Bearing processing capacity optimization and production scheduling system

Abstract

The invention relates to the technical field of production management, and particularly discloses a bearing processing productivity optimization and production scheduling system, which comprises a process state sensing module, a dynamic bottleneck identification module, a self-adaptive buffer control module, a collaborative scheduling decision module and a productivity optimization execution module. Through cooperative work of the dynamic bottleneck identification module and the adaptive buffer control module, a static management mode of a traditional fixed buffer area is thoroughly changed. The system can sense the change of a production flow in real time, accurately identify a bottleneck, and dynamically adjust the target inventory level of a buffer area based on a fuzzy control algorithm, so that a buffer strategy can adaptively match a current production state. The problem that upstream and downstream processes wait for each other due to improper arrangement of a buffer area is solved, and passive blockage and material interruption between the processes are converted into active and flexible rhythm cooperation, so that systematic equipment waiting time is remarkably shortened, hidden capacity is released, and the equipment comprehensive utilization rate and output efficiency of a whole production line are improved.

Description

Technical Field

[0001] This invention belongs to the field of production management technology, specifically relating to a bearing processing capacity optimization and production scheduling system. Background Technology

[0002] In the complex production environment of discrete manufacturing, production scheduling and capacity optimization are core elements for improving enterprise operational efficiency and reducing manufacturing costs. This field involves the entire process from order receipt to product delivery, aiming to meet changing customer demands by rationally allocating production resources and optimizing process flows. The key lies in achieving a dynamic balance among multiple objectives such as equipment utilization, order delivery rate, and work-in-process inventory levels.

[0003] As a fundamental component of precision machinery, bearings exhibit typical discrete manufacturing characteristics in their machining process, which usually involves multiple key steps such as turning, grinding, and heat treatment. The goal of production scheduling is to efficiently coordinate the production rhythm of each step while meeting process constraints and delivery deadlines, ensuring smooth production flow and full utilization of resources, thereby maximizing overall capacity.

[0004] Existing technologies typically employ scheduling methods based on fixed rules or simple mathematical models, which have significant limitations in adaptability to complex scenarios with multiple processes and constraints, such as bearing machining. Specifically, at the junction of critical processes such as turning and grinding, the work-in-process inventory level fluctuates significantly due to the instability of the output cycle time of the preceding process and the fluctuation of the processing time of the subsequent process.

[0005] Existing systems mostly use a fixed-capacity buffer for simple control. This static strategy cannot dynamically respond to real-time changes in the production site and is prone to two adverse situations: when the buffer is too full, the upstream process is forced to stop and wait, resulting in the upstream equipment being idle. When the buffer zone is too empty, subsequent processes wait due to material shortages, causing downstream equipment to stop. This systemic equipment waiting time caused by poor coordination between processes severely restricts the release of the overall production line capacity, and a fixed buffer strategy is difficult to achieve global optimization under changing production orders and equipment conditions. Summary of the Invention

[0006] The present invention aims to provide a bearing processing capacity optimization and production scheduling system to solve the problems in the prior art, such as limited overall production line capacity, excessive equipment waiting time, and drastic fluctuations in work-in-process inventory levels caused by poor coordination between processes and the use of static buffering strategies.

[0007] The technical solution of this invention is a bearing processing capacity optimization and production scheduling system, which includes: The process status sensing module is used to collect processing status data, equipment operation data, and work-in-process flow data of each process in the bearing processing production line in real time. The dynamic bottleneck identification module is used to dynamically identify bottleneck and non-bottleneck processes in the current production process by calculating the real-time load rate and theoretical cycle time deviation of each process based on the real-time data collected by the process status perception module. The adaptive buffer control module is used to calculate and dynamically adjust the target inventory level of the work-in-process buffer located between the bottleneck process and the non-bottleneck process in real time based on the output of the dynamic bottleneck identification module. The collaborative scheduling decision module is used to generate collaborative production scheduling instructions for non-bottleneck processes based on the difference between the target inventory level output by the adaptive buffer control module and the real-time work-in-process inventory, combined with the current order queue and equipment status. The capacity optimization execution module receives and executes scheduling instructions generated by the collaborative scheduling decision module, driving the corresponding processing equipment, material handling devices, or production control systems to achieve precise control of the production cycle of non-bottleneck processes.

[0008] Furthermore, the specific identification process of the dynamic bottleneck identification module is as follows: the module continuously receives real-time data streams from the process status perception module, and calculates the average equipment utilization rate within a preset time window for each processing step on the production line as the real-time load rate. At the same time, the module calculates the absolute deviation between the real-time processing time and the theoretical cycle time based on the standard processing time of the process and the preset theoretical cycle time of the production line. This module has built-in bottleneck determination logic, which determines the process that has a real-time load rate exceeding the first preset threshold and a cycle time deviation value exceeding the second preset threshold as the bottleneck process of the current production process. Processes with a real-time load rate lower than the third preset threshold or a cycle time deviation value lower than the fourth preset threshold are identified as non-bottleneck processes in the current production process. The first preset threshold is set to 85%, the second preset threshold is set to 15% of the standard processing time, the third preset threshold is set to 70%, and the fourth preset threshold is set to 5% of the standard processing time.

[0009] Furthermore, the dynamic adjustment process of the adaptive buffer control module is as follows: the module receives the identification information of bottleneck processes and non-bottleneck processes output by the dynamic bottleneck identification module; This module manages the work-in-process buffer in the logical sense for each pair of bottleneck and non-bottleneck processes that have a sequential relationship. Based on the real-time load rate of the bottleneck process, the real-time load rate of the non-bottleneck process, and the actual inventory of the current buffer, this module calculates the target inventory level of the buffer in real time using a preset fuzzy control algorithm. The fuzzy control algorithm uses the bottleneck process load rate, non-bottleneck process load rate, and inventory deviation as input variables, and the adjustment amount of the target inventory level as the output variable. It contains a series of fuzzy rule bases preset based on the experience of production experts.

[0010] Furthermore, the universe of discourse and fuzzy subsets of the input and output variables of the fuzzy control algorithm are defined as follows: the universe of discourse for the bottleneck process load rate is 0% to 100%, and its fuzzy subset includes "low", "medium", and "high"; the universe of discourse for the non-bottleneck process load rate is 0% to 100%, and its fuzzy subset includes "low", "medium", and "high"; the universe of discourse for the inventory deviation is from the negative maximum value to the positive maximum value, and its fuzzy subset includes "negative large", "negative small", "zero", "positive small", and "positive large"; the universe of discourse for the target inventory level adjustment amount is from the negative maximum adjustment amount to the positive maximum adjustment amount, and its fuzzy subset includes "significantly reduced", "slightly reduced", "maintained", "slightly increased", and "significantly increased".

[0011] Furthermore, the fuzzy rule base contains multiple rules such as "If the bottleneck process load rate is high and the non-bottleneck process load rate is low and the inventory deviation is negative, then the target inventory level adjustment amount is a slight increase". Within each control cycle, this module performs three steps: fuzzification, rule reasoning, and defuzzification. It ultimately outputs the precise target inventory level adjustment amount, which is then added to the target inventory level of the previous cycle to obtain the target inventory level for the current cycle.

[0012] Furthermore, the decision-making process of the collaborative scheduling decision module is as follows: the module obtains the target inventory level set by the adaptive buffer control module for each buffer in real time, and obtains the actual work-in-process inventory quantity of each buffer through the process status perception module; This module calculates the inventory level deviation for each buffer, which is the difference between the target inventory level and the actual inventory quantity. Based on the sign and magnitude of the inventory level deviation, this module generates specific collaborative scheduling instructions in conjunction with the current queue of pending orders, equipment readiness status, and material availability for the corresponding non-bottleneck process. When the inventory level deviation is negative and the absolute value exceeds the preset deviation threshold, it indicates that the buffer inventory is too high and there is a risk of blocking the upstream bottleneck process. At this time, the module generates an instruction to reduce the production priority of the corresponding non-bottleneck process or temporarily slow down its production cycle. When the inventory level deviation is positive and the absolute value exceeds the preset deviation threshold, it indicates that the buffer inventory is too low and there is a risk of material shortage in the downstream bottleneck process. At this time, the module generates an instruction to increase the production priority of the corresponding non-bottleneck process or accelerate its production cycle.

[0013] Furthermore, the specific forms of the collaborative scheduling instructions include, but are not limited to: adjusting the operating speed percentage of non-bottleneck process processing equipment, inserting or delaying the processing sequence of specific production orders in non-bottleneck processes, and triggering the material handling system to replenish or remove raw materials or semi-finished products to non-bottleneck processes.

[0014] Furthermore, the capacity optimization execution module includes an instruction parsing unit and a device driving unit; The instruction parsing unit is used to convert the abstract scheduling instructions issued by the collaborative scheduling decision module into low-level control commands that can be recognized by specific production equipment or control systems. The equipment drive unit sends low-level control commands to the corresponding CNC machine tool, robot or programmable logic controller via industrial fieldbus or industrial Ethernet to physically change the production cycle or operation sequence of non-bottleneck processes.

[0015] Furthermore, the system operates in a hierarchical closed-loop control architecture, which includes a perception layer, an analysis layer, a decision-making layer, and an execution layer. The perception layer corresponds to the process status perception module, which collects production site data at a frequency of seconds or milliseconds. The analysis layer corresponds to the dynamic bottleneck identification module and the adaptive buffer control module, which perform bottleneck identification and buffer calculation at a frequency of minutes. The decision-making layer corresponds to the collaborative scheduling decision-making module, which generates scheduling instructions at a frequency of minutes. The execution layer corresponds to the capacity optimization execution module, which executes instructions in real time and provides feedback on the results, thus forming a closed-loop optimization system that continuously perceives, dynamically analyzes, makes real-time decisions, and executes precisely.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention, through the collaborative operation of a dynamic bottleneck identification module and an adaptive buffer control module, completely transforms the traditional static management mode of fixed buffers. The system can perceive changes in the production flow in real time, accurately identify bottlenecks, and dynamically adjust the target inventory level of the buffer based on a fuzzy control algorithm, enabling the buffer strategy to adaptively match the current production state. This mechanism fundamentally solves the problem of upstream and downstream processes waiting for each other due to improper buffer settings, transforming passive blockages and material shortages between processes into proactive and flexible rhythm coordination. This significantly reduces systemic equipment waiting time, releases hidden capacity, and improves the overall equipment utilization rate and output efficiency of the entire production line.

[0017] 2. The collaborative scheduling decision module proposed in this invention does not optimize a single process in isolation, but rather aims to maximize overall capacity by proactively serving and coordinating the production rhythm of bottleneck processes by adjusting the production pace of non-bottleneck processes. This bottleneck-process-centric collaborative scheduling strategy ensures that the most critical links in the value stream of the production system are always operating efficiently, avoiding global degradation caused by local optimization. Simultaneously, the decision mechanism based on real-time inventory deviation makes scheduling instructions highly responsive and accurate, effectively mitigating drastic fluctuations in work-in-process inventory levels and reducing inventory holding costs and production management complexity.

[0018] 3. This invention constructs a fully closed-loop automated optimization system from data perception to instruction execution. This system seamlessly integrates real-time data acquisition, intelligent analysis algorithms, and production control execution, achieving automation and intelligence in the capacity optimization process. Production managers no longer need to rely on experience for frequent manual intervention in scheduling; the system can autonomously cope with internal disturbances such as processing time fluctuations and order changes, continuously driving the production system towards its optimal operating state. This not only significantly reduces reliance on manpower and decision-making delays but also enables complex bearing processing lines to operate stably and efficiently under the modern manufacturing model of multi-variety, small-batch, and fast delivery, enhancing the company's market responsiveness and competitiveness. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the overall technical solution architecture of the bearing processing capacity optimization and production scheduling system proposed in this invention; Figure 2 This is a schematic diagram of the core principle framework of the collaborative operation of dynamic bottleneck identification and adaptive buffer control in this invention; Figure 3 This is a schematic diagram of the dynamic adjustment process framework of the adaptive buffer control module based on fuzzy logic in this invention; Figure 4 This is a logical flow diagram of the collaborative scheduling decision module in this invention that generates instructions based on inventory deviation; Figure 5 This is a schematic diagram of the hierarchical closed-loop control architecture and data flow from perception to execution in this invention. Detailed Implementation

[0020] Example 1: This invention provides a bearing processing capacity optimization and production scheduling system. Please refer to the appendix for the overall technical solution architecture. Figures 1 to 5 This system constructs a complete closed loop from real-time perception of production site data to automatic execution of optimization instructions. It aims to dynamically identify bottleneck constraints in the bearing processing production line and achieve continuous optimization and stable output of the entire production line's capacity by adaptively adjusting the work-in-process buffer level between processes and collaboratively scheduling the production rhythm of non-bottleneck processes.

[0021] This system completely abandons the traditional model that relies on fixed buffer strategies and manual experience scheduling. Through data-driven and intelligent decision-making, the production line can autonomously adapt to internal processing fluctuations and external order changes, ultimately achieving the core goals of improving the overall utilization rate of equipment, reducing work-in-process inventory, and shortening order delivery cycles.

[0022] The bearing processing capacity optimization and production scheduling system typically includes, in terms of physical deployment, various data acquisition terminals deployed on the production site, edge computing servers or industrial computers located in the workshop control room, and an industrial network connecting various processing equipment and material handling systems.

[0023] The system's software logic is functionally divided into five core modules: process status perception module, dynamic bottleneck identification module, adaptive buffer control module, collaborative scheduling decision module, and capacity optimization execution module. These modules operate collaboratively within a hierarchical closed-loop control architecture, which clearly distinguishes between data flow and control flow layers. Please refer to the appendix. Figure 5 .

[0024] The perception layer is responsible for high-frequency data acquisition, the analysis layer performs mid-frequency intelligent calculations, the decision layer generates optimization instructions, and the execution layer completes the physical execution and feedback of the instructions, thus forming a continuously operating optimization closed loop.

[0025] First, the process status perception module forms the data foundation of the entire system, corresponding to the perception layer in the hierarchical closed-loop control architecture. This module's task is to comprehensively and accurately capture the real-time status of the entire bearing processing production line at extremely high frequencies, typically on the order of seconds or even milliseconds. Its data acquisition covers three dimensions: processing status data, equipment operation data, and work-in-process flow data. The processing status data targets each specific process, such as turning, grinding, heat treatment, and assembly, and includes, but is not limited to: the currently executing processing task number, the timestamp indicating that the task has started processing, the standard processing time quota for this process, the batch number and serial number of the currently processed part, and the processing progress percentage read directly through the equipment interface.

[0026] The equipment operation data is obtained from the control system of the processing equipment itself. Key parameters include: the start and stop status of the spindle of the CNC machine tool or special machine, the ratio of the actual spindle speed to the set speed, the actual feed speed of the feed axis, the load current percentage of the servo drive system, the equipment alarm codes and alarm history, and the planned and unplanned downtime records of the equipment.

[0027] The collection of work-in-process flow data relies on automatic identification technology and sensor networks deployed at key nodes of the production line. For example, RFID readers or QR code scanners are installed at the entrance and exit of each process to record the precise timestamps of each bearing semi-finished or finished product entering and leaving the process. At the same time, photoelectric sensors or visual counting systems are deployed in buffer or temporary storage areas between processes to count the number of work-in-process items waiting to be processed in the buffer in real time.

[0028] To ensure reliable data acquisition, the process status sensing module must integrate multiple industrial communication protocol adapters. These adapters can communicate with different types of device controllers and sensors. For example, they can read real-time data from CNC systems via manufacturing technology protocols, exchange information with programmable logic controllers via an open platform communication unified architecture, and receive asynchronous signals from RFID readers and photoelectric sensors via Ethernet or serial ports using transmission control protocols.

[0029] All collected raw data must undergo a rigorous data cleaning and formatting sub-process before entering subsequent modules for processing. This sub-process verifies the integrity of the data, filters out obvious outliers such as rotational speed readings that exceed the physical range, synchronizes data from different time sources to a unified system time base, and converts all data into a standardized structure required for internal processing. Typical fields include data source identifier, timestamp, data type, value, unit, and data quality flag.

[0030] The processed real-time data stream will be continuously pushed to the dynamic bottleneck identification module via message queue or shared memory.

[0031] The dynamic bottleneck identification module is the key starting point for the system's intelligent analysis, located in the analysis layer of the hierarchical closed-loop control architecture. Its core function is to abandon static, historically experience-based bottleneck judgment methods and instead dynamically and quantitatively identify, based on the real-time data stream provided by the process status perception module, the critical processes (bottleneck processes) that currently restrict the output rate of the entire production line, as well as those processes with current capacity margins (non-bottleneck processes). The identification process is a continuously running, periodic calculation task, with a calculation cycle typically set to 1 to 5 minutes to balance real-time performance and computational overhead.

[0032] The dynamic bottleneck identification module executes the same analysis algorithm for each independent processing step on the production line within each calculation cycle. This algorithm includes two core calculation metrics: real-time load rate and theoretical cycle time deviation. The real-time load rate is calculated to reflect the activity level of the equipment in that process over a recent period. The module extracts the state sequence of the equipment in that process from the data stream within a preset time window, typically 30 minutes in length.

[0033] By analyzing the sequence, the module calculates the total time the equipment is in the "processing" state, and divides this total time by the time window length to obtain the average equipment utilization rate of this process within this window, i.e., the real-time load rate. The calculation of the theoretical cycle time deviation is used to measure the degree of matching between the actual processing speed of this process and the overall design rhythm of the production line.

[0034] The module first retrieves the standard processing time for the current product model from the process database. Simultaneously, the system maintains the theoretical cycle time of the production line, which is the ideal output interval set based on market demand and capacity planning. For each completed workpiece, the module calculates its actual processing time using the start and end timestamps recorded by the process status sensing module. Subsequently, it calculates the absolute difference between the actual processing time and the theoretical cycle time, and averages this deviation across all completed workpieces within the calculation period to obtain the theoretical cycle time deviation value for that period.

[0035] Based on the calculated deviation between the real-time load rate and the theoretical cycle time, the dynamic bottleneck identification module invokes its built-in bottleneck determination logic for final judgment. This determination logic employs a dual-threshold joint judgment strategy. Specifically, the logic presets four key thresholds: a first preset threshold, a second preset threshold, a third preset threshold, and a fourth preset threshold. According to the invention, the first preset threshold is set to 85%, the second preset threshold is set to 15% of the standard processing time, the third preset threshold is set to 70%, and the fourth preset threshold is set to 5% of the standard processing time.

[0036] The determination rule is as follows: For a process, if its real-time load rate exceeds the first preset threshold of 85%, and its theoretical cycle time deviation exceeds the second preset threshold of 15% of the standard processing time, then the process is determined to be a bottleneck process. This definition means that the process is not only very busy, but its actual processing speed is also significantly slower than the ideal rhythm of the production line, so it is most likely to slow down the output of the entire production line.

[0037] Conversely, for a given process, if its real-time load rate is below 70% of the third preset threshold, or its theoretical cycle time deviation is below the fourth preset threshold (5% of the standard processing time), then the process is considered a non-bottleneck process. This indicates that either the equipment utilization rate of the process is low, or its processing speed can keep up with or even exceed the production line cycle time, thus it has capacity margin and can be scheduled to support bottleneck processes.

[0038] All processes that neither meet the bottleneck nor non-bottleneck criteria are marked as "normal processes," and their status is under monitoring but they are not currently involved in proactive collaborative scheduling within this cycle. After completing the assessment of all processes, the dynamic bottleneck identification module will output a structured list, which clearly lists the identifiers of all bottleneck and non-bottleneck processes identified in the current cycle, along with their calculated index values. This list is one of the core inputs for subsequent modules to make decisions.

[0039] The adaptive buffer control module acts as a bridge connecting bottleneck identification and scheduling decisions. It is also located in the analysis layer of the hierarchical closed-loop control architecture. For its collaborative working principle framework with the dynamic bottleneck identification module, please refer to the appendix. Figure 2 The core responsibility of this module is to manage the work-in-process buffer, which exists virtually or physically between bottleneck processes and non-bottleneck processes.

[0040] Unlike traditional static buffer strategies that set fixed safety stock, this module dynamically calculates and adjusts the target inventory level of each buffer based on real-time production status. This allows the buffer to effectively isolate bottleneck processes from upstream fluctuations while avoiding excessive inventory that leads to tied-up production funds and space.

[0041] The adaptive buffer control module requires receiving a list of bottleneck and non-bottleneck processes from the dynamic bottleneck identification module before it can start operating. During system initialization, this module predefines all possible inter-process buffers based on the production line's process flow diagram, especially those buffers located between identified non-bottleneck processes and downstream bottleneck processes.

[0042] For each pair of non-bottleneck and bottleneck processes that have a sequential relationship, this module independently manages a logical buffer object. The status attributes of this buffer object include: a unique identifier for the buffer, the identifier of the associated preceding non-bottleneck process, the identifier of the associated following bottleneck process, the current actual work-in-process inventory, the current target inventory level, and the historical inventory level.

[0043] The dynamic adjustment process of this module is executed once within each control cycle, which is typically synchronized with the calculation cycle of the dynamic bottleneck identification module. The core of the adjustment process is a pre-defined fuzzy control algorithm. The algorithm's input consists of three key production state variables, and its output is the adjustment amount for the target inventory level. Please refer to the appendix for the framework of the fuzzy control algorithm. Figure 3 .

[0044] The three input variables and their universes of discourse and fuzzy subsets are defined as follows: The first input variable is the bottleneck process load rate, with a universe of discourse ranging from 0% to 100%, and its fuzzy subset is divided into "low," "medium," and "high." The second input variable is the non-bottleneck process load rate, with its universe of discourse also ranging from 0% to 100%, and its fuzzy subset is divided into "low," "medium," and "high." The third input variable is the inventory deviation, with its universe of discourse ranging from the negative maximum value to the positive maximum value, and its fuzzy subset is divided into "large negative," "small negative," "zero," "small positive," and "large positive." The formula for calculating the inventory deviation is: current target inventory level minus current actual work-in-process inventory. The output variable is the target inventory level adjustment amount, with its universe of discourse ranging from the negative maximum adjustment amount to the positive maximum adjustment amount, and its fuzzy subset is divided into "significantly reduced," "slightly reduced," "maintained," "slightly increased," and "significantly increased."

[0045] The specific scope of all domains and the shape of the membership functions of fuzzy subsets, such as triangular or trapezoidal membership functions, need to be parameterized during system deployment based on the specific production line's capacity, buffer physical capacity, and production management strategy.

[0046] The intelligence of fuzzy control algorithms lies in their internally encapsulated fuzzy rule base, derived from the experience of production experts. This rule base contains dozens of rules in the form of "if the premise is true, then the conclusion is true," used to describe how to adjust buffer inventory under different production scenarios. The combination of rules covers a wide range of possible input states. For example, a typical rule is: "If the bottleneck process has a high load rate and the non-bottleneck process has a low load rate and the inventory deviation is negative and small, then the target inventory level adjustment is a small increase."

[0047] This rule describes a production scenario where the downstream bottleneck process is very busy, while the upstream non-bottleneck processes are relatively idle, and the current actual inventory in the buffer is slightly lower than the target level. In this case, the system determines that the target inventory in the buffer should be appropriately increased to incentivize the upstream non-bottleneck processes to slightly accelerate production, thus reserving more work-in-process inventory for the upcoming high-load operation of the bottleneck process and preventing material shortages.

[0048] Conversely, another rule might be: "If the bottleneck process has a medium load rate and the non-bottleneck process has a high load rate and the inventory deviation is positive, then the target inventory level adjustment should be a significant reduction." This corresponds to upstream non-bottleneck processes producing too quickly, causing the buffer inventory to be far higher than the target level, posing a risk of excess inventory clogging the production line. Therefore, a significant reduction in the target inventory is needed to slow down upstream production.

[0049] Within each control cycle, the execution flow of the adaptive buffer control module strictly follows the three steps of fuzzy logic control: fuzzification, rule reasoning, and defuzzification. First, the module reads the precise input values ​​at the current moment: the real-time load rate of the bottleneck process and non-bottleneck processes comes from the output of the dynamic bottleneck identification module, the current actual inventory comes from the sensor data of the process status perception module, and the current target inventory level is the result of the previous cycle.

[0050] These precise values ​​are input into the membership functions of their respective variables to obtain their membership degrees to each fuzzy subset; this process is called fuzzification. Next, rule reasoning is performed. The module traverses all rules in the fuzzy rule base, matching and logically operating the fuzzy membership degree of the current input variable with the premise of each rule, typically using minimum operations to determine the activation strength of each rule. Then, based on the activation strength of each rule, its conclusion part is pruned to output the membership function of the fuzzy set. The output fuzzy sets of all activated rules are then aggregated to obtain a comprehensive output fuzzy set, which represents the fuzzy distribution of the variable "target inventory level adjustment amount". Finally, defuzzification is performed.

[0051] The module applies a defuzzification algorithm, such as the centroid method, to the aggregated output fuzzy set to calculate a precise adjustment value that can be used for practical control. This adjustment value is then algebraically added to the target inventory level determined in the previous period to obtain the new target inventory level for that buffer in the current period. This module performs the above calculations in parallel for all active buffers and outputs a list containing the latest target inventory level values ​​for each buffer.

[0052] The collaborative scheduling decision module is the "brain" of the system, generating specific optimization instructions, and is located at the decision layer of the hierarchical closed-loop control architecture. This module receives target inventory levels from each buffer from the adaptive buffer control module, and actual work-in-process inventory quantities from each buffer from the process status perception module. Its core decision logic is to generate collaborative production scheduling instructions for non-bottleneck processes based on inventory deviations. Please refer to the appendix for the logic flow. Figure 4 .

[0053] The collaborative scheduling decision module first calculates a key indicator for each managed buffer: inventory level deviation. This deviation value equals the target inventory level set by the adaptive buffer control module minus the actual work-in-process inventory quantity monitored by the process status perception module. A positive deviation value indicates insufficient actual inventory, requiring replenishment; a negative deviation value indicates excess actual inventory, requiring consumption.

[0054] The decision engine has a built-in preset deviation threshold to determine whether the deviation is significant enough to trigger scheduling intervention. This threshold is typically set based on the buffer capacity, the unit value of the material, and historical statistics of production fluctuations. The decision process is as follows: For each buffer, the module checks whether the absolute value of its inventory level deviation exceeds the preset deviation threshold.

[0055] If the inventory level is not exceeded, the current inventory status is considered to be within an acceptable range and no immediate adjustment is required. The non-bottleneck processes corresponding to this buffer can continue to produce according to the original plan or the default pace.

[0056] If the absolute value of the inventory level deviation exceeds the preset deviation threshold, the module initiates a detailed scheduling decision sub-process. This sub-process not only considers the magnitude and direction of the deviation, but also must comprehensively evaluate the current operating context of the corresponding non-bottleneck process.

[0057] The module will query the following information through the process status perception module: the current operating status of the non-bottleneck process equipment is processing, standby, or faulty; the current queue of orders to be processed for this process, including the priority, quantity, and estimated processing time of each order in the queue; the availability status of the raw materials or semi-finished products from the previous process required to produce this batch of workpieces in the material library; and whether the equipment is within the planned maintenance window.

[0058] Based on the deviation direction and context information, the module generates specific collaborative scheduling instructions. When the inventory level deviation is negative and its absolute value exceeds the threshold, it indicates that the inventory in the buffer is too high, posing a risk of blocking upstream processes or even indirectly affecting more upstream bottleneck processes. In this case, the instructions generated by the module aim to reduce the production intensity of the non-bottleneck processes corresponding to the buffer.

[0059] The specific form of the instruction may include: sending a command to the CNC system of the process to reduce its spindle speed or feed rate by a percentage, such as reducing it to 90% of the rated speed; or modifying its production scheduling queue to temporarily remove some low-priority orders from the current processing sequence, delaying their processing time; or triggering the material handling system to suspend feeding materials to the process.

[0060] When the inventory level deviation is positive and its absolute value exceeds the threshold, it indicates that the inventory in the buffer zone is too low, and the downstream bottleneck process faces the risk of material shortage and shutdown. At this time, the module generates instructions designed to increase the production intensity of this non-bottleneck process. Specific forms of these instructions may include: increasing the operating speed percentage of the equipment in this process; inserting high-priority rush orders into its current production queue; Alternatively, the material handling system can be instructed to prioritize replenishing raw materials for this process and ensure that the handling equipment at its output end is in the most efficient state, so as to transport the produced work-in-process to the downstream buffer as soon as possible.

[0061] The instructions generated by the collaborative scheduling decision module are relatively high-level instructions oriented towards business logic, such as "increase the production speed of process A005 to 105%" or "advance the processing sequence of process A003 in order BATCH_202 by two positions". These instructions are encapsulated into structured messages and sent to the capacity optimization execution module.

[0062] The capacity optimization execution module is the last link in the system closed loop and is also the key to converting digital instructions into physical actions. It is located in the execution layer of the hierarchical closed-loop control architecture.

[0063] This module is responsible for receiving, parsing, and ultimately executing all scheduling instructions issued by the collaborative scheduling decision module, ensuring that optimization strategies are implemented. This module typically consists of two core sub-units: an instruction parsing unit and a device driver unit.

[0064] The instruction parsing unit acts as a "translator." It receives abstract scheduling instructions from the decision-making layer, which describe "what to do" but do not specify "how to do it." Internally, the instruction parsing unit maintains a device instruction mapping library, which defines how each type of scheduling instruction is translated into low-level control commands or parameter sets that can be recognized and executed by a specific model of device controller.

[0065] For example, for the instruction "increase the production speed of process A005 to 105%", the instruction parsing unit needs to query the mapping library to find that the equipment corresponding to process A005 is a specific model of CNC grinding machine. Its speed control can be achieved by modifying the feed rate F code in the CNC program or by directly writing the speed multiplier parameter through the manufacturing technology protocol. The parsing unit will generate one or more precise control command frames that conform to the communication protocol of the equipment based on the instruction content.

[0066] The device drive unit acts as the "executor." It establishes a physical connection with the actual processing equipment, robots, programmable logic controllers, and material handling system controllers on the production line via industrial fieldbus or industrial Ethernet. After the instruction parsing unit generates the low-level control commands, the device drive unit is responsible for reliably sending these commands to the target device through the appropriate network protocol.

[0067] For example, modified CNC program blocks can be sent to the CNC system via Ethernet, or Boolean variables controlling the start and stop of the material conveyor can be written to the programmable logic controller via fieldbus. The equipment drive unit also needs basic communication status monitoring and error retry mechanisms. After sending a command, it listens for feedback signals from the equipment or indirectly verifies whether the command has been executed through a process status sensing module. If no confirmation is received within a certain time or the status does not change as expected, the drive unit will record an error log and retry or report the exception according to a preset strategy.

[0068] Through the actions of the capacity optimization execution module, the production cycle time or work sequence of non-bottleneck processes is physically altered, directly impacting the inventory level of the work-in-process buffer. This change is then captured by the process status perception module, generating new real-time data and initiating the next cycle's perception, analysis, decision-making, and execution loop. The entire system operates continuously in this way, dynamically tracking production bottlenecks, adaptively adjusting buffer strategies, and collaboratively scheduling production resources, ultimately driving the bearing processing line to always operate towards maximizing capacity and optimizing inventory.

[0069] Example 2: Based on the system described in Example 1, this example provides a specific implementation of a bearing processing capacity optimization and production scheduling system for a mixed production mode of multiple varieties and small batches. While maintaining the core architecture, this implementation focuses on enhancing the algorithms and strategies of the dynamic bottleneck identification module and the collaborative scheduling decision module to address challenges such as large differences in processing time and frequent production preparation activities caused by frequent product switching.

[0070] In this embodiment, the process status sensing module needs to additionally collect data related to product changeover. This data includes: the currently processed product model code, the standard processing time matrix for that model in each process, historical data on the preparation time required for changeover, and the switching status of molds, fixtures, or programs involved in the changeover process. This data provides crucial input for subsequent modules to perform more refined bottleneck identification and scheduling decisions.

[0071] The identification logic of the dynamic bottleneck identification module needs adaptive expansion. When calculating the real-time load rate, all equipment state times cannot be simply considered as load. The module needs to distinguish between "processing time" and "changeover preparation time." Within the calculation cycle, the time the equipment is in the "processing" state is included in the effective load, while the time in states such as "changeover setup" and "debugging" should be included in auxiliary time. The formula for calculating the real-time load rate is adjusted to: effective processing time divided by the total calculation cycle time. Simultaneously, the calculation of theoretical cycle time deviation also needs to be grouped and statistically analyzed by product model.

[0072] The module needs to maintain a deviation record table with product model as the key. It calculates the deviation between the actual processing time and the theoretical cycle time of the production line for each product model, and when identifying bottlenecks, it considers the currently mainstream product models for weighted or prioritized evaluation. Bottleneck determination thresholds, especially the second and fourth preset thresholds, can be set to be dynamically related to the standard processing time. However, for production lines with frequent product changes, these thresholds may need to be fine-tuned based on the processing complexity of the product family. For example, for product families with large processing time variances, a more lenient deviation threshold can be used.

Claims

1. A bearing processing capacity optimization and production scheduling system, characterized in that, include: The process status sensing module is used to collect processing status data, equipment operation data, and work-in-process flow data of each process in the bearing processing production line in real time. The dynamic bottleneck identification module is used to dynamically identify bottleneck and non-bottleneck processes in the current production process by calculating the real-time load rate and theoretical cycle time deviation of each process based on the real-time data collected by the process status perception module. An adaptive buffer control module is used to calculate and dynamically adjust the target inventory level of the work-in-process buffer located between the bottleneck process and the non-bottleneck process in real time based on the output of the dynamic bottleneck identification module. The collaborative scheduling decision module is used to generate collaborative production scheduling instructions for the non-bottleneck process based on the difference between the target inventory level output by the adaptive buffer control module and the real-time work-in-process inventory, combined with the current order queue and equipment status. The capacity optimization execution module is used to receive and execute the scheduling instructions generated by the collaborative scheduling decision module, drive the corresponding processing equipment, material handling device or production control system, and realize precise control of the production cycle of the non-bottleneck process.

2. The bearing processing capacity optimization and production scheduling system according to claim 1, characterized in that, The identification process of the dynamic bottleneck identification module is as follows: The dynamic bottleneck identification module continuously receives real-time data streams from the process status sensing module, and calculates the average equipment utilization rate within a preset time window for each processing step on the production line as the real-time load rate. Meanwhile, the dynamic bottleneck identification module calculates the absolute deviation between the real-time processing time and the theoretical cycle time based on the standard processing time of the process and the preset theoretical cycle time of the production line. The dynamic bottleneck identification module has a built-in bottleneck determination logic. The bottleneck determination logic determines the process with a real-time load rate exceeding a first preset threshold and an absolute deviation value exceeding a second preset threshold as the bottleneck process of the current production process. Processes with a real-time load rate lower than the third preset threshold or an absolute deviation value lower than the fourth preset threshold are determined to be non-bottleneck processes in the current production process.

3. The bearing processing capacity optimization and production scheduling system according to claim 2, characterized in that, The first preset threshold is set to 85%, the second preset threshold is set to 15% of the standard processing time, the third preset threshold is set to 70%, and the fourth preset threshold is set to 5% of the standard processing time.

4. The bearing processing capacity optimization and production scheduling system according to claim 1, characterized in that, The dynamic adjustment process of the adaptive buffer control module is as follows: The adaptive buffer control module receives the identification information of bottleneck processes and non-bottleneck processes output by the dynamic bottleneck identification module. The adaptive buffer control module manages the work-in-process buffer in logic for each pair of bottleneck and non-bottleneck processes that have a sequential relationship. The adaptive buffer control module calculates the target inventory level of the buffer in real time based on the real-time load rate of the bottleneck process, the real-time load rate of the non-bottleneck process, and the actual inventory of the current buffer using a preset fuzzy control algorithm.

5. The bearing processing capacity optimization and production scheduling system according to claim 4, characterized in that, The fuzzy control algorithm uses the bottleneck process load rate, the non-bottleneck process load rate, and the inventory deviation as input variables, and the adjustment amount of the target inventory level as the output variable. The universes of discourse and fuzzy subsets of the input and output variables of the fuzzy control algorithm are defined as follows: The domain of the bottleneck process load rate is 0% to 100%, and its fuzzy subset includes low, medium, and high; the domain of the non-bottleneck process load rate is 0% to 100%, and its fuzzy subset includes low, medium, and high. The universe of discourse for inventory deviation is from the negative maximum value to the positive maximum value, and its fuzzy subset includes negative large, negative small, zero, positive small, and positive large. The domain of the target inventory level adjustment is from the negative maximum adjustment to the positive maximum adjustment, and its fuzzy subset includes significant reduction, slight reduction, maintenance, slight increase, and significant increase.

6. The bearing processing capacity optimization and production scheduling system according to claim 5, characterized in that, The fuzzy control algorithm contains a fuzzy rule library preset based on the experience of production experts; the fuzzy rule library contains multiple rules that indicate a small increase in the target inventory level when the bottleneck process load rate is high, the non-bottleneck process load rate is low, and the inventory deviation is negative. The adaptive buffer control module performs three steps—fuzzification, rule reasoning, and defuzzification—in each control cycle, and finally outputs the precise target inventory level adjustment amount, which is then added to the target inventory level of the previous cycle to obtain the target inventory level for the current cycle.

7. The bearing processing capacity optimization and production scheduling system according to claim 1, characterized in that, The decision-making process of the collaborative scheduling decision module is as follows: the collaborative scheduling decision module obtains the target inventory level set by the adaptive buffer control module for each buffer in real time, and obtains the actual work-in-process inventory quantity of each buffer through the process status perception module. The collaborative scheduling decision module calculates the inventory level deviation for each buffer, which is the difference between the target inventory level and the actual inventory quantity. The collaborative scheduling decision module generates specific collaborative scheduling instructions based on the sign and magnitude of the inventory level deviation, combined with the current queue of orders to be processed for the corresponding non-bottleneck process, the equipment readiness status, and the material availability.

8. The bearing processing capacity optimization and production scheduling system according to claim 7, characterized in that, When the inventory level deviation is negative and the absolute value exceeds the preset deviation threshold, the collaborative scheduling decision module generates an instruction to reduce the production priority of the corresponding non-bottleneck process or temporarily slow down its production cycle. When the inventory level deviation is positive and the absolute value exceeds the preset deviation threshold, the collaborative scheduling decision module generates an instruction to increase the production priority of the corresponding non-bottleneck process or accelerate its production cycle.

9. The bearing processing capacity optimization and production scheduling system according to claim 1, characterized in that, The capacity optimization execution module includes an instruction parsing unit and a device driving unit; The instruction parsing unit is used to convert the scheduling instructions issued by the collaborative scheduling decision module into low-level control commands that can be recognized by specific production equipment or control systems. The device drive unit then sends the underlying control commands to the corresponding CNC machine tool, robot, or programmable logic controller via industrial fieldbus or industrial Ethernet.

10. The bearing processing capacity optimization and production scheduling system according to claim 1, characterized in that, The system operates on a hierarchical closed-loop control architecture, which includes a perception layer, an analysis layer, a decision-making layer, and an execution layer. The perception layer corresponds to the process status perception module, which collects production site data at a frequency of seconds or milliseconds. The analysis layer corresponds to the dynamic bottleneck identification module and the adaptive buffer control module, and performs bottleneck identification and buffer calculation at a frequency of minutes. The decision-making layer corresponds to the collaborative scheduling decision module, which generates scheduling instructions at a minute-level frequency. The execution layer corresponds to the capacity optimization execution module, which executes instructions in real time and provides feedback on the results.

Images