calculation method can capture the energy consumption changes caused by load fluctuations at each moment, and can distinguish the energy consumption differences of different working conditions such as steady-state cutting, load fluctuation cutting, and slight overload cutting, so as to obtain a dynamic power consumption conversion value that fits the actual working conditions.

[0065] The preset tool life depreciation calculation rules are executed to dynamically calculate the tool wear depreciation value. These rules are based on real-time wear and can dynamically match depreciation costs according to the actual wear increment in the process, conforming to the true wear pattern of non-uniform tool wear. The calculation rules use the tool's rated total life, tool purchase cost, and the total real-time cumulative wear in the process as the core calculation basis. The specific tool wear depreciation calculation formula is as follows: ,in This represents the converted tool wear value corresponding to the current process. The total tool wear increment across all time-sequence sub-intervals during the entire processing time of a representative process is the sum of all tool wear increments, i.e., the total tool wear caused by this process. This represents the rated total service life of the cutting tool as specified by the manufacturer. This represents the unit cost price of a single new tool purchase. This calculation method can quantify the differences in tool wear caused by real-time cutting conditions in each process. Processes with large load fluctuations and fast wear rates correspond to higher tool depreciation costs, while processes with stable loads and slight wear correspond to lower tool depreciation costs. The calculated dynamic energy consumption conversion value and dynamic tool wear conversion value are bound one by one and summarized in a structured manner. Taking a single process as an independent accounting unit, a two-dimensional basic direct cost dataset is constructed, which includes the direct energy consumption cost of the process and the direct tool wear cost of the process.

[0066] Step 332: Input the basic direct cost data into the preset equipment health assessment rules, calculate the probability of spindle overload warning and the probability of tool failure, and perform risk value conversion based on the preset downtime loss rate per unit time to obtain detailed cost data for all elements, including failure risk costs. Specifically, the equipment health assessment rules are not a single fixed threshold judgment standard, but a hierarchical quantitative assessment rule system built on long-term operating big data of all types of CNC machine tools in the workshop, historical abnormal working condition records of equipment, spindle overload failure samples, tool chipping failure samples, and equipment downtime maintenance ledgers. The whole system includes a complete four-layer architecture: data statistical calibration, limit threshold solution, abnormal working condition cumulative weighting, and risk quantification mapping. The core limit threshold is calibrated in batches. For each CNC machine tool and matching tool in the workshop, normal steady-state machining data and critical machining data throughout the equipment's life cycle are collected. Data and failure data are analyzed using normal distribution statistics and the 3σ anomaly detection rule to eliminate experimental interference and random fluctuations, thereby obtaining the equipment's safe operating boundary parameters. These parameters include the spindle long-term operating load fluctuation limit threshold, the spindle speed decay limit threshold, and the tool wear rate limit threshold. The spindle load fluctuation limit threshold is the maximum safe upper limit of the load fluctuation variance under long-term steady-state machining conditions. If the real-time variance exceeds this value, it indicates severe spindle load disturbance and instability in the cutting condition. The spindle speed decay limit threshold is the maximum allowable speed decay slope for continuous spindle cutting. Exceeding this threshold indicates a decline in spindle transmission performance and severe overload. The tool wear rate limit threshold is the upper limit of the safe wear rate under the matching of tool material and process parameters. Exceeding this threshold indicates that the tool is in a state of high-intensity abnormal wear, which is highly prone to sudden failure.

[0067] The rules for accumulating abnormal operating conditions are designed to address the fact that momentary, minor fluctuations in operating conditions do not directly cause equipment failure, but rather that continuously accumulated over-limit operating conditions are the core cause of failure. This rule incorporates a time-series abnormality accumulation judgment mechanism. It accumulates statistics on the time intervals within the entire machining process where three parameters—load fluctuation variance, speed decay slope, and tool wear rate—exceed safety thresholds. Differentiated weighting coefficients are assigned based on the duration of the abnormality and the magnitude of the over-limit, achieving low-weight accumulation for short-term, minor abnormalities and high-weight accumulation for long-term, significant abnormalities. The rule system is layered and solidified. The final equipment health assessment rule system consists of three levels of standards: spindle overload quantification judgment standard, tool sudden failure judgment standard, and operating condition abnormality accumulation weight standard. It can adaptively match corresponding thresholds and weighting coefficients based on the cutting parameters of different processes, the equipment performance of different machine tools, and the material characteristics of different tools.

[0068] Based on the solidified equipment health assessment rules and real-time processing status characteristics, the probability of spindle overload warning is calculated by coupling the calculation. It integrates two core abnormal features: the degree of spindle load fluctuation dispersion and the degree of speed decay. The comprehensive overload risk probability is quantified through a weighted coupling method based on the over-limit ratio. The specific calculation formula is as follows: ,in The value is the spindle overload warning probability, with a fixed range of 0 to 1. The closer the value is to 1, the higher the risk of the spindle becoming unstable due to overload. , These are the load fluctuation risk weighting coefficient and the speed decay risk weighting coefficient, respectively, and the weights satisfy the normalization constraint. + =1, a fixed value is determined based on the contribution rate of historical equipment failure samples; This represents the variance of the spindle load fluctuation corresponding to the current time series sub-interval; The pre-calibrated spindle load fluctuation limit safety variance threshold in the equipment health assessment rules; This is the absolute value of the spindle speed decay slope in the current process. The maximum permissible spindle speed decay slope threshold is pre-defined in the equipment health assessment rules. Based on the same equipment health assessment rule system, the probability of sudden tool failure is quantitatively calculated. The average wear rate is obtained by solving the tool wear increment sequence throughout the entire process. Combined with the tool's safe wear limit threshold, the risk level of sudden tool chipping, breakage, and premature failure is quantified by the rate exceeding the limit. The specific calculation formula is as follows: ,in The value range is fixed between 0 and 1, representing the probability of sudden tool failure. The closer the value is to 1, the higher the risk of sudden tool failure. The arithmetic mean of the tool wear rate over the entire machining period of the current process is obtained by weighting the tool wear increment sequence obtained in step 330 with the machining time of each period. The maximum permissible safe wear rate limit threshold for cutting tools, pre-calibrated in the equipment health assessment rules.

[0069] The downtime loss rate is used to calculate the risk value of failure. This rate is a standardized parameter derived from a weighted statistical average of five core cost data categories: average equipment maintenance cost, revenue per unit time, fixed labor costs, order delay compensation costs, and material stagnation loss costs. It quantifies the comprehensive economic loss caused by equipment downtime per unit of time. Combining industry standards for small and medium-sized precision CNC machining workshops with actual operational data, the rate is fixed at 80-300 RMB / hour. For high-precision imported machine tools, the rate is 200-300 RMB / hour, and for ordinary domestic CNC machine tools, it is 80-200 RMB / hour. Different machine models correspond to fixed rate parameters, which are consistently applied and do not change randomly. Based on the obtained spindle overload warning probability, tool failure probability, and standardized downtime loss rate, and combined with the actual effective processing time of the process, the expected failure risk cost of the process is calculated. The specific calculation formula is as follows: ,in This represents the expected failure risk cost corresponding to the current process. The actual effective cutting time of the process is expressed in hours. The preset downtime loss rate is matched to the machine tool model. , These are the normalized spindle overload probability and tool failure probability, respectively. The process failure risk cost, the converted value of power consumption, and the converted value of tool wear are integrated and structured across the entire domain to form detailed cost data for all elements.

[0070] Step 333 involves assigning cost weights and normalizing the detailed cost data of all factors to form a quantitative value representing the comprehensive economic cost of a single process under a specific matching state, thus obtaining micro-dynamic processing cost data. Specifically, this includes: combining the real cost composition characteristics, cost fluctuation sensitivity, and historical accounting data statistical patterns of a multi-variety, small-batch CNC flexible production workshop, fixing the weights of three dimensions: electricity consumption cost, tool wear cost, and failure risk cost. All weight coefficients are fitted and calibrated using massive production sample data from the workshop. Among these, tool wear cost is a high-frequency variable core cost in the workshop. The core cost is most significantly affected by fluctuations in machine tool load, changes in cutting conditions, and tool wear, exhibiting the highest sensitivity to cost fluctuations and having the largest weight in the overall cost of a single machining operation. Electricity consumption cost is a basic, rigid, variable cost that fluctuates slightly with machining time and load, with a moderate impact from operating condition disturbances, and thus has the second-highest weight. Failure risk cost is a probabilistic, implicit loss cost, calculated based on the probability of equipment malfunctions, exhibiting range-bound fluctuations, and has the lowest overall weight. Based on statistical results of extensive CNC production operating condition cost data, specific value ranges are defined for the three types of weight coefficients, and they strictly adhere to normalization constraints. Among them, the weighting coefficient of electricity consumption cost The value range is 0.25-0.35; tool wear cost weighting coefficient. The value range is 0.45-0.55; Failure risk cost weighting coefficient. The value range is 0.10-0.20. Each weight coefficient is set within the corresponding interval based on the workshop production process type, equipment precision level, and tool wear characteristics. Since the original calculated values ​​of power consumption cost, tool wear cost, and failure risk cost have differences in magnitude and dimension, the cost values ​​of different dimensions do not have the conditions for direct weighted fusion. Therefore, it is necessary to perform global extreme value normalization preprocessing on the three types of original cost values ​​to uniformly map the cost values ​​of all dimensions to the standard dimensionless value range of 0 to 1, eliminate dimensional barriers and magnitude deviations, and realize the same-dimensional comparability, superposition, and fusion of multi-dimensional cost data.

[0071] The three types of standardized cost values, after normalization preprocessing, are weighted and fused with differential weighting coefficients determined by preset interval calibration to construct a multi-dimensional cost aggregation architecture. This architecture yields a unique quantitative value that uniformly represents the comprehensive economic cost of a single process under specific machine tool, cutting tool, and order matching conditions. The formula for calculating the comprehensive micro-dynamic processing cost of a process is as follows: ,in This represents the comprehensive cost of the final micro-dynamic processing of the process. The weighting coefficient representing the cost of electricity consumption ranges from 0.25 to 0.35. This represents the normalized cost of electricity consumption. The weighting coefficient representing the cost of tool wear has a value range of 0.45-0.55. This represents the normalized tool wear cost. The weighting coefficient representing the cost of failure risk ranges from 0.10 to 0.20. This represents the normalized value of the failure risk cost.

[0072] Improve the control of production costs in the workshop and optimize the scheduling of operations to reduce the overall operating costs under flexible production of multiple varieties and small batches.

[0073] In a preferred embodiment of the present invention, step 4 above may include:

[0074] In this embodiment of the invention, step 440 involves using micro-dynamic machining cost data as input parameters to construct a multi-dimensional decision matrix that includes process topology constraints and machine tool capacity constraints, forming an optimization solution space. The machine tool allocation sequence, tool matching combination, and feed trajectory parameters are mapped to a chromosome coding structure to obtain an initial scheduling population. Specifically, this includes: completing the collection of basic data and the pre-calibration of constraint parameters. All pre-calibrated parameters are based on workshop equipment nameplate parameters, process specifications, and historical production big data statistics, possessing clear engineering basis. The comprehensive micro-dynamic machining cost data of all processes to be processed within the current production batch are uniformly collected and recorded as single-process benchmark cost parameters, fully covering all cost samples corresponding to different machine tools, different tools, and different feed trajectory parameters for each process. Simultaneously, based on the workshop equipment's factory rated parameters and long-term operation and maintenance production data, the standardized pre-calibration of two types of core constraint parameters is completed. For process topology constraint parameters, based on the part machining process card and process flow specifications, the pre-constraint relationships of all processes are sorted out, a process topology association matrix is ​​constructed, and a 0-1 quantitative constraint mechanism is used to define the legality of process processing, setting topology constraint judgment coefficients. , used to characterize the The first process is relative to the second process. The legality of the processing sequence of a process is determined by the following rules: if the process... For process If the preceding process is not completed, then =0 indicates an invalid scheduling plan; if the process If the preceding processing is completed or there are no preceding constraints, then =1 indicates that the scheduling scheme is valid.

[0075] Based on the machine tool capacity constraints, the maximum effective machining time limit per unit production cycle of a single machine tool is statistically determined, taking into account the rated effective working hours of the CNC machine tool at the factory, the average daily maintenance time of the equipment, and the standard production cycle time of the workshop. Maximum number of processes subscript This is a unique machine tool identification number. This parameter represents the physical limit constraint of the equipment, and is preset based on the machine tool's rated load parameters and workshop production statistics. It is used to limit the overload distribution of a single machine tool due to excessive working hours, excessive processes, and excessive loads. The machine tool capacity load determination formula is defined as follows: ,in For machine tools Total processing time for all assigned processes; For machine tools Maximum permissible processing time per unit cycle; For machine tools The total number of processes currently assigned; For machine tools The maximum allowed number of processing steps per unit cycle is determined. If the dual-formula condition is met, the production capacity is deemed compliant; otherwise, it is considered illegal scheduling exceeding capacity. This matrix is ​​a four-dimensional coupled decision matrix, strictly corresponding to the core elements of production scheduling. The matrix row dimension is uniformly set to all pending processing steps in the current batch, with each row uniquely corresponding to one independent processing step. The matrix column dimension includes four sub-dimensions: machine tool equipment dimension, tool combination dimension, feed trajectory parameter dimension, and overall cost dimension, fully covering all decision variables for scheduling optimization. Any cell in the matrix is ​​recorded as... , representing the Matching the first process Taiwanese machine tool, the first Group of cutting tools, the first The scheduling unit information corresponding to the group feed trajectory parameters; the core assignment content of the matrix unit is the comprehensive micro-dynamic machining cost under the corresponding working condition. Simultaneously, compliance corrections are made by binding topology constraint coefficients and capacity constraint coefficients, and the final effective assignment formula for the cell is as follows: ,in These are the topological constraint determination coefficients; This is the machine tool capacity constraint judgment coefficient; when capacity is compliant... =1, when production capacity exceeds the limit =0; For the first The comprehensive cost of micro-dynamic machining of each process under the matching of corresponding equipment, tools, and trajectory parameters; according to the above assignment formula, all matrix units that violate topological timing constraints or exceed machine tool capacity constraints are assigned a value of 0 and are uniformly marked as invalid and illegal scheduling units; all matrix units with non-zero values ​​are legal and valid scheduling combinations that simultaneously satisfy process logic, equipment capacity, and actual working conditions. The set of all valid scheduling units constitutes the complete optimization solution space.

[0076] A multi-segment real-number encoding structure is adopted, with encoding length and range strictly matching the effective solution space parameter range of the multi-dimensional decision matrix. The three core multi-dimensional decision parameters included in the scheduling scheme are uniformly mapped to complete and continuous chromosomal gene sequences. Each chromosome uniquely corresponds to a complete workshop scheduling scheme. Specific encoding mapping objects strictly correspond to the matrix decision dimensions, including three independent gene segments: machine tool allocation sequence, tool matching combination, and feed trajectory parameters. The number of gene positions in the machine tool allocation sequence gene segment is consistent with the total number of processes to be processed. Each gene value corresponds to the valid machine tool number for process matching in the multi-dimensional decision matrix, realizing the mapping and binding of processes and equipment resources. The tool matching combination gene segment corresponds to the valid tool combination encoding within the matrix according to the process adaptation requirements of each process, realizing the final matching mapping of processes and tool consumable resources. The feed trajectory parameter gene segment is taken from the effective solution space of the matrix. The system employs refined trajectory parameters such as feed rate, feed path range, and cutting feed duration to encode and map micro-machining process parameters. All encoded values ​​are strictly limited to the effective units of the multi-dimensional decision matrix. Within the defined legal optimization solution space, a uniform random sampling mechanism is used to traverse all effective scheduling unit combinations of the multi-dimensional decision matrix, generating batches of independent, parameter-differentiated, and non-repeating chromosome individuals. All individuals in the initial population are adapted to the process topology constraints and machine tool capacity constraints, strictly corresponding to non-zero effective units in the matrix, and there are no illegal codes, invalid parameters, or illegal scheduling combinations. The number of individuals in the population is a preset fixed hyperparameter, based on the scale of the number of workshop processes and the algorithm's optimization accuracy requirements, ensuring that the initial population can fully cover high-quality scheduling combinations within the solution space, possessing sufficient scheme diversity and a global search basis.

[0077] Step 441 involves mapping the individual codes in the initial scheduling population back to the optimization solution space for full-path cost accounting, calculating the reciprocal of the total operating cost corresponding to each scheduling sequence, quantifying the quality of individuals in the solution space, and obtaining a set of individual fitness values. Specifically, this includes: traversing all chromosome individuals in the initial scheduling population, performing segmented decoding and parsing of the three-segment gene structure of each chromosome, reading the gene values ​​of the machine tool allocation sequence gene segments one by one to reconstruct the processing machine tool equipment corresponding to each process; reading the gene values ​​of the tool matching combination gene segments one by one to reconstruct the tool assembly matching scheme corresponding to each process; reading the gene values ​​of the feed trajectory parameter gene segments one by one to reconstruct the complete refined feed processing parameters for each process. After decoding, according to the temporal logic of the process process topology constraints, all processes are sorted and reorganized to completely reconstruct the complete set of process processing sequences, resource matching relationships, and micro-processing parameters corresponding to the current chromosome individual, forming an entity scheduling sequence scheme.

[0078] Based on the processing sequence determined by the topology constraints, the micro-dynamic comprehensive processing cost values ​​corresponding to the matching machine tool, matching cutting tool, and matching trajectory parameters for each process are retrieved. Relying entirely on the dynamic cost data output from previous steps, the total operating cost of the entire scheduling scheme is calculated by accumulating costs across each process. The global total operating cost calculation formula is as follows: ,in This represents the total global operating cost of a single complete scheduling scheme for the workshop. This represents the total number of all processes in the current batch awaiting processing in the workshop. Representing the The micro-dynamic comprehensive machining cost of each process under the current matching state of machine tools, cutting tools, and trajectory parameters; with minimizing the overall total operating cost of the workshop as the core optimization objective, the lower the total operating cost, the higher the resource utilization rate of the scheduling scheme, the lower the production loss, the lower the failure risk, and the better the overall economic benefits. To adapt to the iterative logic of maximizing optimization, a fitness calculation architecture based on the reciprocal of the total cost is constructed, while a minimum constant is introduced to avoid calculation anomalies. The individual fitness calculation formula is as follows: ,in This represents the fitness value of the scheme corresponding to the currently scheduled individual; The minimum value is represented by a zero constant, which is close to 0 but not zero. It is used to prevent calculation errors when the total operating cost is extremely low, thus ensuring the stability of the iterative calculation process. The initial population is decoded, cost is calculated, and fitness is calculated for each individual. The fitness values ​​of all individuals are collected in an orderly manner and arranged according to the order of the individuals in the population to form a complete set of individual fitness values. This quantifies the comprehensive performance of each scheduling scheme in the optimization solution space.

[0079] Step 442: Based on the set of individual fitness values, perform roulette wheel selection and adaptive crossover operations on the initial scheduling population, retaining high-fitness gene fragments, and performing local trajectory parameter perturbation on the crossover individuals to obtain the next generation iterative population. Specifically, this includes: performing a roulette wheel selection operation to complete the population's selection and update through natural selection. The roulette wheel selection mechanism uses the ratio of a single scheduling individual's fitness value to the total fitness value of all individuals in the population as the core selection probability criterion, strictly following the population iterative evolution logic of natural selection. For high-quality scheduling individuals with better adaptability to working conditions, lower overall operating costs, and higher fitness, the probability of them being selected, replicated, and completely retained into the next generation population is higher. For low-quality scheduling individuals with poor adaptability to working conditions, higher overall operating costs, and lower fitness, the probability of them being selected, eliminated, and removed from the population is higher. The formula for calculating the selection probability of a single individual is as follows: ,in The first in the representative population The probability of each individual choosing to retain their choice; Representing the The fitness value of each individual; This represents the sum of the fitness of all individuals within the current population. Through a global probability traversal screening mechanism, high-quality machine tool allocation sequence gene fragments, tool matching combination gene fragments, and feed trajectory parameter gene fragments corresponding to individuals with high fitness are prioritized for retention. At the same time, inferior scheduling gene fragments with defects such as high processing costs, high equipment risks, and unreasonable resource matching are completely eliminated. This completes the first overall performance optimization of the population, making the overall baseline performance of the selected and retained population significantly better than that of the original initial population.

[0080] Adaptive crossover operations are performed to achieve recombination and fusion of high-quality gene fragments from different individuals. An adaptive crossover probability control mechanism is established that can be dynamically adjusted according to the population iteration progress and the degree of dispersion of individual fitness. This adapts to the optimization needs of different iteration stages. In the early stage of iteration, when the individual parameters within the population are significantly different and the effective parameter search space is wide, the system automatically increases the crossover probability to enhance the gene recombination between different individuals, quickly traverse various high-quality scheduling parameter combinations in the solution space, and accelerate the accumulation of high-quality genes. In the middle and late stages of iteration, as the performance of individual individuals in the population gradually becomes more uniform and the structure of high-quality scheduling genes becomes more stable, the system automatically reduces the crossover probability to avoid high-frequency crossover from destroying the already converged and optimized high-quality gene structure, thus preventing population iteration oscillations and performance decline.

[0081] After screening and optimization, individuals in the population are randomly paired. For each paired individual, segmented single-point crossover is performed on the machine tool allocation gene segment, tool matching gene segment, and trajectory parameter gene segment. This recombines and merges the advantageous gene fragments carried by different high-quality individuals, generating new scheduling individuals that simultaneously possess the advantages of multiple groups of individuals, have more reasonable resource matching, lower processing costs, and stronger adaptability to working conditions. This effectively enriches the structural types of high-quality scheduling schemes within the population and enhances the overall optimization potential of the population. For the new individuals generated after gene crossover, strict adherence to process topology and timing constraints and not exceeding the rated capacity of the machine tool are required. Under the premise of quantitative constraints and without altering the established high-quality equipment matching genes and tool matching genes, only small-scale, controllable, and refined random perturbation fine-tuning is carried out on the feed trajectory parameter gene fragment. By making small and reasonable changes to micro-process parameters such as process feed speed, effective cutting range, and processing dwell time, the final matching combination of micro-process parameter dimensions is deeply explored while retaining the macro-high-quality resource scheduling framework. After completing the three-layer progressive evolution process of selection, crossover, and perturbation mutation, the number of individuals in the current population is standardized and unified to strictly ensure that the overall size of the population in each generation remains constant, and finally, a complete next-generation iteration population is generated.

[0082] Step 443: Determine whether the iteration rounds of the next generation of iterative population have reached a preset threshold or whether the fitness variance is lower than the convergence tolerance. If the convergence conditions are met, extract the final fitness individuals for chromosome reverse decoding to restore the matching matrix and trajectory parameter set of the process-equipment-tool, thus obtaining the initial flexible job scheduling scheme. Specifically, this includes setting two types of parallel effective iteration convergence judgment conditions: the maximum iteration round threshold condition and the population fitness variance convergence tolerance condition. Meeting any single condition is sufficient to determine iteration convergence and terminate the loop iteration. All preset parameters are based on the process scale and equipment computing power of the multi-variety, small-batch CNC flexible production workshop. The statistical calibration of the measured patterns of flat and iterative optimization has clear engineering application basis and is not a random fixed value. Among them, the maximum iteration round threshold is a limited iteration upper limit adapted to the scale of workshop scheduling solution and computing efficiency. It is comprehensively calibrated based on the number of routine processing operations in the workshop, the population size, and the calculation time of a single iteration round. For the process scheduling scenario of small and medium-sized flexible CNC machining workshops, the preset value range of the maximum iteration round threshold is 80-150 rounds. The upper limit of the range is selected when the number of processes is large and the resource matching dimension is complex, and the lower limit of the range is selected when the number of processes is small and the scheduling dimension is simple, so as to balance the optimization accuracy and computing efficiency.

[0083] The population fitness variance convergence tolerance is the critical deviation threshold for determining whether the overall scheduling performance of the population tends to stabilize. It is used to quantify the performance dispersion of all individual schedulers within the population. Based on steady-state convergence data from a large number of field iterative experiments, it can determine whether the algorithm has completed global optimization and whether the solution tends to the global final state. The preset range of this convergence tolerance is 0.001-0.005. The smaller the tolerance value, the higher the requirement for consistency in the performance of individual populations and the higher the optimization accuracy. It can be adaptively selected within the range according to the accuracy requirements of workshop cost optimization. The population fitness variance is used to quantify the performance dispersion of the scheduling schemes of all individuals in the current population. The larger the variance value, the greater the cost-performance difference among individual schedulers within the population, the more sufficient the parameter optimization space, and the need for continuous iterative optimization. The smaller the variance value, the more uniform the performance of the individual scheduling schemes in the population, the more stable the parameter structure, and the basic achievement of the final steady state. The formula for calculating the population fitness variance is: ,in Represents the variance of the fitness of the current iteration population; Represents the total number of individuals in the current iteration population; Represents the first in the population The fitness value of each individual; This represents the arithmetic mean of the fitness of all individuals in the current population. After each iteration, the system automatically accumulates the number of iterations completed by the current algorithm, continuously comparing it with the preset maximum iteration threshold. Simultaneously, it calculates the fitness variance of the current population in real time for each iteration, continuously comparing it with the preset convergence tolerance, achieving dual-condition synchronous monitoring and judgment. If the current accumulated iteration count reaches the preset maximum iteration threshold, regardless of the dispersion of the current population's fitness variance, the iteration process is deemed sufficiently complete, and the iteration loop is immediately terminated to avoid redundant computational waste caused by excessive iteration. If the current... If the population fitness variance value continues to decrease and falls below the preset convergence tolerance, it means that the overall cost and performance of all scheduling schemes within the population are basically consistent. The algorithm has fully traversed the solution space and locked the global final parameter range. There is no room for further optimization. Similarly, it is determined that the iteration has converged and the iteration loop is terminated. If neither of the above two convergence conditions is met, it is determined that the current scheduling scheme optimization is not sufficient and the population still has a large optimization space. The next generation of iterative population generated is taken as a brand new initial population, and the process returns to step 442 to continue to execute a new round of iterative evolution operation, continuously iterating and optimizing the population gene structure and scheduling scheme performance.

[0084] After the iteration process meets the convergence condition and terminates the iteration loop, all scheduling individuals within the final convergent population are completely traversed. The individual with the largest fitness value is selected as the global final scheduling individual. This final individual corresponds to the final scheduling combination with the lowest comprehensive economic cost, compliance with production constraints, and strongest adaptability to on-site implementation within the solution space. The selected final individual undergoes complete chromosome encoding reverse decoding analysis. The actual production parameters corresponding to the machine tool allocation gene, tool matching gene, and feed trajectory parameter gene are decomposed segment by segment according to the gene segmentation structure. The final process-equipment-tool resource matching matrix, the standardized processing time sequence of the entire process, and the refined feed trajectory parameter set are restored. After decoding, secondary constraint verification and data normalization are performed on all restored scheduling parameters to verify the compliance of process topology constraints and machine tool capacity constraints. Extremely low probability parameter deviations and boundary abnormal values ​​are eliminated. Finally, a standardized initial flexible operation scheduling scheme that takes into account process compliance, equipment capacity rationality, production economy, and operational stability is generated. This provides data decision support for on-site operation execution, dynamic allocation of production resources, and dynamic rescheduling decisions under working condition disturbances in the workshop for flexible production of multiple varieties and small batches.

[0085] Reduce overall workshop operating costs, decrease the probability of equipment downtime, and improve tool resource utilization and equipment output efficiency.

[0086] In a preferred embodiment of the present invention, step 5 above may include:

[0087] In this embodiment of the invention, step 550 involves real-time acquisition of actual execution feedback messages after the initial flexible job scheduling plan is issued to the physical entities in the workshop. The planned progress nodes are compared with the actual completion status, and time offsets and equipment anomaly identifiers are calculated to obtain the process execution deviation sequence. Specifically, this includes: after the initial flexible job scheduling plan is officially issued to the workshop manufacturing execution terminal and the process processing is started, real-time acquisition of operation feedback messages from all physical entities in the workshop, such as CNC machine tools, tool actuators, and workpiece conveying equipment, is performed. The message data comprehensively includes all dimensions of execution information for each process, such as planned start time, actual start time, planned completion time, actual completion time, real-time equipment operating conditions, process completion verification status, spindle load operating data, and real-time tool wear data, ensuring that the collected data matches the actual processing conditions. The system is synchronized in real time. It uses the fixed process sequence progress nodes in the initial flexible operation scheduling plan as the sole planning benchmark. The preset basis of this plan progress node is the standard working hours of the part processing process specification, the rated processing cycle of the workshop equipment, the order delivery cycle constraints, and the statistical working hours of steady-state processing in batch production. It is a standardized progress benchmark that is comprehensively verified by combining process compliance, equipment capacity limit, and order delivery requirements. It has a clear engineering production basis and is not randomly set. It matches the actual completion status of each process to be executed, in execution, and completed on site one by one, and establishes a one-to-one correspondence between the set of planned progress parameters and the set of actual execution parameters. It covers the status comparison of the time dimension, equipment operation dimension, and processing completion dimension of all processes, and screens for progress deviations and equipment operation abnormalities that occur during on-site processing.

[0088] Time offset is used to reflect the degree of deviation between the actual execution progress of a process and the preset planned progress. The preset allowable deviation range of the planned progress is determined based on the long-term steady-state processing data of flexible production of multiple varieties and small batches in the workshop and the normal fluctuation range allowed by the process technology. It distinguishes between normal fine-tuning fluctuations in production and abnormal progress disturbances. The degree of deviation is calculated by the difference between the actual execution time node and the planned time node of the process. The formula for calculating the time offset of a single process is as follows: ,in Representing the The time offset of each process, in seconds; Representing the The actual completion time of each process on site; Representing the The planned completion time nodes preset in the initial flexible operation scheduling scheme for each process are as follows: when the time offset value is greater than 0, it means that the process is lagging behind the planned progress; when the time offset value is equal to 0, it means that the process progress is in line with the planned standard; when the time offset value is less than 0, it means that the process is ahead of the planned progress.

[0089] Combining the spindle load fluctuation data, speed decay data, and tool wear rate data from the feedback messages, anomaly determination is made by comparing them with the equipment's healthy operating baseline safety thresholds. All safety thresholds are uniformly calibrated based on the CNC machine tool's factory rated parameters, spindle steady-state operating parameters, wear test data of commonly used tool materials in the workshop, and the allowable fluctuation range for precision machining conditions in the industry. They possess standardized engineering basis, and their specific numerical ranges are clearly defined and fixed. Specifically, the spindle load fluctuation safety threshold is 12% of the equipment's rated load; a spindle load fluctuation amplitude exceeding 12% of the rated load during on-site machining is considered an abnormal load. The spindle speed decay safety threshold... The value is 6% of the equipment's rated speed. If the actual spindle speed attenuation exceeds 6% of the rated speed, it is considered an abnormal speed operation. The tool wear rate safety threshold is 1.3 times the wear rate of normal steady-state machining. If the actual tool wear rate exceeds 1.3 times the standard steady-state wear rate, it is considered abnormal tool wear. Based on the above quantified safety thresholds, abnormal equipment states during single-process machining are marked, generating binary quantified equipment abnormality identifiers. If any operating parameter exceeds the corresponding safety threshold, an abnormal working condition is determined and marked as an abnormal state. If all operating parameters are within the safety threshold range, it is marked as a normal state. The time offset and equipment abnormality identifiers corresponding to each process are sequentially organized and sorted according to the process execution sequence to form a sequential, complete, and one-to-one corresponding process execution deviation sequence, fully recording the progress deviation and equipment abnormality deviation information of the workshop machining.

[0090] Step 551: When the time offset or equipment anomaly flag in the process execution deviation sequence exceeds the preset tolerance threshold, a dynamic disturbance event is triggered, and the scheduling status snapshot of the completed processes is frozen. Real-time operating parameters of the remaining processes to be executed are extracted to obtain the real-time status data of the remaining processes. Specifically, the set time offset tolerance threshold and equipment anomaly tolerance threshold are both comprehensively calibrated based on long-term steady-state production big data statistics from the CNC flexible machining workshop, allowable fluctuation standards for precision machining processes, CNC machine tool safety operating condition specifications, and tool wear limit test data. They are suitable for multi-variety, small-batch flexible production scenarios, have standardized engineering application basis, are not randomly set by humans, and can effectively distinguish between normal production fine-tuning fluctuations and malicious abnormal disturbances. Specific threshold parameters, calibration basis, and numerical ranges are detailed below. The following are the key parameters: First, the time offset tolerance threshold. This threshold represents the maximum allowable positive or negative fluctuation in the planned processing progress of a process. It is used to distinguish between normal progress adjustments and abnormal progress delays. The calibration is based on the steady-state working time fluctuation range of batch production in the workshop, the order delivery buffer margin specifications, and the timing constraints of precision process processing. The preset time offset tolerance threshold has a uniform value range of ±300 seconds. That is, the maximum allowable advance of the actual completion time of a process relative to the planned completion time is 300 seconds, and the maximum allowable lag is 300 seconds. When the process time offset is within the range of -300 seconds to +300 seconds, it is determined to be a normal fluctuation in production progress, and no rescheduling is required. When the process time offset is less than -300 seconds or greater than +300 seconds, it is determined to be an excessive progress deviation, indicating an abnormal progress disturbance.

[0091] Second, there is the equipment anomaly tolerance threshold. This threshold represents the maximum safe fluctuation range allowed for the machining conditions of machine tools and tool systems. It is used to distinguish between fluctuations in normal machining conditions and severe disturbances caused by equipment failure or abnormal tool wear. The calibration basis is the rated load parameters of CNC machine tools, the steady-state operating standard of spindle speed, and the conventional wear test data of carbide tools. The equipment anomaly tolerance threshold includes three core quantitative judgment criteria, with the following specific value ranges: the spindle load fluctuation tolerance threshold is 12% of the rated machining load of the equipment; the spindle speed decay tolerance threshold is 6% of the rated speed of the equipment; and the tool abnormal wear tolerance threshold is 1.3 times the wear rate of conventional steady-state machining. If the real-time fluctuation amplitude of any one of the three parameters exceeds the corresponding proportional threshold, the equipment condition is judged to be abnormal and exceeds the standard. The two types of thresholds are independent of each other and judged in parallel, together constituting the complete quantitative judgment benchmark for dynamic disturbance events of this invention. As long as any deviation index of a single or multiple processes in the process execution deviation sequence exceeds the corresponding tolerance threshold, a dynamic disturbance event can be judged to have occurred in on-site production.

[0092] The system iterates through all time-series deviation data in the process execution deviation sequence one by one, and simultaneously performs dual-dimensional threshold verification. It checks whether the time offset of each process exceeds the ±300-second time offset tolerance threshold, and at the same time checks whether the spindle load fluctuation, spindle speed decay, and tool wear rate corresponding to each process exceed the equipment abnormality tolerance threshold standard. If any single process or multiple processes have excessive progress offset or abnormal equipment condition, it directly determines that the current production process in the workshop has triggered a dynamic disturbance event. At this time, the actual production conditions in the workshop have deviated from the steady-state conditions adapted to the initial scheduling plan. The original initial flexible operation scheduling plan can no longer adapt to the real-time production status on site, and the dynamic rescheduling process must be initiated to optimize and adjust subsequent production tasks. If the progress deviation indicators and equipment condition deviation indicators of all processes fall within the corresponding tolerance threshold range, it is determined that the on-site production conditions are stable and there is no abnormal production disturbance. There is no need to trigger the rescheduling mechanism, and the original initial flexible operation scheduling plan continues to be used to execute production and processing.

[0093] At the instant the dynamic disturbance event is determined to be triggered, a snapshot freeze of the scheduling status is immediately performed on all completed and processed processes before the current time point. This freezes and saves all historical processing data for completed processes, including equipment matching relationships, tool matching combinations, machining feed trajectory parameters, actual machining time, micro-dynamic machining costs, process completion time points, real-time equipment operating status, and workpiece machining completion status. After the snapshot freeze is completed, all scheduling statuses and processing data for completed processes remain fixed throughout the process and will no longer participate in any subsequent dataset updates, cost recalculations, or rescheduling optimization iterations. The instant the dynamic disturbance event is triggered is used as the dividing point. The system filters all remaining processes after the current node that have not yet started processing, are in a pending processing state, or have started but not yet completed. It comprehensively extracts all dimensions of operating condition parameters and resource status parameters corresponding to the remaining processes in real time. Specifically, this includes real-time equipment occupancy status, remaining effective tool life, real-time spindle load status, standard process parameters of unexecuted processes, remaining order delivery period constraints, real-time remaining capacity of machine tools, workpiece pending processing status, and on-site resource occupancy, etc. All extracted real-time parameters are then organized, collected, and structured to form real-time status data of the remaining processes, representing the actual operating conditions of the remaining production tasks in the workshop after a disturbance occurs.

[0094] Step 552: Update the micro-process-level production dataset based on the real-time status data of the remaining processes, recalculate the micro-dynamic processing cost data of the remaining processes to be executed, and perform local optimization calculation with the goal of minimizing the sum of the total cost of the remaining processes and the order delay penalty to obtain the rescheduling candidate sequence. Specifically, this includes: based on the real-time status data of the remaining processes, performing targeted iterative updates on the original global micro-process-level production dataset, replacing outdated operating condition parameters, equipment status parameters, tool wear parameters, and capacity constraint parameters of the remaining processes to be executed, retaining the frozen historical data of completed processes, and ensuring that the updated dataset fully adapts to the actual production conditions on site after the disturbance occurs, covering core parameters such as equipment health status, remaining tool life, real-time energy consumption characteristics, and failure risk probability of the remaining processes, and eliminating static parameters that are lagging and failing, so as to ensure that subsequent cost accounting and optimization calculations are consistent with the real-time operating conditions.

[0095] Based on the updated micro-level production dataset, the cost accounting logic of steps 331 to 333 is reused to recalculate the energy consumption, tool wear, and failure risk costs of all remaining unexecuted processes. The comprehensive micro-dynamic processing cost of each remaining process is updated using a normalized weighted aggregation method, quantifying the true economic cost of the remaining processes under perturbation conditions. With the dual optimization objectives of minimizing the total comprehensive cost of the remaining processes and minimizing order delay penalties, a local optimization calculation criterion is constructed, and the overall optimization objective calculation formula is as follows: ,in This represents the overall cost of optimizing the remaining processes. This represents the cumulative total of the updated micro-dynamic processing costs for all remaining pending steps. The cost of order delay penalty is determined based on the delay duration of the process and the order priority. Based on this dual optimization objective, local optimization iterations are carried out on the resource matching combination, processing sequence, and trajectory parameters of the remaining unexecuted processes. Only the scheduling parameters of the unexecuted processes are optimized and adjusted, without changing the status of the frozen completed processes. While effectively reducing the computational load of rescheduling, the system maximizes the adaptation to on-site disturbance conditions. Through multiple rounds of local iterative optimization, multiple sets of feasible scheduling combinations that meet the basic constraints and have different performance characteristics are obtained, and these combinations are regularized to form a complete set of rescheduling candidate sequences.

[0096] Step 553 involves verifying the capacity conflict and the legality of the process route for the candidate rescheduling sequences, eliminating scheduling paths that violate equipment capacity constraints, replacing the corresponding unexecuted nodes in the original plan with sequences that pass verification, obtaining an updated flexible operation scheduling scheme, and issuing it to the workshop manufacturing execution terminal to control the processing flow of physical entities in the workshop; specifically, this includes: the system traversing each scheduling scheme in the candidate sequence set, conducting comprehensive capacity verification on the machine tool equipment allocation relationship, process load quantity, and processing time allocation for the remaining processes in the scheme, strictly benchmarking against the machine tool's preset maximum processing time capacity per unit cycle, maximum number of processes, and real-time remaining capacity parameters of the equipment, and checking one by one whether each candidate sequence has capacity violations such as single machine tool time exceeding the limit, process allocation quantity overload, and overlapping processing time conflicts. The first layer of solution selection and optimization involves eliminating scheduling paths that conflict with equipment capacity or exceed the physical carrying capacity of the equipment, retaining only candidate sequences that meet equipment capacity constraints and have no resource conflicts. Based on capacity compliance, the second layer further follows the machining process specifications and topological constraints of the parts, verifying the processing sequence of the remaining processes in each candidate sequence, the dependencies between preceding processes, and the matching rules of subsequent processes. It focuses on screening for violations such as process reversal, process skipping, executing subsequent processes before completing preceding processes, and disordered process routes. This ensures that the process arrangement of the rescheduled sequence conforms to the inherent process logic of machining, guaranteeing that each remaining candidate sequence has on-site process feasibility, thus completing the second layer of solution selection and optimization. Finally, a valid rescheduled sequence that meets both constraints is obtained.

[0097] The system selects the sequence with the lowest overall cost and the best adaptability from the valid rescheduling sequences that have passed double-layer verification as the final rescheduling scheme. This scheme replaces all unexecuted process scheduling nodes in the initial flexible operation scheduling scheme, fully retains the frozen scheduling data and processing status of completed processes, and achieves partial updates to the scheduling scheme. It generates a new flexible operation scheduling scheme that adapts to the dynamic disturbance conditions on site, and distributes the updated flexible operation scheduling scheme to all manufacturing execution terminals in the workshop in real time. This synchronously updates the processing sequence, resource matching, and trajectory processing parameters of physical entities such as CNC machine tools, tool actuators, and workpiece conveying equipment in the workshop, and adjusts the processing flow rhythm and resource matching relationship of subsequent processes in the workshop. This enables adaptive dynamic scheduling under production disturbance conditions and ensures the continuous and stable operation of the flexible production process in the workshop.

[0098] It adapts to various dynamic disturbances in flexible production scenarios with multiple varieties and small batches, enabling adaptive iterative updates of scheduling schemes and reducing overall production losses and order delay risks in the workshop.

[0099] like Figure 2 As shown, embodiments of the present invention also provide a CNC shop flexible operation planning cost optimization management system, including:

[0100] The data acquisition module is used to collect spindle operating status data, feed speed data, process execution status data, RFID-read tool remaining life data, and order delivery date data of heterogeneous CNC machine tools through the workshop Internet of Things, so as to obtain multi-source heterogeneous data.

[0101] The fusion module is used to perform timestamp alignment and feature fusion on multi-source heterogeneous data to obtain micro-level production datasets.

[0102] The calculation module is used to extract the spindle load timing characteristics and tool wear rate characteristics of a specific process on a specific machine tool based on the micro-process-level production dataset, and calculate the corresponding dynamic energy consumption cost, tool wear cost and failure risk cost, forming micro-dynamic machining cost data with process, machine tool, tool matching and tool feed trajectory parameters as independent variables.

[0103] The optimization module is used to construct an optimization solution space based on micro-dynamic processing cost data, with the objective function of minimizing the overall operating cost and the constraints of the process sequence and equipment capacity. It encodes the matching relationship between different processes, machine tools and cutting tools and the tool feed trajectory parameters as population individuals, calculates the fitness value of each population individual in the optimization solution space, and iteratively updates the population through selection, crossover and mutation operations until the preset convergence condition is met, and obtains the initial flexible operation scheduling scheme.

[0104] The response module is used to monitor dynamic disturbance events in the processing process in real time according to the initial flexible operation scheduling scheme. When a dynamic disturbance event is triggered, it recalculates the micro-dynamic processing cost data of the remaining unexecuted processes based on the real-time updated micro-process-level production dataset, and performs adaptive rescheduling calculation on the initial flexible operation scheduling scheme to obtain the updated flexible operation scheduling scheme, thereby controlling the processing flow of physical entities in the workshop.

[0105] It should be noted that this system is a system corresponding to the above method. All implementation methods in the above method embodiments are applicable to this embodiment and can achieve the same technical effect.

[0106] Embodiments of the present invention also provide a computing device, including: a processor and a memory storing a computer program, wherein the computer program, when executed by the processor, performs the method described above. All implementations in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0107] Embodiments of the present invention also provide a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the method described above. All implementations in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0108] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for optimizing the operation and cost management of flexible work planning in CNC workshops, characterized in that, The method includes: Step 1: Collect spindle running status data, feed speed data, process execution status data, RFID-read tool remaining life data, and order delivery date data of heterogeneous CNC machine tools through the workshop Internet of Things to obtain multi-source heterogeneous data; Step 2: Perform timestamp alignment and feature fusion on the multi-source heterogeneous data to obtain a micro-process-level production dataset; Step 3: Based on the micro-process-level production dataset, extract the spindle load timing characteristics and tool wear rate characteristics of a specific process on a specific machine tool, calculate the corresponding dynamic energy consumption cost, tool wear cost and failure risk cost, and form micro-dynamic machining cost data with process, machine tool, tool matching and tool feed trajectory parameters as independent variables. Step 4: Based on the micro-dynamic processing cost data, construct an optimization solution space with the objective function of minimizing the overall operating cost and the constraints of the process sequence and equipment capacity. Encode the matching relationship between different processes, machine tools and cutting tools and the tool feed trajectory parameters into population individuals. Calculate the fitness value of each population individual in the optimization solution space, and iteratively update the population through selection, crossover and mutation operations until the preset convergence condition is met to obtain the initial flexible operation scheduling scheme. Step 5: Based on the initial flexible operation scheduling scheme, monitor dynamic disturbance events in the processing process in real time. When a dynamic disturbance event is triggered, recalculate the micro-dynamic processing cost data of the remaining unexecuted processes based on the real-time updated micro-process-level production dataset, and perform adaptive rescheduling calculation on the initial flexible operation scheduling scheme to obtain the updated flexible operation scheduling scheme, thereby controlling the processing flow of physical entities in the workshop.

2. The CNC workshop flexible operation planning cost optimization management method according to claim 1, characterized in that, By performing timestamp alignment and feature fusion on multi-source heterogeneous data, a micro-level production dataset at the process level is obtained, including: Using the actual start time of the process as a unified time reference, the time series labels in the multi-source heterogeneous data are subjected to sliding window interpolation and sampling rate resampling to obtain the time-aligned data sequence. Cross-domain feature mapping is performed on the time-aligned data sequence to establish a multi-dimensional topological association between spindle running fluctuations, tool consumption trajectory and order delivery nodes. Feature dimensions representing the resource interaction state of a single process are extracted to obtain the process-level association feature matrix. Based on the process-level associated feature matrix, feature vectorization is performed according to the matching and combination relationship of machine tool-tool-order, and continuous numerical representation reflecting the comprehensive state of the process processing environment is extracted to obtain the process-level fusion feature set; The process-level fusion feature set is dimensionality reduced, aggregated, and restructured to remove redundant dimensions and retain core processing state parameters, forming a micro-process-level production dataset that includes process time sequence fragments and related features.

3. The CNC workshop flexible operation planning cost optimization management method according to claim 2, characterized in that, Cross-domain feature mapping is performed on the time-aligned data sequence to establish a multi-dimensional topological association between spindle operation fluctuations, tool consumption trajectories, and order delivery nodes. Feature dimensions representing the resource interaction state of a single process are extracted to obtain a process-level association feature matrix, including: The spindle dynamic load data in the time-aligned data sequence is statistically analyzed in the time domain and the main frequency component is extracted in the frequency domain to obtain the spindle load fluctuation feature set. The remaining tool life data in the time-aligned data sequence is calculated by differential degradation according to the operation execution time window, and the rate is fitted by combining the past wear accumulation curve to obtain the tool wear rate feature set. The spindle load fluctuation feature set and the tool wear rate feature set are synchronized and matched on the time axis according to the same process execution time. The relative time deviation between the matched feature pairs and the planned completion time in the order delivery data is calculated. A multi-dimensional topology graph is constructed with the actual processing time of the process as the topology edge and the three-element combination of machine tool, tool and order as the topology node. Based on the multidimensional topological association graph, the load disturbance amplitude, tool wear increment, and delivery margin deviation values ​​of each process node under its corresponding machine tool and tool matching state are extracted and then scaled and compressed using dimensionless scaling and principal component compression to form a process-level association feature matrix with process index as the row and machine tool and tool combination state features as the column.

4. The CNC workshop flexible operation planning cost optimization management method according to claim 3, characterized in that, The multidimensional topology graph includes the correlation between load disturbance propagation path, tool decay process, and delivery urgency.

5. The CNC workshop flexible operation planning cost optimization management method according to claim 2, characterized in that, Based on the micro-level production dataset, the spindle load timing characteristics and tool wear rate characteristics of specific processes on specific machine tools are extracted. The corresponding dynamic energy consumption cost, tool wear cost, and failure risk cost are calculated, forming micro-level dynamic machining cost data with process, machine tool, tool matching, and tool feed trajectory parameters as independent variables. This includes: Extract the spindle load fluctuation variance sequence, speed decay slope sequence, and tool wear increment sequence of a specific process on a specific machine tool from the micro-process-level production dataset, and perform multi-dimensional feature splicing and quantization calculation to obtain the process-level machining state feature group. Based on the process-level machining status feature group, the power consumption integral mapping is performed in combination with the preset machine tool energy efficiency benchmark curve to obtain the power consumption conversion value. The tool wear conversion value is obtained by performing life depreciation calculation through the preset tool life depreciation accounting rules. The power consumption conversion value and the tool wear conversion value are combined to obtain the basic direct cost data. Input the basic direct cost data into the preset equipment health assessment rules, calculate the probability of spindle overload warning and the probability of tool failure, and combine the preset downtime loss rate per unit time to perform risk value conversion, and obtain full-element cost details including failure risk cost; By performing cost weight allocation and normalization aggregation on the detailed data of all factors of cost, a quantitative value representing the comprehensive economic cost of a single process under a specific matching state is formed, thus obtaining micro-dynamic processing cost data.

6. The CNC workshop flexible operation planning cost optimization management method according to claim 5, characterized in that, Based on micro-dynamic processing cost data, an optimization solution space is constructed with the objective function of minimizing overall operating cost and constraints of process sequence and equipment capacity. The matching relationships between different processes, machine tools, and cutting tools, as well as the tool feed trajectory parameters, are encoded as population individuals. The fitness value of each population individual in the optimization solution space is calculated, and the population is iteratively updated through selection, crossover, and mutation operations until a preset convergence condition is met, resulting in an initial flexible job scheduling scheme, including: Using micro-dynamic machining cost data as input parameters, a multi-dimensional decision matrix including process topology constraints and machine tool capacity constraints is constructed to form an optimization solution space. The machine tool allocation sequence, tool matching combination and feed trajectory parameters are mapped to chromosome coding structure to obtain the initial scheduling population. The individual codes in the initial scheduling population are mapped back to the optimization solution space for full path cost accounting. The inverse of the total operating cost corresponding to each scheduling sequence is calculated to quantify the quality of individuals in the solution space and obtain the set of individual fitness values. Based on the set of individual fitness values, roulette wheel selection and adaptive crossover operations are performed on the initial scheduling population. High-fitness gene segments are retained, and local trajectory parameter perturbation is performed on the crossover individuals to obtain the next generation iterative population. Determine whether the iteration rounds of the next generation of iterative population have reached the preset threshold or whether the fitness variance is lower than the convergence tolerance. If the convergence condition is met, extract the final fitness individuals for chromosome reverse decoding, restore the matching matrix and trajectory parameter set of process-equipment-tool, and obtain the initial flexible operation scheduling scheme.

7. The CNC workshop flexible operation planning cost optimization management method according to claim 6, characterized in that, Based on the initial flexible job scheduling scheme, dynamic disturbance events during the processing are monitored in real time. When a dynamic disturbance event is triggered, the micro-dynamic processing cost data of the remaining unexecuted processes is recalculated based on the real-time updated micro-process-level production dataset. An adaptive rescheduling calculation is then performed on the initial flexible job scheduling scheme to obtain an updated flexible job scheduling scheme, controlling the processing flow of physical entities in the workshop, including: Real-time acquisition of actual execution feedback messages after the initial flexible operation scheduling plan is issued to the physical entity in the workshop; comparison of planned progress nodes with actual completion status; calculation of time offset and equipment anomaly identifier; and obtaining the process execution deviation sequence. When the time offset or equipment anomaly flag in the process execution deviation sequence exceeds the preset tolerance threshold, a dynamic disturbance event is triggered, and the scheduling status snapshot of the completed process is frozen. The real-time operating parameters of the remaining processes to be executed are extracted to obtain the real-time status data of the remaining processes. The micro-process-level production dataset is updated based on the real-time status data of the remaining processes. The micro-dynamic processing cost data of the remaining processes to be executed is recalculated. Local optimization calculation is performed with the goal of minimizing the sum of the total cost of the remaining processes and the order delay penalty to obtain the rescheduling candidate sequence. The candidate sequences for rescheduling are checked for capacity conflicts and the legality of process routes. Scheduling paths that violate equipment capacity constraints are eliminated. The sequences that pass the verification replace the corresponding unexecuted nodes in the original plan, resulting in an updated flexible operation scheduling scheme, which is then sent to the workshop manufacturing execution terminal to control the processing flow of physical entities in the workshop.

8. A CNC workshop flexible operation planning cost optimization management system, wherein the system implements the method as described in any one of claims 1 to 7, characterized in that, include: The data acquisition module is used to collect spindle operating status data, feed speed data, process execution status data, RFID-read tool remaining life data, and order delivery date data of heterogeneous CNC machine tools through the workshop Internet of Things, so as to obtain multi-source heterogeneous data. The fusion module is used to perform timestamp alignment and feature fusion on multi-source heterogeneous data to obtain micro-level production datasets. The calculation module is used to extract the spindle load timing characteristics and tool wear rate characteristics of a specific process on a specific machine tool based on the micro-process-level production dataset, and calculate the corresponding dynamic energy consumption cost, tool wear cost and failure risk cost, forming micro-dynamic machining cost data with process, machine tool, tool matching and tool feed trajectory parameters as independent variables. The optimization module is used to construct an optimization solution space based on micro-dynamic processing cost data, with the objective function of minimizing the overall operating cost and the constraints of the process sequence and equipment capacity. It encodes the matching relationship between different processes, machine tools and cutting tools and the tool feed trajectory parameters as population individuals, calculates the fitness value of each population individual in the optimization solution space, and iteratively updates the population through selection, crossover and mutation operations until the preset convergence condition is met, and obtains the initial flexible operation scheduling scheme. The response module is used to monitor dynamic disturbance events in the processing process in real time according to the initial flexible operation scheduling scheme. When a dynamic disturbance event is triggered, it recalculates the micro-dynamic processing cost data of the remaining unexecuted processes based on the real-time updated micro-process-level production dataset, and performs adaptive rescheduling calculation on the initial flexible operation scheduling scheme to obtain the updated flexible operation scheduling scheme, thereby controlling the processing flow of physical entities in the workshop.

9. A computing device, characterized in that, include: One or more processors; A storage device for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to implement the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program that, when executed by a processor, implements the method as described in any one of claims 1 to 7.