A production rate and demand intensity joint rule generation method considering gear shifting physical constraints and edge manufacturing execution system

Abstract

The application discloses a production rate and demand intensity joint rule generation method considering gear switching physical constraints and an edge manufacturing execution system, and belongs to the technical field of industrial process control, manufacturing execution system and edge computing. The method determines the hysteresis bandwidth, the protection zone bandwidth and the upper limit of the switching rate according to the production gear switching wear matrix and the demand intensity gear switching cost matrix; the joint value iteration is performed on the inventory equivalent interval to obtain a benchmark gear mapping; adjacent states with the same joint gear are compressed into segments, and the touch boundary is asymmetrically adjusted; when the production gears are different, the touch boundary is contracted to the low inventory side, and when only the demand intensity gears are different, the touch boundary is expanded to the high inventory side; then, the two ends of each segment are inwardly contracted to form a gap area; the edge node checks the table in the event step cycle, the inventory equivalent falls into the gap area, the last step gear is kept, the count reaches the upper limit, the limit is kept, and the production line channel and the demand side channel are issued in parallel. The application can reduce wear and energy consumption peaks.

Description

Technical Field

[0001] This invention belongs to the fields of industrial process control, manufacturing execution systems and edge computing technology. Specifically, it relates to a method and edge manufacturing execution system for generating and executing joint control rules of production rate level and demand intensity level in spot production scenarios based on inventory equivalent, production line shift telemetry, energy consumption changes and demand-side configuration execution records. Background Technology

[0002] Spot-based production systems typically have discrete production rate settings such as stop, low speed, and high speed, as well as discrete demand intensity settings such as high, medium, and low, corresponding to different order arrival and acceptance intensities. Inventory equivalent changes with completion events, order arrival events, and under-delivery status. When inventory equivalent fluctuates repeatedly near its limits, if control rules only switch settings based on single-point thresholds, it will cause repeated actuator operation, accumulated mechanical wear, frequent energy consumption spikes, and execution channel jitter, thereby shortening equipment lifespan and affecting the stability of on-site control.

[0003] Existing technologies mainly include the following four categories: The first type is an inventory threshold control scheme that only targets production rate. This type of scheme switches between several production levels based on inventory levels, but it does not include demand intensity levels, nor does it form a joint output between production levels and demand intensity levels. Therefore, it cannot coordinate and suppress switching between the capacity side and the demand side.

[0004] The second category comprises theoretical models that combine capacity and demand-side flow control. These models mathematically combine capacity and demand-side patterns to derive an abstract optimal strategy function. For example, the paper "Inventory Management Problems in Stochastic Differential Game Models: Markov Chain Approximation and Optimal Strategy" (Hans Publishers OA, Chinese Journal) discloses a method for jointly solving production and demand-side control variables under stochastic demand conditions. It uses dynamic programming to derive coupled nonlinear integral-differential equations and employs Markov chain approximation to approximate the value function and optimal control, providing convergence analysis. Another example is the paper "Control policies for single-stage production systems with perishable inventory and customer impatience" (Annals of Operations Research), which discloses a joint strategy for inventory control and order access control in spot-type production systems. This strategy suspends production when inventory reaches a safe level and implements access judgment for arriving orders during stockout periods. The aforementioned literature discloses mathematical methods for jointly solving capacity and demand-side control quantities, but all assume that gear switching can be completed instantaneously, outputting an abstract strategy function. They do not disclose the segmented lookup table structure for edge nodes, the gap preservation area formed by the shrinking of segment boundaries, or the hard limit of the switching rate based on event steps.

[0005] The third category involves control schemes that use hysteresis intervals or dead zones to suppress frequent actuator movements. In the field of electrical drives and process control, to reduce the frequency of actuator movements, hysteresis intervals or dead zones are typically set on both sides of the setpoint of the controlled variable. The actuator is only triggered when the controlled variable exceeds this interval, thereby reducing switching losses and suppressing the reciprocating oscillations of the controlled variable near the setpoint. In this type of scheme, the hysteresis interval or dead zone is symmetrically set around the fixed setpoint of a single controlled variable. Its bandwidth value depends on manual tuning based on experience, lacking a quantitative correlation with the actual wear and energy consumption of the controlled actuator: a smaller bandwidth results in insufficient oscillation suppression, while a larger bandwidth leads to sluggish response. Furthermore, this type of scheme is designed for a single controlled variable. When two execution channels exist simultaneously—one on the production side and one on the demand side—the trigger points of the two channels are concentrated in the same state neighborhood, mutually stimulating each other, and repeated shifting still occurs.

[0006] The fourth category consists of general decision process solving tools or numerical calculation tools. These tools can output state-action mappings, but their outputs do not contain any mechanisms to inhibit frequent actuator actions, nor do they contain constraints on the number of actions per unit time. They also lack overload protection when the actuators are already under high load. Their output is a state-by-state mapping table, which requires a large amount of storage and has a long retrieval time, making it unsuitable for direct deployment at field control nodes.

[0007] In summary, existing technologies still have the following shortcomings in spot-production scenarios: First, although the control variables on the capacity and demand sides can be jointly solved, the solution process does not take into account the mechanical wear and energy consumption peaks generated by actuator movements. Equipment losses cannot be observed separately, and there is no basis for constraining control rules. Second, the hysteresis band or dead zone bandwidth used to suppress frequent movements depends on empirical tuning and lacks a corresponding relationship with the actual wear of the actuators. Moreover, it only applies to a single controlled variable, and in dual-channel execution scenarios, the movements of the two channels mutually incentivize each other. Third, the solution results are mostly abstract policy functions or state-by-state mapping tables, requiring real-time calculation by a host computer or cloud. Field control nodes cannot perform local closed-loop execution, limiting both real-time performance and reliability. Fourth, there is a lack of constraints on the number of movements per unit time, and actuators are at risk of overload under high-load conditions. Therefore, when the inventory equivalent fluctuates repeatedly near the limit, existing solutions still suffer from frequent shifting, wear accumulation, prominent energy consumption peaks, and execution channel jitter, which urgently need to be addressed. Summary of the Invention

[0008] To address the shortcomings of the existing technology, this invention provides a method for generating joint rules for production rate and demand intensity that considers the physical constraints of gear shifting, and an edge manufacturing execution system. The specific objectives are as follows: Firstly, after obtaining the baseline boundary through joint value iteration, a deployable gap holding area is formed by post-processing of the asymmetric protection band and the shrinking of the segmented boundary. This ensures that when the inventory equivalent falls into the gap near the boundary, it maintains the joint position of the previous event step, thereby suppressing boundary oscillations. Secondly, the edge manufacturing execution node completes the local closed loop of hysteresis lookup, switch count judgment and dual-channel distribution in a cycle of event steps, without relying on real-time inference in the cloud; Third, when the candidate gear changes, a hard limit is applied to the number of joint switching times within a single statistical period to avoid actuator overload; Fourth, control rules are regenerated according to the same technology chain after demand intensity shifts, and the original rules can be replaced after phased trial operation.

[0009] To achieve the above objectives, this invention provides a method for generating joint rules for production rate and demand intensity that considers the physical constraints of gear shifting, comprising the following steps: S1, Configure the production level set, demand intensity level set, and production level switching wear matrix K. u Demand intensity level switching cost matrix K d And the peak energy consumption proxy value for a single transaction, and based on K u With K d Determine the hysteresis bandwidth δ based on K u Determine the protection band bandwidth g and the upper limit of the switching rate N. max ; S2. Within the inventory equivalent range, using inventory equivalent xt For the state, the joint value iteration is performed with the production level u and the demand intensity level d as the joint control variables to obtain the benchmark joint level mapping; the instantaneous cost function in the joint value iteration consists of the inventory bias term, the production level operating load term and the production and demand intensity mismatch term, but does not include the monetary price term and the sales profit term. S3. Compress adjacent inventory states with the same joint level in the benchmark joint level mapping into segments, and make asymmetric protection band adjustments to the boundary between adjacent segments: when the production level of adjacent segments is different, regardless of whether the demand intensity level is the same, the boundary shrinks by g units towards the low inventory side; when only the demand intensity level is different, the boundary expands by g units towards the high inventory side; when the joint level is the same, the boundary remains unchanged; it should be noted here that the boundary refers to the boundary inventory equivalent value between adjacent segments. S4. Shrink the two ends of each effective interval of each segment after adjustment by S3 by δ units to form a non-covered gap between adjacent segments, thus obtaining the segmented lookup table rule. S5. Execute a local closed loop at the edge manufacturing execution node with an event step cycle: Collect the current inventory equivalent x t When x t When falling into the gap region, take the joint position of the previous event step as the candidate joint position; otherwise, take the x position from the segmented lookup table rule. t The joint gearing position corresponding to the segment is used as the candidate joint gearing position; when the switching count value C in the current statistical period t Reaching N max If the candidate joint level is different from the joint level of the previous event step, the joint level of the previous event step is output; otherwise, the candidate joint level is output and issued in parallel through the production line execution channel and the demand-side execution channel. Whenever the actually issued joint level is different from the joint level of the previous event step, the counter value C is switched. t Add 1.

[0010] Furthermore, in S5, the statistical period is reset every set number of event steps, switching the count value C. t The count is then reset to zero; and if either the production line execution channel or the demand-side execution channel fails to issue a response, both channels revert and maintain the joint gear of the previous event step. This issuance is not counted in the switching count value C. t .

[0011] The present invention also provides an edge manufacturing execution system for executing the above-described method for generating joint rules of production rate and demand intensity, including a physical parameter configuration module, a joint value iteration module, a segmented compression and protection band module, a hysteresis rule extraction module, an edge limiting execution module, and a drift update module.

[0012] The physical parameter configuration module is used to configure K u Kd And the single-time energy consumption peak proxy value, and determine δ, g, and N. max ; The joint value iteration module is used to perform joint value iteration over the inventory equivalent range to obtain the benchmark joint level mapping; The segmented compression and protection band module is used to compress adjacent inventory states to obtain segments, and to make asymmetrical protection band adjustments at the boundary. The hysteresis rule extraction module is used to shrink the two ends of the effective interval of each segment by δ to form a gap area and output the segment lookup table rules; The edge limiting execution module is set at the edge manufacturing execution node and is used to perform table lookup, gap maintenance, switching limiting judgment and dual-channel parallel distribution in a period of event steps; The drift update module is used to regenerate the segmented lookup table rules when the demand intensity vector is updated.

[0013] The beneficial effects of this invention are as follows: 1. The physical constraints of gear switching are structured into rule generation through parameter mapping. Wear costs, demand-side switching costs, and energy consumption peaks can be accumulated and observed separately, and can be recalibrated in reverse by field operation records to form a parameter closed loop. 2. The asymmetric protection zone and hysteresis contraction are implemented in layers, and their contributions can be verified separately; simulation shows that the main effect of reducing switching speed comes from hysteresis contraction and gap maintenance. 3. The gap preservation area is deployed in the form of segmented table lookup. The average time for a single table lookup at the edge node is about 0.000935ms, which is suitable for real-time execution on site and does not rely on cloud inference. 4. The switching rate hard limit can be triggered under pressure conditions to maintain the limit and prevent actuator overload; 5. After the demand intensity drifts, the recalculation rules are based on the same technology chain, and the threshold is automatically adjusted as the parameters drift. Replacement is supported after phased trial operation. Attached Figure Description

[0014] Figure 1 This is a flowchart of the method for generating joint rules of production rate and demand intensity according to the present invention; Figure 2 This is a schematic diagram illustrating the principle of adjustment and hysteresis retraction of the asymmetric protective zone at the contact point of the present invention. Figure 3 This is a structural block diagram of the edge manufacturing execution system of the present invention; Figure 4 This is a single-event step logic flowchart for edge limiting execution in this invention. Detailed Implementation

[0015] The present invention will be further described below with reference to the accompanying drawings and embodiments. The following embodiments are for illustrative purposes only and are not intended to limit the scope of the invention.

[0016] like Figure 1 As shown, the present invention provides a method for generating joint rules of production rate and demand intensity, including steps S1 to S5, and may further include step S6, as follows: I. Controlled Objects and Parameter Configuration Corresponding step S1: Configure the production level set, demand intensity level set, and production level switching wear matrix K. u Demand intensity level switching cost matrix K d And the peak energy consumption proxy value for a single transaction, and based on K u With K d Determine the hysteresis bandwidth δ based on K u Determine the protection band bandwidth g and the upper limit of the switching rate N. max .

[0017] This embodiment uses inventory equivalent x t The status input is used, with production level u and demand intensity level d as the joint control output. The production level set includes stop, low speed, and high speed; the demand intensity level set includes high demand intensity, medium demand intensity, and low demand intensity. Inventory equivalent x t This is a unified converted value for available inventory, under-delivery status, and production status. The main control quantities and their technical meanings are shown in Table 1.

[0018] Table 1 Record Sheet of Main Control Quantities and Technical Meanings Production gears are determined by u i Switch to u j At that time, the wear matrix is ​​recorded according to equation (1): K ij u =w ij wear (1) In equation (1), w ij wear For production gears, by u i Switch to u j The equivalent mechanical wear generated during gear shifting is calculated from component wear, impact load, and lubrication loss during the shifting process.

[0019] Demand intensity level from d j Switch to d k When switching, the switching cost is calculated according to equation (2): K jk d =η jk lat +η jk fail (2) In equation (2), ηjk lat Configure the latency cost for the demand side, η jk fail The cost of failure to issue the document.

[0020] Specifically, the hysteresis bandwidth δ, the protection band bandwidth g, and the upper limit of the switching rate N. max These are not independently tuned empirical parameters, but rather determined by K. u With K d The reverse determination method is as follows: (3) δ=⌈κ1·(max K u +max K d )⌉,g=⌈κ2·max K u ⌉, N max =⌊W lim / max K u ⌋ (3) In equation (3), max K u For the maximum wear equivalent in the production gear shift wear matrix, max K d W represents the maximum switching cost in the demand intensity level switching cost matrix. lim The maximum allowable cumulative wear within a single statistical period is defined by κ1 and κ2, which are calibration coefficients. ⌈·⌉ and ⌊·⌋ represent rounding up and rounding down, respectively. As shown in equation (3), the greater the physical cost of gear switching, the wider the hysteresis bandwidth and protection bandwidth, and the fewer the allowed switching times. This allows physical constraints to directly determine the shape of the rules in a structured manner, without having to convert equipment wear into a scalar penalty term and mix it into the optimization objective. In this embodiment, κ1 = 0.5, κ2 = 1.2, and W... lim =75、max K u =2.5, max K d =1.2, corresponding to δ=2, g=3, N max =30.

[0021] II. Joint Value Iteration and Asymmetric Guard Band Post-processing Corresponding step S2: Within the inventory equivalent range, using inventory equivalent x t Given the state, the production level u and the demand intensity level d are used as joint control variables to perform joint value iteration, which yields the baseline joint level mapping. The immediate cost function in the joint value iteration consists of inventory bias, production level operating load, and production-demand intensity mismatch, but does not include monetary price or sales profit.

[0022] Also corresponding to step S3: Compress adjacent inventory states with the same joint level in the benchmark joint level mapping into segments, and make asymmetric protection band adjustments to the boundary between adjacent segments: When the production level of adjacent segments is different, regardless of whether the demand intensity level is the same, the boundary shrinks by g units towards the low inventory side; when only the demand intensity level is different, the boundary expands by g units towards the high inventory side; when the joint level is the same, the boundary remains unchanged.

[0023] In the inventory status range [x min x max Perform joint value iteration on the above to obtain the baseline joint gear mapping. The joint value iteration is expressed by equation (4): V r+1 (x)=max u，d {[-J(x, u, d)+λ d ·V r (x - )+μ(u)·V r (x + )] / [α+λ d +μ(u)]} (4) In equation (4), λ d Let d represent the order arrival intensity corresponding to the demand intensity level, μ(u) represent the production capacity at the production level u, and x represent the order arrival intensity. - and x + For adjacent inventory states and truncated at the endpoints of the interval, x - =max(x−1,x min ), x + =min(x+1, x) max ), where α is the stabilization coefficient, taking a value greater than 0, used to ensure the convergence of the value iteration in the infinite time domain; in this embodiment, α is taken as 0.05; r is the iteration number, V r+1 V is a function of the value of the inventory state x obtained in the (r+1)th iteration. r (x - Let x be the inventory state obtained in the r-th iteration. - Value function, V r (x + Let x be the inventory state obtained in the r-th iteration. + The value function; max u，d This indicates that the maximum value is taken over all combinations of production level sets and demand intensity level sets; J(x, u, d) is the immediate cost function when using the combined level (u, d) under inventory state x. As defined by equation (5), it consists only of three terms: inventory bias, production level operating load, and production-demand intensity mismatch, and does not include monetary price or sales profit terms. J(x, u, d) = h(x) + e(u) + w imb ·(λ d -μ(u))2 (5) The inventory bias penalty is determined according to formula (6): h(x) = h h ·max(x, 0)+h b ·max(-x, 0) (6) In equation (5), e(u) is the operating load proxy value of production level u, w imb λ is the weight for the mismatch between supply and demand intensity. d Let h be the order arrival intensity corresponding to demand intensity level d, and μ(u) be the production capacity of production level u; in equation (6), h h h is the inventory bias coefficient. b This is the undercrossing bias coefficient.

[0024] After solving the state-by-state strategy, adjacent inventory states with the same joint gear (u, d) are compressed into segments, and the boundary is adjusted asymmetrically according to equation (7): In equation (7), b is the original contact limit obtained by compression. When the production level changes, the contact limit shrinks by g units towards the low inventory side, so that the capacity increase action is postponed until the inventory is even lower, avoiding the start of high-wear production line shifting when the inventory is still bearable; when only the demand intensity level changes, the contact limit expands by g units towards the high inventory side, so that the demand intensity adjustment is postponed until the inventory is even higher, avoiding frequent rewriting of the demand side configuration when the inventory is low. The above-mentioned opposite adjustments make the two types of execution actions staggered on the same inventory state axis, thereby eliminating the situation where the production line channel and the demand side channel jitter simultaneously in the same inventory neighborhood. The rule generation process is as follows: joint value iteration, segmented compression, asymmetric protection band post-processing, and hysteresis shrinkage, without including additional monotonic projection steps.

[0025] III. Hysteresis Retraction and Gap Maintenance Corresponding step S4: Shrink both ends of the effective interval of each segment by δ after adjustment in S3, so that a non-covered gap area is formed between adjacent segments, and the segment lookup table rule is obtained.

[0026] like Figure 2 As shown, for each segment of the protective belt after adjustment, its effective range is shrunk inward according to formula (8): [l k +δ, r k -δ] (8) In equation (8), l k With r k δ represents the lower and upper boundaries of the k-th segment after adjustment by the protection band according to formula (7), and δ is the hysteresis bandwidth.

[0027] The non-covered inventory interval formed after the adjacent segments are condensed is denoted as the gap area G. The lag lookup table is performed according to formula (9): In equation (9), x t For the inventory equivalent of the current event step, (u t d t ) represents the joint position of the current event step, (u t-1 d t-1 ) represents the joint gear of the previous event step; u*(x t ) and d*(x t (x) represents the inventory equivalent in the segmented lookup table rules. t The production level and demand intensity level corresponding to the segment are indicated by an asterisk, which indicates that it is the baseline level determined by the offline rule generation process.

[0028] As can be seen from equation (9), the gap area is not an implicit judgment condition, but an explicitly stored interval in the segmented lookup table rule, whose output is to keep the position of the previous event step. Therefore, it can be downloaded to the edge node along with the other segments and processed by the same lookup table process.

[0029] IV. Edge Limiting Implementation and Parameter Calibration like Figure 4 As shown, corresponding to step S5: Execute a local closed loop at the edge manufacturing execution node with an event step cycle: collect the current inventory equivalent x t When x t When falling into the gap region G, the joint position of the previous event step is taken as the candidate joint position; otherwise, the position x in the segmented lookup table rule is taken. t The joint gearing position corresponding to the segment is used as the candidate joint gearing position; when the switching count value C in the current statistical period t Reaching N max If the candidate joint position is different from the joint position of the previous event step, the joint position of the previous event step is output; otherwise, the candidate joint position is output and issued in parallel through the production line execution channel and the demand side execution channel.

[0030] The edge limiting execution module is located at the edge manufacturing execution node and includes a segmented lookup table unit, a gap holding unit, a switching counting unit, and a dual-channel distribution unit. Within each event step, the segmented lookup table unit determines the current inventory equivalent x. t The segment to which it belongs, the gap-preserving unit is in x t When falling into the gap region G, the combined gear position of the previous event step is output. The two are first obtained according to equation (9) to obtain the candidate gear position (u). t d t Then, perform the switching limit according to formula (10): In equation (10), C t N represents the switching count value within the current statistical period. max The upper limit of the switching rate is determined according to equation (3). This refers to the actual joint allocation after the threshold is limited. The statistical period is reset every fixed number of event steps. (C) t The timeout is then reset; this embodiment uses 100 event steps as a statistical cycle. When the candidate level changes but does not reach the limit, the system issues a response in parallel through the production line execution channel and the demand-side execution channel. .

[0031] To avoid a mismatch between production capacity and demand intensity caused by inconsistent results from the two channels, the dual-channel distribution unit is equipped with a dual-channel interlock: the switchover is only counted as C when both channels return a successful distribution. t And update the gear of the previous event step; if either channel fails to send or times out, both channels will roll back and maintain the joint gear of the previous event step (u). t-1 d t-1 This issuance will not be counted in C. t This interlocking mechanism ensures that the production capacity and demand sides are always in the same combination of levels determined by the same decision-making cycle.

[0032] In addition, N max It can be configured with two levels: a normal switching rate upper limit and a pressure switching rate upper limit. When the production line switching telemetry or energy consumption change collected by the edge node exceeds the set threshold, indicating that the actuator is under high load, N will be... max The normal switching rate limit is switched to the pressure switching rate limit to further reduce the number of allowed switching operations.

[0033] Edge manufacturing execution nodes synchronously perform parameter calibration during operation. Each time a production gear change occurs, the start and end gears of that change are set at K. u Extract the corresponding elements and sum them to obtain the cumulative wear value for production gear switching; for each demand intensity gear switching, calculate the wear value based on the start and end gears of that switching at K. d The corresponding elements are extracted and summed to obtain the cumulative value of the cost of switching demand intensity levels; for each joint level switch, the peak energy consumption surrogate value is accumulated to obtain the cumulative value of the peak energy consumption. The above three types of cumulative values ​​are stored separately for each level and are not merged, so equipment losses can be observed separately.

[0034] When the cumulative number of switching samples reaches the set value, K is recalibrated based on the actual cumulative result. u K d Single-cycle energy consumption peak proxy value: The recalibrated K is obtained by dividing the cumulative wear value of a certain gear pair by the actual number of switching times of that gear pair. ij uThe recalibrated K is obtained by dividing the cumulative switching cost of a certain gear pair by the actual number of switching operations for that gear pair. jk d Divide the cumulative energy consumption peak value by the total number of combined gear switching times to obtain the recalibrated single energy consumption peak proxy value. After recalibration, return to S1 and redetermine δ, g, and N according to formula (3). max Then, S2, S3 and S4 are executed again in sequence, so that the hysteresis bandwidth, protection bandwidth and switching rate limit are automatically adjusted according to the actual wear of the actuator, forming a parameter closed loop.

[0035] In one embodiment, K u With K d The initial value is given by the equipment maintenance manual and the latency statistics of the demand-side interface, corresponding to max K. u =2.5, max K d =1.2; After running 100,000 event steps, recalibrate as described above to obtain max K. u =2.95, maxK d =1.3, according to equation (3), we can redetermine δ=3, g=4, N max =25, meaning that the contact boundary and gap area widen as wear increases, and the number of switching operations allowed in a single statistical period decreases accordingly.

[0036] V. Drift Recalculation and System Composition Corresponding step S6: When the demand intensity vector is updated, S2, S3 and S4 are executed again in sequence to obtain new segmented lookup table rules. After phased trial operation, the new segmented lookup table rules replace the original segmented lookup table rules.

[0037] When the demand intensity vector is updated to λ′, the recalculation rule is applied according to equation (11): S′=H δ (G g (V(λ′))) (11) In equation (11), λ′ is the updated demand intensity vector, and its elements are the order arrival intensities corresponding to each updated demand intensity level; V represents the joint value iteration of equation (4), and G g The asymmetric guard band post-processing of expression (7), H δ The hysteresis extraction of equations (8) and (9) is represented by S′, which is the new segmentation rule after drift. The new rule replaces the original rule after phased trial operation.

[0038] like Figure 3As shown in Table 2, the edge manufacturing execution system of the present invention includes a physical parameter configuration module, a joint value iteration module, a segmented compression and guard band module, a hysteresis rule extraction module, an edge limiting execution module, and a drift update module. The physical parameter configuration module, joint value iteration module, segmented compression and guard band module, hysteresis rule extraction module, and drift update module generate control rules offline and can be deployed on a host computer or on the edge manufacturing execution node. The edge limiting execution module is located on the edge manufacturing execution node, receives the segmented lookup table rules downloaded from the hysteresis rule extraction module, and performs local closed-loop execution with an event step cycle. Its execution process does not rely on real-time inference from the host computer or the cloud.

[0039] Table 2 System Module Composition Record Table The edge limiting execution module further includes a segmented lookup unit, a gap holding unit, a switching counting unit, and a dual-channel delivery unit. The segmented lookup unit determines the segment to which the current inventory equivalent belongs; the gap holding unit outputs the joint gear position of the previous event step when the current inventory equivalent falls into the gap area; the switching counting unit counts the switching count value within the current statistical period, and when this value reaches N... max The output is limited and held; the dual-channel distribution unit is used to distribute joint gear positions in parallel through the production line execution channel and the demand-side execution channel, and triggers synchronous rollback of both channels if distribution fails or times out in either channel. The above four units are configured as follows: Figure 4 The single-event-step logic shown is executed sequentially, and the next event step is initiated after each execution is completed.

[0040] VI. Examples In one embodiment, the production levels are stop, low speed, and high speed; the demand intensity levels are high demand intensity, medium demand intensity, and low demand intensity, corresponding to arrival intensities of 3.7, 2.1, and 1.4, respectively; and the inventory range is [-20, 30]. After joint value iteration, asymmetric protection band post-processing, and hysteresis shrinkage, the segmentation rules shown in Table 3 are obtained. In this embodiment, the production boundary is approximately located around -5, and the demand intensity boundary includes the boundaries around -19 and 2.

[0041] Table 3. Record of Segmented Lookup Rules for Implementation Examples VII. Simulation Experiment Verification This invention has been verified through simulation experiments. The parameters and results listed in this section are all derived from simulation experiments. The simulation experiment parameters are shown in Table 4.

[0042] Table 4 Simulation Experiment Parameter Recording Table The simulation event model is as follows: each event step has a probability λ. d / (λ d +μ(u)) An order arrival event decreases inventory by 1, otherwise a completion event increases inventory by 1, with inventory limited to [-20, 30]. Boundary oscillation is defined as |x t |≤δ+2 and the actual execution level changes relative to the previous event step. The ablation comparison results for 100,000 event steps are shown in Table 5.

[0043] Table 5. Results of ablation using the 100,000 event-step strategy. Ablation results show that when only the asymmetric protection band post-processing is added, the number of joint switching operations decreases from 76,128 to 75,856, a limited reduction. After adding hysteresis contraction and gap maintenance, the number of joint switching operations decreases to 10,784, a reduction of 85.83% compared to the strategy without hysteresis, a 74.35% reduction in wear cost, and a 92.92% reduction in boundary oscillation from 76,127 to 5,391. Under normal limiting conditions, no limiting maintenance is required, indicating that hysteresis contraction has suppressed the switching rate to N. max The following results show that under stress test conditions, the limit was triggered 22,347 times, and the number of joint switching was further reduced to 5,722, verifying the protective effect of the hard limiter under high load on the actuator.

[0044] The above results show that the main switching effect of the present invention comes from hysteresis shrinkage and gap maintenance. The asymmetric protection band post-processing is responsible for offsetting the two types of execution actions on the inventory state axis. The hard limit is responsible for overload protection under pressure conditions. The functions of the three can be verified and independently tuned.

[0045] In terms of edge lookup performance, the average time for 102,000 hysteresis lookups is approximately 0.000935ms per lookup, indicating that segmented lookup, gap preservation, and amplitude limiting judgment are suitable for local real-time execution at the edge manufacturing execution node.

[0046] It should be noted that the average technical evaluation score after protection band post-processing and hysteresis retraction in the simulation is lower than that of the strategy without hysteresis. This is because the present invention prioritizes equipment protection and execution stability, rather than the extreme values ​​of output indicators. For production lines with significant shift wear, limited energy consumption peaks, or tight execution channel bandwidth, this trade-off has clear engineering value.

[0047] Before on-site deployment, K should be recalibrated using actual maintenance records, gear shift events, energy consumption curves, and demand-side configuration execution records. u K d Based on the single energy consumption peak proxy value, δ, g, and N are re-determined according to equation (3). max And after phased trial operation, formal control rules will be released.

[0048] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for generating joint rules for production rate and demand intensity that considers the physical constraints of gear shifting, characterized in that, Includes the following steps: S1, Configure the production level set, demand intensity level set, and production level switching wear matrix K. u Demand intensity level switching cost matrix K d And the peak energy consumption proxy value for a single transaction, and based on K u With K d Determine the hysteresis bandwidth δ based on K u Determine the protection band bandwidth g and the upper limit of the switching rate N. max ; S2. Within the inventory equivalent range, using inventory equivalent x t For the state, using the production level u and the demand intensity level d as joint control variables, perform joint value iteration to obtain the baseline joint level mapping; The immediate cost function in the joint value iteration consists of the inventory bias term, the production level operating load term, and the supply-demand intensity mismatch term, but does not include the monetary price term and the sales profit term. S3. Compress adjacent inventory states with the same joint level in the benchmark joint level mapping into segments, and make asymmetric protection zone adjustment on the boundary between adjacent segments: when the production level of adjacent segments is different, regardless of whether the demand intensity level is the same, the boundary shrinks by g units towards the low inventory side. When only the demand intensity level differs, the threshold expands outward by g units towards the high inventory side; when the combined level is the same, the threshold remains unchanged. S4. Shrink the two ends of each effective interval of each segment after adjustment by S3 by δ units to form a non-covered gap between adjacent segments, thus obtaining the segmented lookup table rule. S5. Execute a local closed loop at the edge manufacturing execution node with an event step cycle: Collect the current inventory equivalent x t When x t When falling into the gap region, take the joint position of the previous event step as the candidate joint position; otherwise, take the x position from the segmented lookup table rule. t The joint gearing position corresponding to the segment is used as the candidate joint gearing position; when the switching count value C in the current statistical period t Reaching N max If the candidate joint level is different from the joint level of the previous event step, the joint level of the previous event step is output; otherwise, the candidate joint level is output and issued in parallel through the production line execution channel and the demand-side execution channel. Whenever the actually issued joint level is different from the joint level of the previous event step, the counter value C is switched. t Add 1.

2. The method for generating joint rules of production rate and demand intensity according to claim 1, characterized in that, In S1, the value of δ is related to K. u The maximum wear equivalent quantity and K in d The sum of the maximum switching costs in the process is positively correlated; the value of g is positively correlated with K. u The maximum wear equivalent quantity is positively correlated with N; max The value of is positively correlated with the upper limit of cumulative wear allowed within a single statistical period and with K. u The maximum wear equivalent quantity is negatively correlated.

3. The method for generating joint rules of production rate and demand intensity according to claim 1, characterized in that, In S2, the instantaneous cost function J is determined by the following formula: J(x, u, d)=h h max(x, 0) + h b max(-x, 0) + e(u) + w imb ·(λ d -μ(u)) 2 In the formula, h h h is the inventory bias coefficient. b Here, e(u) is the under-competition bias coefficient, e(u) is the operating load proxy value of production level u, and w is the under-competition bias coefficient. imb λ represents the weighting of the mismatch between supply and demand intensity. d Let μ(u) be the order arrival intensity corresponding to the demand intensity level d, and let μ(u) be the production capacity of the production level u.

4. The method for generating joint rules of production rate and demand intensity according to claim 1, characterized in that, In S5, the statistical period is reset every set number of event steps, and the count value C is switched. t The count is then reset to zero; and if either the production line execution channel or the demand-side execution channel fails to issue a response, both channels revert and maintain the joint gear of the previous event step. This issuance is not counted in the switching count value C. t .

5. The method for generating joint rules of production rate and demand intensity according to claim 1, characterized in that, In S5, N max Configured as the upper limit of normal switching rate and the upper limit of pressure switching rate; When the changes in production line shifting or energy consumption collected by the edge manufacturing execution node exceed a set threshold, N will be... max The normal switching rate limit is switched to the pressure switching rate limit.

6. The method for generating joint rules of production rate and demand intensity according to claim 1, characterized in that, It also includes step S6: when the demand intensity vector is updated, S2, S3 and S4 are executed again in sequence to obtain new segmented lookup table rules. The new segmented lookup table rules replace the original segmented lookup table rules after phased trial operation.

7. The method for generating joint rules of production rate and demand intensity according to claim 1, characterized in that, It also includes a parameter calibration step: during the operation of the edge manufacturing execution node, press K u Accumulate the wear and tear costs of production gear switching respectively, by K d The switching costs for demand intensity level shifts are accumulated separately, and the peak energy consumption is accumulated based on the single energy consumption peak proxy value. K is then recalibrated using the accumulated results. u K d And the single energy consumption peak proxy value, then return to S1.

8. The method for generating joint rules of production rate and demand intensity according to claim 1, characterized in that, The production speed range includes stop, low speed, and high speed; the demand intensity range includes high demand intensity, medium demand intensity, and low demand intensity; inventory equivalent x t It is a unified conversion value for available inventory, under-delivery status, and production status.

9. An edge manufacturing execution system, characterized in that, The method for generating a joint rule for production rate and demand intensity according to any one of claims 1 to 8 includes: The physical parameter configuration module is used to configure K. u K d And the single-time energy consumption peak proxy value, and determine δ, g, and N. max ; The joint value iteration module is used to perform joint value iteration over the inventory equivalent range to obtain the benchmark joint level mapping; The segmented compression and protection band module is used to compress adjacent inventory states to obtain segments and to make asymmetrical protection band adjustments at the boundary. The hysteresis rule extraction module is used to shrink the two ends of the effective interval of each segment by δ to form a gap area and output the segment lookup table rules; The edge limiting execution module is set at the edge manufacturing execution node and is used to perform table lookup, gap maintenance, switching limiting judgment and dual-channel parallel distribution in a periodic event step; The drift update module is used to regenerate the segmented lookup table rules when the demand intensity vector is updated.

10. The edge manufacturing execution system according to claim 9, characterized in that, The edge limiting execution module includes: a segmented lookup unit for determining the segment to which the current inventory equivalent belongs; a gap holding unit for outputting the joint gear of the previous event step when the current inventory equivalent falls into the gap area; and a switching count unit for counting the switching count value within the current statistical period and reaching N. max It outputs a limiting and holding signal; a dual-channel sending unit is used to send joint gears in parallel through the production line execution channel and the demand side execution channel, and to trigger synchronous back-off of both channels when the sending fails in either channel.

Images