A high-frequency high-speed PCB impedance adaptive regulation control system and method

Abstract

This invention relates to the field of printed circuit board manufacturing management data processing, and discloses a high-frequency, high-speed PCB impedance adaptive control system and method, including a material process feature aggregation unit, an impedance prediction analysis unit, and a business management decision scheduling unit. The material process feature aggregation unit acquires the factory and monitoring feature data of the current batch of materials and outputs a state feature tensor. The impedance prediction analysis unit calculates an impedance prediction indicator factor based on this data. The business management decision scheduling unit calculates the capital loss when the factor crosses a threshold, and rewrites the underlying routing label to change the order attribute when the loss is below the diversion threshold. This invention transforms defect interception into a closed-loop adjustment that interlocks loss prediction and production scheduling, avoiding the scrapping of work-in-process, converting stagnant materials into degradeable assets, and maintaining the continuity of production scheduling decisions.

Description

Technical Field

[0001] This invention belongs to the field of printed circuit board manufacturing management data processing technology, and particularly relates to a high-frequency, high-speed PCB impedance adaptive control system and method. Background Technology

[0002] Currently, in the production scheduling management system of printed circuit board manufacturing, the customized production scheduling and online quality supervision of high-frequency and high-speed printed circuit boards fall under the control of the manufacturing management system. Conventional management models guide the flow of materials between processes by configuring static calibration process routes and rely on end-point physical interception to isolate non-compliant assets in order to maintain the yield of finished products. With the increasing requirements for the continuity accuracy of printed circuit impedance in high-frequency and high-speed signal transmission, the intrinsic dispersion of the electrical properties of copper-clad laminate raw materials and the accumulation of deviations in lamination and etching composite processes have limited the control of rigid control strategies at the physical level. Open-loop supervision methods such as on-line edge sampling are lagging behind the temporal evolution of material processing. This makes it impossible for the management decision-making end to perceive in real time when high-value substrate materials experience batch impedance drift, thus causing large-scale work-in-process scrap losses and disordered mismatch of production plans.

[0003] Industry attempts have attempted to improve physical interception accuracy by simply tightening physical milling tolerances or stacking local physical sensors. However, this approach lacks practical value for large-scale application due to hardware modification costs and equipment pump wear. In addition to physical hardware constraints, existing software monitoring and evaluation strategies also have shortcomings when dealing with complex production scheduling logic. For example, Chinese invention patent CN120911397B discloses a PCB impedance characteristic prediction and evaluation method based on neural networks. It improves the impedance prediction accuracy under specific operating conditions through physical prior regularization constraints. Essentially, it is anchored to a single-dimensional quality diagnosis function and fails to penetrate physical layer data and transform it into business layer scheduling logic. In large-scale flexible manufacturing scenarios, when there is a lack of real-time rewriting capability of the underlying routing labels of the manufacturing management system, even if impedance deviation is predicted, high-value work-in-process still faces rigid scrapping due to impedance non-compliance on the original high-standard order path due to the lack of degradeable asset diversion paths. This cannot solve the deep mismatch between production planning and the real-time physical and chemical state of materials, and it is difficult to overcome the constraints of production planning management and physical mismatch of material characteristics. It also cannot meet the needs of continuous flexible production scheduling and the reprocessing of special materials.

[0004] Therefore, the technical problem to be solved by this invention is how to use multi-task sequence characteristic flow to predict impedance error probability to assess the remaining carry-over value of material assets, and how to achieve adaptive diversion scheduling of pre-production scheduling by rewriting the underlying routing labels of the manufacturing management system. Summary of the Invention

[0005] This invention aims to solve the problem of predicting impedance deviation risk using multi-task sequence characteristic flow and achieving dynamic production scheduling adjustment through underlying route rewriting.

[0006] In this technical solution, a high-frequency, high-speed PCB impedance adaptive control system is provided, the system comprising: The system includes a material process feature aggregation unit, an impedance prediction and analysis unit, and a business management decision-making and scheduling unit. The material process feature aggregation unit is connected to the impedance prediction and analysis unit, which in turn is connected to the business management decision-making and scheduling unit. The material process feature aggregation unit acquires the substrate factory feature data and line-edge physical monitoring feature data of the current batch of materials processed for high-frequency and high-speed printed circuit boards, and completes tensor alignment according to the timestamp sequence to assemble and output the state feature tensor. The impedance prediction analysis unit receives the state feature tensor and calculates the impedance prediction indicator factor reflecting the degree of impedance deviation of the current batch of materials through the impedance regression model. When the impedance prediction indicator factor crosses the safety deviation threshold, the business management decision scheduling unit calculates the capital loss carried over to the current node by combining the real-time bill of materials cost data of the current batch of materials. When the capital loss is lower than the low economic diversion threshold, the underlying routing label of the manufacturing management system is automatically rewritten, changing the target order attribute of the current batch of materials from server motherboard order to low frequency communication backplane order, and allocating the current batch of materials to the low frequency printed circuit board production queue.

[0007] Preferably, when the business management decision-making and scheduling unit calculates the capital loss carried over to the current node, the system controls the compensation timing of sudden disturbances in the sensor network: the business management decision-making and scheduling unit extracts the impedance prediction indicator factor and identifies sudden jumps greater than 15% as transient disturbances; the business management decision-making and scheduling unit sets the sampling timing deviation parameter according to the transient disturbance, modifies the sampling window timing of the next cycle sensor, and calls the historical rolling average of the current batch of materials at the data link end to complete the smoothing filter replacement.

[0008] Preferably, under the condition of system data flow, the system controls network interruption degradation scheduling: when the communication interruption is within 50μs, the business management decision scheduling unit automatically extracts the historical rolling average of the current batch of materials and completes smooth replacement; when the network interruption duration exceeds the 100ms threshold, the business management decision scheduling unit cuts off the continuous prediction state and reduces the dimension to the batch-level risk compliance judgment mode based on the static factory characteristics of raw materials.

[0009] Preferably, the process of the material process feature aggregation unit acquiring line-edge physical monitoring feature data includes: the material process feature aggregation unit collecting dielectric layer thickness deviation data and etching linewidth fluctuation data of the current batch of materials; the material process feature aggregation unit acquiring copper foil roughness parameters of the current batch of materials, and combining the dielectric layer thickness deviation data, etching linewidth fluctuation data, and copper foil roughness parameters into the line-edge physical monitoring feature data.

[0010] Preferably, the process of the material process feature aggregation unit acquiring substrate factory feature data includes: the material process feature aggregation unit reading the original dielectric constant and original dielectric loss factor of the copper clad laminate of the current batch of materials; the material process feature aggregation unit retrieving the factory resin content and fiberglass cloth weaving density of the current batch of materials, and combining the original dielectric constant, original dielectric loss factor, factory resin content and fiberglass cloth weaving density into the substrate factory feature data.

[0011] Preferably, the process of the impedance prediction analysis unit running the impedance regression model includes: the impedance prediction analysis unit inputting the substrate manufacturing characteristic data and the line-side physical monitoring characteristic data into a multi-layer nonlinear regression network; the impedance prediction analysis unit using the multi-layer nonlinear regression network to determine the nonlinear mapping relationship between the substrate manufacturing characteristic data and the line-side physical monitoring characteristic data and the target impedance, and outputting an impedance prediction indicator factor.

[0012] Preferably, the process by which the business management decision scheduling unit determines the safety deviation threshold includes: the business management decision scheduling unit establishing a first safety deviation upper limit boundary and a first safety deviation lower limit boundary for the high-level server motherboard order as the safety deviation threshold; when the impedance prediction indicator factor exceeds the first safety deviation upper limit boundary or falls below the first safety deviation lower limit boundary, the business management decision scheduling unit determines that the current batch of materials has generated impedance deviation scrap loss and triggers subsequent cost accounting.

[0013] Preferably, the process by which the business management decision-making scheduling unit controls the reorganization and circulation status when the capital loss is higher than the low economic diversion threshold includes: when the capital loss is greater than or equal to the low economic diversion threshold, the business management decision-making scheduling unit determines that the current batch of materials does not have the value of being downgraded and realized assets; the business management decision-making scheduling unit synchronously issues a stop circulation instruction to the underlying scheduling route, suspends the circulation queue of the current batch of materials in place at the manufacturing management system end, and generates a special material scrap warning signal at the system end.

[0014] Preferably, the system also includes a data synchronization module: the data synchronization module is connected to both the material process feature aggregation unit and the business management decision scheduling unit; while the material process feature aggregation unit is acquiring feature data, the data synchronization module collects the real-time process status of the upstream multi-task sequence work-in-process and transmits the real-time process status synchronously to the business management decision scheduling unit, so that the business management decision scheduling unit can mark the work-in-process asset carry-over amount on the multi-task sequence processing path when calculating the capital loss carried over at the current node.

[0015] A high-frequency, high-speed PCB impedance adaptive regulation and control method, which operates within a high-frequency, high-speed PCB impedance adaptive regulation and control system, includes the following steps: Step 1001: The material process feature aggregation unit acquires the substrate factory feature data and line-side physical monitoring feature data of the current batch of materials, and completes tensor alignment according to the timestamp sequence to assemble and output the state feature tensor. Step 1002: The impedance prediction analysis unit receives the state feature tensor and calculates the impedance prediction indicator factor through the impedance regression model. Step 1003: When the impedance prediction indicator factor crosses the safety deviation threshold, the business management decision scheduling unit calculates the capital loss carried over to the current node by combining the real-time bill of materials cost data of the current batch of materials. When the capital loss is lower than the low economic diversion threshold, the unit automatically rewrites the underlying routing label of the manufacturing management system, changes the target order attribute of the current batch of materials from server motherboard order to low frequency communication backplane order, and allocates the current batch of materials to the low frequency printed circuit board production queue.

[0016] Compared with existing technologies, the high-frequency, high-speed PCB impedance adaptive control system and method of the present invention have the following advantages: 1. In high-frequency and high-speed PCB impedance adaptive control, the material process feature aggregation unit collects the intrinsic parameters of the copper-clad laminate and the processing features of the process sensor network online. The data is aligned through the timestamp sequence to reconstruct the state feature tensor. The impedance prediction analysis unit determines the operator by the operation rules and generates the impedance prediction indicator factor. The business management decision scheduling unit compares the factor with the safety deviation threshold range. When the indicator factor crosses the truncation boundary, the flow of high-specification server motherboard orders is stopped. By rewriting the underlying routing label of the production and manufacturing management system, the current processing batch is changed to a low-specification low-frequency communication backplane order. This avoids the scrapping of the entire board caused by the dispersion of raw materials and the accumulation of processing defects from the source. It transforms the end-point physical interception and supervision mode into a closed-loop regulation that interlocks online loss prediction and dynamic production scheduling.

[0017] 2. To address packet loss caused by strong electromagnetic interference and thickness deviation caused by sensor temperature drift, the system relies on transient data verification gating to filter online feature streams in real time. When communication interruption is within 50μs, it automatically extracts the rolling average of historical materials in the batch to complete smooth replacement. When the network interruption duration exceeds the 100ms threshold, it adaptively cuts off the continuous prediction state and reduces the dimension to a batch-level risk compliance judgment mode based on the static factory characteristics of raw materials. This dual defense mechanism effectively maintains the continuity of production scheduling decisions, eliminates downtime interference caused by false anomalies, and smoothly absorbs the underlying random physical noise and long-term network failures at the data processing layer.

[0018] 3. The business management decision-making and scheduling unit integrates cost accounting operators, combining impedance prediction drift probability with real-time bill of materials cost data for the current batch of materials to calculate the capital loss carried over to the current node. When the loss is lower than the preset low economic diversion threshold, the conditional control flow is automatically triggered to reconstruct the network and issue a downgrade usage instruction, automatically allocating the batch of materials to the low-frequency printed circuit board production queue, realizing the conversion of the residual value of special material assets. This process transforms simple quality defect interception into scheduling decisions based on material cost loss, establishing a causal transmission chain between quality dispersion and asset carryover. While maintaining the final impedance yield rate to 99.5%, it converts stagnant work-in-process on the multi-task processing path into compliant assets that can be downgraded and consumed. Attached Figure Description

[0019] Figure 1 This is a decision flow diagram of the high-frequency and high-speed PCB impedance adaptive regulation and control method of the present invention; Figure 2 This is a module architecture diagram of the high-frequency, high-speed PCB impedance adaptive control system of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0021] A high-frequency, high-speed PCB impedance adaptive control system, the system comprising: The system includes a material process feature aggregation unit, an impedance prediction and analysis unit, and a business management decision-making and scheduling unit. The material process feature aggregation unit is connected to the impedance prediction and analysis unit, which in turn is connected to the business management decision-making and scheduling unit. The material process feature aggregation unit acquires the substrate factory feature data and line-edge physical monitoring feature data of the current batch of materials processed for high-frequency and high-speed printed circuit boards, and completes tensor alignment according to the timestamp sequence to assemble and output the state feature tensor. The impedance prediction analysis unit receives the state feature tensor and calculates the impedance prediction indicator factor reflecting the degree of impedance deviation of the current batch of materials through the impedance regression model. When the impedance prediction indicator factor crosses the safety deviation threshold, the business management decision scheduling unit calculates the capital loss carried over to the current node by combining the real-time bill of materials cost data of the current batch of materials. When the capital loss is lower than the low economic diversion threshold, the underlying routing label of the manufacturing management system is automatically rewritten, changing the target order attribute of the current batch of materials from server motherboard order to low frequency communication backplane order, and allocating the current batch of materials to the low frequency printed circuit board production queue.

[0022] Preferably, when the business management decision-making and scheduling unit calculates the capital loss carried over to the current node, the system controls the compensation timing of sudden disturbances in the sensor network: the business management decision-making and scheduling unit extracts the impedance prediction indicator factor and identifies sudden jumps greater than 15% as transient disturbances; the business management decision-making and scheduling unit sets the sampling timing deviation parameter according to the transient disturbance, modifies the sampling window timing of the next cycle sensor, and calls the historical rolling average of the current batch of materials at the data link end to complete the smoothing filter replacement.

[0023] Preferably, under the condition of system data flow, the system controls network interruption degradation scheduling: when the communication interruption is within 50μs, the business management decision scheduling unit automatically extracts the historical rolling average of the current batch of materials and completes smooth replacement; when the network interruption duration exceeds the 100ms threshold, the business management decision scheduling unit cuts off the continuous prediction state and reduces the dimension to the batch-level risk compliance judgment mode based on the static factory characteristics of raw materials.

[0024] Preferably, the process of the material process feature aggregation unit acquiring line-edge physical monitoring feature data includes: the material process feature aggregation unit collecting dielectric layer thickness deviation data and etching linewidth fluctuation data of the current batch of materials; the material process feature aggregation unit acquiring copper foil roughness parameters of the current batch of materials, and combining the dielectric layer thickness deviation data, etching linewidth fluctuation data, and copper foil roughness parameters into the line-edge physical monitoring feature data.

[0025] Preferably, the process of the material process feature aggregation unit acquiring substrate factory feature data includes: the material process feature aggregation unit reading the original dielectric constant and original dielectric loss factor of the copper clad laminate of the current batch of materials; the material process feature aggregation unit retrieving the factory resin content and fiberglass cloth weaving density of the current batch of materials, and combining the original dielectric constant, original dielectric loss factor, factory resin content and fiberglass cloth weaving density into the substrate factory feature data.

[0026] Preferably, the process of the impedance prediction analysis unit running the impedance regression model includes: the impedance prediction analysis unit inputting the substrate manufacturing characteristic data and the line-side physical monitoring characteristic data into a multi-layer nonlinear regression network; the impedance prediction analysis unit using the multi-layer nonlinear regression network to determine the nonlinear mapping relationship between the substrate manufacturing characteristic data and the line-side physical monitoring characteristic data and the target impedance, and outputting an impedance prediction indicator factor.

[0027] Preferably, the process by which the business management decision scheduling unit determines the safety deviation threshold includes: the business management decision scheduling unit establishing a first safety deviation upper limit boundary and a first safety deviation lower limit boundary for the high-level server motherboard order as the safety deviation threshold; when the impedance prediction indicator factor exceeds the first safety deviation upper limit boundary or falls below the first safety deviation lower limit boundary, the business management decision scheduling unit determines that the current batch of materials has generated impedance deviation scrap loss and triggers subsequent cost accounting.

[0028] Preferably, the process by which the business management decision-making scheduling unit controls the reorganization and circulation status when the capital loss is higher than the low economic diversion threshold includes: when the capital loss is greater than or equal to the low economic diversion threshold, the business management decision-making scheduling unit determines that the current batch of materials does not have the value of being downgraded and realized assets; the business management decision-making scheduling unit synchronously issues a stop circulation instruction to the underlying scheduling route, suspends the circulation queue of the current batch of materials in place at the manufacturing management system end, and generates a special material scrap warning signal at the system end.

[0029] Preferably, the system also includes a data synchronization module: the data synchronization module is connected to both the material process feature aggregation unit and the business management decision scheduling unit; while the material process feature aggregation unit is acquiring feature data, the data synchronization module collects the real-time process status of the upstream multi-task sequence work-in-process and transmits the real-time process status synchronously to the business management decision scheduling unit, so that the business management decision scheduling unit can mark the work-in-process asset carry-over amount on the multi-task sequence processing path when calculating the capital loss carried over at the current node.

[0030] A high-frequency, high-speed PCB impedance adaptive control method includes the following steps: Step 1001: The material process feature aggregation unit acquires the substrate factory feature data and line-side physical monitoring feature data of the current batch of materials, and completes tensor alignment according to the timestamp sequence to assemble and output the state feature tensor. Step 1002: The impedance prediction analysis unit receives the state feature tensor and calculates the impedance prediction indicator factor through the impedance regression model. Step 1003: When the impedance prediction indicator factor crosses the safety deviation threshold, the business management decision scheduling unit calculates the capital loss carried over to the current node by combining the real-time bill of materials cost data of the current batch of materials. When the capital loss is lower than the low economic diversion threshold, the unit automatically rewrites the underlying routing label of the manufacturing management system, changes the target order attribute of the current batch of materials from server motherboard order to low frequency communication backplane order, and allocates the current batch of materials to the low frequency printed circuit board production queue.

[0031] Example 1: In a customized flexible manufacturing scenario involving multiple specifications of high-frequency and high-speed printed circuit boards, the manufacturing management system faces frequent switching between high-value server motherboard orders and low-frequency communication backplane orders. At this time, due to the implicit dielectric constant dispersion fluctuations within upstream copper-clad laminate raw material batches, and the cumulative non-ideal processing mechanical deformation and noise interference in the midstream lamination and etching processes, the conventional production line's reliance on end-of-line sampling inspection for monitoring and interception results in management time lags. Consequently, when batch impedance drift occurs, timely asset path interruption and material scheduling intervention cannot be implemented at the management and scheduling end. This continuously leads to the ineffective flow of expensive special substrate materials along process paths with systemic quality deviations, causing significant work-in-process scrap losses and production schedule disruptions. To address the management challenges of large-scale delayed scrapping of work-in-process, this system operates continuously during the production process. The material process feature aggregation unit collects substrate outgoing feature data and line-edge physical monitoring feature data of the current batch of materials online through the standard data interface of the factory manufacturing execution system. The substrate outgoing feature data includes the original dielectric constant data of the copper-clad laminate after desensitization labeling and the nominal substrate thickness data of the process. The line-edge physical monitoring feature data includes the online thickness feature deviation flow of the lamination process and the real-time feature sequence of the spray pressure of the etching process. The material process feature aggregation unit extracts the above-mentioned multi-variable heterogeneous non-electrical variables and completes tensor alignment according to the timestamp sequence, assembling and outputting a standardized state feature tensor representing the current online asset quality status.

[0032] The impedance prediction analysis unit receives standardized state characteristic tensors via a high-speed local data bus. These tensors are then input into a built-in differential impedance prediction attenuation network. Utilizing a white-box control flow based on domain physicochemical common sense to eliminate the mathematical black-box nature of multidimensional function optimization, the online thickness characteristic deviation flow of the lamination process is established as the core information source determining subsequent impedance evolution. After passing a fusing verification regarding incremental complexity, the real-time characteristic sequence of the etching process spray pressure is introduced as a second auxiliary information source. From this, an impedance prediction indicator factor reflecting the degree of impedance deviation in the current batch of material is calculated. The algebraic relationship of this factor is expressed as the original dielectric of the copper-clad laminate... The electrical constant data is multiplied by the sum of the nominal substrate thickness data and the online thickness characteristic deviation flow of the lamination process. The resulting product is then divided by the real-time characteristic sequence of the etching process's spray pressure. This scheme integrates algebraic relationship calculations with a multi-layer nonlinear regression network in a hierarchical cascade manner. The initial result output from the algebraic relationship calculation is not used as the final judgment index, but is defined as a one-dimensional initial benchmark quality operator. This operator is then combined with the fundamental physical quantities extracted from the state feature tensor to form a composite feature vector, which is then input into the multi-layer nonlinear regression network. This utilizes the nonlinear fitting capability of the neural network to correct for mechanical deformation and environmental factors. Noise-induced algebraic formula residuals are addressed, eliminating structural mutual exclusion between analytical and statistical learning methods. A one-dimensional matrix is ​​constructed by extracting the original dielectric constant of the copper-clad laminate and the online thickness characteristic deviation flow from the lamination process from the state feature tensor. This matrix is ​​then input into a multi-layer nonlinear regression network comprising a six-node input layer, a twelve-node first hidden layer, a six-node second hidden layer, and a single-node output layer. The node gain of the first hidden layer is calculated using a linear rectified function, and the node gain of the second hidden layer is calculated using a hyperbolic tangent function. The node weights of the multi-layer nonlinear regression network are adjusted via backpropagation within a local computing unit with a floating-point throughput of at least 10 billion operations per second. The bias parameter is set until the root mean square error between the network output value at 100 consecutive sampling points and the actual impedance measurement value in the training input source is less than 0.005 ohms. The mapping relationship between the substrate factory characteristic data, the line-side physical monitoring characteristic data and the target impedance is determined to output the impedance prediction indicator factor. The business management decision scheduling unit extracts the impedance prediction indicator factor and inputs it into the conditional control flow reconstruction network. It is compared with the safety deviation threshold range formed by the first safety deviation upper limit boundary and the first safety deviation lower limit boundary established for the high-level server motherboard order. When the impedance prediction indicator factor is between 0.85 and 1.When the data is within the first safety deviation threshold range of 15, the system determines that the current online asset status meets the highest technical specifications of the target order and maintains the normal flow of the original production scheduling sequence. However, when transient disturbances occur in the data due to strong electromagnetic interference from the sensor network or environmental temperature drift, causing a sudden jump of more than 15% in the impedance prediction indicator factor, the business management decision scheduling unit automatically triggers the built-in secondary bottleneck precision hedging strategy. It extracts the transient disturbance and quantifies it into a sampling timing deviation parameter to adaptively modify the sampling window timing of the next cycle sensor. At the same time, it calls the historical rolling average of the current batch of materials at the data link end to complete the smoothing filter replacement. Thus, under the condition that the communication interruption is within 50μs, random physical noise is eliminated to maintain the continuity of production scheduling decisions. The business management decision scheduling unit and the data acquisition board with a sampling frequency of not less than 50 MHz establish a bidirectional asynchronous data interaction bus through the hard real-time edge computing gateway. When the hard real-time edge computing gateway captures a single-point jump amplitude of more than 15% in the impedance prediction indicator factor and determines it as a transient disturbance, it triggers the internal real-time interrupt service program and directly rewrites the timer matching register of the data acquisition board. The clock comparison value in the timer causes the sensor sampling window timing of the next cycle to generate a displacement on the time axis equivalent to the sampling timing deviation parameter. Specifically, the ratio between the jump amplitude and the timing adjustment amount is pre-fixed inside the hard real-time edge computing gateway. When a single-point jump amplitude of ΔP is captured, for example, ΔP is greater than 15%, the system calculates the corresponding sampling time delay Δt through a preset sensitivity gain coefficient. Then, the delay Δt is divided by the hard real-time crystal oscillator period of the main control board, which is 20ns in this embodiment, to obtain the discrete integer pulse of the timer comparison value. The system calculates and accumulates the count value, writing it directly into the timer matching register. This achieves a precise physical translation of amplitude disturbances into time axis window displacement. It then obtains the historical rolling average of processed batch materials stored in a local high-speed cache via the fieldbus network. This historical rolling average replaces burst sampling data with single-point jumps exceeding 15%. Under conditions where communication interruptions are within 50 microseconds, the system delivers a smoothed and filtered data feature matrix to the business management decision-making scheduling unit, maintaining the scheduling continuity of orders whose underlying routing labels in the manufacturing management system have changed to low-frequency communication backplane orders.

[0033] As midstream process deviations accumulate, when the impedance prediction indicator factor crosses the hard cutoff boundary of the safety deviation threshold, the business management decision scheduling unit determines that the current batch of materials in the processing batch has incurred impedance over-tolerance scrapping losses and triggers subsequent cost accounting. By integrating financial cost accounting operators and combining the real-time bill of materials cost data of the current batch of materials in the processing batch, the system calculates the capital loss carried forward at the current node. When the calculated capital loss is lower than the preset low economic diversion threshold, the system automatically triggers conditional control flow reconfiguration of the network and issues a downgrade usage instruction. By automatically rewriting the underlying routing labels of the manufacturing management system, the target order attribute corresponding to the current batch of materials is transferred from the high-level service... The server motherboard order was changed to a low-frequency communication backplane order, thereby covering and rewriting the business target parameters in the pre-production scheduling sequence. The materials of the current processing batch were automatically allocated to the general low-frequency printed circuit board production queue with high impedance dispersion tolerance. From the source, the stagnant work-in-process was converted into compliant assets that can be degraded and consumed. The financial cost accounting operator in the business management decision scheduling unit read the real-time bill of materials cost data on the work-in-process path of the current processing batch of materials, and calculated the algebraic sum of the fixed asset value transferred to the current processing step and the material waste loss due to impedance deviation as a quantitative indicator of the capital loss transferred at the current node. When the capital loss is lower than that of the target low-frequency printed circuit board, the capital loss is calculated. When the low economic diversion threshold is determined by reverse calibration of the physical tolerance limit of the interlayer dielectric crosslinking degree that the circuit board processing line can withstand under subsequent lamination thermal stress, control code is sent to the address register of the field programmable logic controller to change the target physical path to the low-frequency communication backplane production line. This drives the automated transfer trolley's mechanical gripper to change the material transport trajectory, moving the processed batch of materials into the low-frequency printed circuit board production queue to change the physical work-in-process processing flow path. When the capital loss is greater than or equal to the low economic diversion threshold, the system synchronously sends a stop flow command to the underlying scheduling route and suspends the flow queue of the current processed batch of materials in place at the manufacturing management system end, with low cost and low data. The data-driven control model for overhead addresses the bottleneck of traditional production scheduling supervision, which lacks reverse reconstruction and adaptive diversion of routes. It maximizes the utilization of the residual value of special materials in the full data flow tracking decision-making process. This non-physical entity data processing architecture, which reversely transmits the discrete information of underlying physical quality to the front end of the supply chain and the production scheduling decision-making end, establishes a strict two-way dependent symbiotic relationship between the hardware asset status and software scheduling logic in the manufacturing system. By transforming the physical processing error into an asset loss risk vector at the management level in real time, it ensures that the final impedance yield is maintained at 99.5% while leveraging the flexible self-healing capability of the entire manufacturing network to adaptively absorb process defects.

[0034] Example 2: When the system faces a test condition where the dielectric constant of the copper-clad laminate substrate fluctuates between 3.21 and 3.68, and the thickness deviation of the lamination process line is accompanied by mechanical displacement noise of ±0.03mm, making it difficult to reconfigure and allocate online assets of customized high-frequency and high-speed printed circuit boards, the high-frequency and high-speed printed circuit board impedance adaptive control system adopts a semi-physical integrated test platform combining full-link online simulation and physical monitoring. The platform outputs continuous material flow characteristics of the production line through data acquisition boards with a sampling rate of not less than 50MHz and a voltage measurement resolution of 0.1mV. The basic data required for its analysis comes from the historical tracking of the full volume of materials in the automated printed circuit board manufacturing plant during 48 hours of continuous operation. The sampling period in the rational data set is controlled by the real-time interaction of high-dimensional data and the data processing load of the central processing unit. When the system detects a material conveying rate of 3.5 m / min, the sampling period converges to its lower limit of 2.5 ms. At this time, the platform superimposes Gaussian white noise with a signal-to-noise ratio of 20 dB and power frequency harmonic electromagnetic interference with a frequency of 50 Hz in the test environment to simulate the industrial site environment and multi-source data aliasing. The test environment is divided into a test group running the complete system, a partially missing control group with the etching spray pressure correction branch removed from the differential impedance regression model, an out-of-range control group with the first safety deviation threshold extended to outside the safety range, and a control group that uses the end physical sampling mode.

[0035] During the test run, the material process feature aggregation unit received the original state voltage input signal containing multi-source harmonic interference. Under the initial material flow conditions, the original state voltage input signal exhibited a non-stationary discrete series with a baseline drift amplitude reaching 14.3 mV. The standardized state feature tensor, containing random electromagnetic disturbance characteristics, was input to the impedance prediction analysis unit. The experimental group used a hierarchical cascaded topology and error hedging operator within the differential impedance regression model to filter out the 14.3 mV random electromagnetic disturbance. The calculated impedance prediction indicator had a measurement dispersion of 0.012 over 100 consecutive sampling points. However, the partially missing control group lacked the collaborative correction of the real-time feature sequence of the etching process spray pressure, and its output impedance prediction indicator experienced amplitude fluctuations during process switching. The signal false overshoot reached 28.4%. Simultaneously, within the parameter nodes defined by the assessment parameter boundaries, the first safety deviation threshold for the out-of-range control group was 0.60 to 1.40. When the impedance prediction indicator factor reached 1.26, the actual interlayer dielectric crosslinking degree of the material at the component level underwent reverse cracking due to subsequent thermal stress overload during pressing, resulting in a 32.6% increase in dielectric loss and the scrapping of work-in-process. Conversely, when the impedance prediction indicator factor decreased to 0.74, the system's control gain entered the nonlinear saturation region of dispersion and tended to flatten, increasing the overall system calculation delay from 12.43 ms to 45.18 ms, causing a delay in production scheduling switching decisions. Therefore, the first safety deviation threshold range was determined to be 0.85 to 1.15 to avoid both the point of increased dielectric loss and the point of increased calculation delay.

[0036] After 48 hours of continuous online comparative verification of the semi-real-world logistics line, the experimental group using the high-frequency, high-speed printed circuit board impedance adaptive control system achieved an order flow scheduling matching rate of 99.53% in online asset restructuring and allocation. The business management decision-making scheduling unit, by rewriting the underlying routing labels of the manufacturing management system, changed stagnant work-in-process assets with impedance dispersion fluctuations in the high-level server motherboard order queue to low-frequency communication backplane orders, achieving an asset residual value recovery rate of 84.21%. In contrast, the control group, which used the original end-point interception management mode, could not reverse-write routing labels or restructure production schedules in real time, resulting in a lower overall asset value recovery rate. The scrap rate of entire boards of materials after batch deviations in work-in-process was 15.34%, and the average waiting time for processes was extended to 2.4 hours due to production scheduling mismatch. The differences in the comprehensive system indicators between the experimental group, the control group, and the control systems of each gradient confirmed the feasibility of the data processing mechanism that translates the discrete characteristics of the underlying physical processing into production scheduling routing control variables in real time in order to avoid the scrapping of work-in-process assets and maintain the continuity of production scheduling decisions in industrial applications. The final operating status of the system showed that the stable control of high-frequency and high-speed signal impedance and the efficient scheduling of customized multi-specification production achieved adaptive engagement balance within a unified data-driven processing framework.

[0037] Example 3: In a customized flexible manufacturing plant for high-frequency, high-speed printed circuit boards, due to the continuous high-load flow of multi-specification work-in-process substrates, and the unpredictable subtle physicochemical fluctuations in copper foil roughness parameters and local residual linewidths within the etching reactor, the multi-layer nonlinear regression network of the manufacturing management system experiences cumulative logical drift in its internal network weights and bias parameters after prolonged uncalibrated operation. This leads to the calculation results of impedance prediction indicator factors deviating from the physical state of the material's process characteristics. This performance degradation of the model caused by long-term operation directly results in false alarms or missed alarms when the system determines the safety deviation threshold range, leading to the loss of good products. Materials are abnormally suspended at the management and scheduling end, or out-of-tolerance materials are not intercepted and downgraded in a timely manner, disrupting the continuity of customized multi-specification scheduling decisions and the economic allocation of losses. To address the aforementioned challenges of model performance degradation caused by long-cycle operation, the impedance prediction analysis unit automatically triggers timeliness assurance and reconstruction procedures in the monitoring and control process. This solution defines a model timeliness deviation degree to characterize the timeliness quality decay of the differential impedance prediction network model itself. To quantify the degree of logical drift of the internal parameters of the differential impedance prediction network model over time, this solution draws on the basic principles of system uncertainty evolution in information theory, addressing the impact of data environment changes over time. An increase in model prediction variance is considered an increase in Shannon information entropy. Therefore, by calculating the statistical distribution dispersion of the residual error between the real-time measured impedance values ​​and the model prediction output values ​​of a specific batch of in-process substrates, an objectively measurable model timeliness deviation is defined. At the initial stage of this process, the data acquisition channel of the hardware-in-the-loop integrated testing platform must maintain a normal enabled connection, and its underlying central processing unit's floating-point throughput must be no less than 10 billion operations per second, serving as the enabling environment functional specification for maintaining the closed-loop computation timing. During the execution of the timeliness assurance and refactoring program, the first step is to read the previous 10 consecutive production batches containing 100... The first step involves analyzing the impedance measurement data stream of a high-frequency, high-speed printed circuit board with 0 sample points, while simultaneously retrieving historical output data from the differential impedance prediction network model for each sample point. The second step involves subtracting the corresponding historical output data from the impedance measurement data stream of each sample point to obtain the single-point residual, and calculating the overall standard deviation of the total single-point residual. The third step involves dividing the overall standard deviation by the nominal span value of the safety deviation threshold interval, and determining the resulting ratio as the model timeliness deviation. This ratio is used for quantification and translation. When the measured overall standard deviation is 0.035 ohms and the nominal span value is 0.30 ohms, the model timeliness deviation output result obtained through the above division operation is 0.117.

[0038] The impedance prediction analysis unit seamlessly integrates the calculated model timeliness deviation into its built-in parameter setting decision loop, executing a process judgment quantification procedure. This loop includes a set of reconstruction triggering criteria based on pure textual quantification deduction logic. The triggering criteria stipulate that when the model timeliness deviation is within the first interval not exceeding the preset reconstruction critical threshold, the system controls the model to continue online impedance prediction while maintaining the existing network weights. However, as process equipment ages and material batch dispersion increases, single-point residual errors continuously expand unidirectionally, causing the model timeliness deviation to cross the preset reconstruction critical threshold of 0.15. At this point, the impedance prediction analysis unit stops the current prediction call and initiates in-situ parameter adaptive calibration reconstruction. In the specific reconstruction calibration action flow, the system introduces a sliding time window mechanism to eliminate and downweight old data in the historical database. Specifically, this involves multiplying the weight coefficients of outdated batch data (more than 24 hours old) by a downstream attenuation operator. Here, the downstream attenuation operator adopts an exponential attenuation form, specifically expressed as the formula... ,in, The number of hours exceeding 24 hours for data storage. The preset attenuation factor constant is limited to a range of 0.05 to 0.12 in this embodiment. The boundary constraint condition of the downlink attenuation operator is that when the excess time... Make the calculated attenuation coefficient When the value is less than the preset cutoff limit of 0.01, the system automatically triggers the hard elimination of historical data, directly clearing the corresponding feature weights to zero. This ensures that the contribution of historical temperature drift data to the current regression gradient correction decreases logarithmically and smoothly over time. At the same time, the system automatically retrieves 100 sets of the latest impedance feature data measured by the line-side precision time-domain reflectometer in the most recent hour as a source of strong constraint boundary input. The system performs gradient correction on the weight matrix of the internal cascaded architecture through the backpropagation algorithm of the multi-layer nonlinear regression network until the prediction variance of the newly trained updated model is reduced to below 0.005. Only then can the updated model be re-input into the online differential impedance prediction stream, thereby achieving the long-term feasibility of the prediction model and decision self-healing.

[0039] This model timeliness guarantee and reconstruction logic, embedded in the data processing architecture, does not require changes to the existing sensor hardware configuration of the production line. By performing periodic calibration and correction on the internal feature weights of the dynamic calculation model at the software management algorithm level, the long-term root mean square error of the regression predictor is successfully controlled within 0.015. This not only eliminates the risk of production scheduling decision chaos caused by the long-term failure of the algorithm black box, but also enables the entire impedance adaptive control system to still produce deterministic, self-consistent management and control decisions that support adaptive routing rewriting of production scheduling when facing long-term physical drift of the manufacturing environment.

[0040] Example 4: When the system is newly installed and deployed on a high-frequency, high-speed printed circuit board flexible manufacturing line, and the initial working condition is that the difference in the surface resin curing shrinkage rate of different batches of copper-clad laminate materials causes a systematic measurement bias in the physical monitoring characteristic data of the line edge, the high-frequency, high-speed printed circuit board impedance adaptive control system, before being put into production, has a set of closed-loop standardized start-up self-test calibration procedures executed by the material process characteristic aggregation unit and the impedance prediction analysis unit. In the specific operation process, the technicians select three sets of calibration test blocks with completely known physical characteristic dimensions and all geometric dimensions in a standard state. The substrate thickness of the calibration test blocks is calibrated to 1.0mm, 1.6mm and 2.4mm, respectively. At the same time, on the integrated test platform, the sampling frequency of the multi-channel high-frequency impedance tester is locked at 100MHz, and the static voltage temperature drift correction coefficient of the data acquisition board is set to the start-up reset reference state, thereby establishing the enabling environment functional specifications required to implement the process parameter calibration process.

[0041] In the flow control of the standardized startup self-test calibration procedure, the material process feature aggregation unit reads each group of calibration test blocks and acquires the initial voltage discrete signal in the sensor path. The impedance prediction and analysis unit calls the differential impedance regression model to translate the initial voltage discrete signal into an initial asset feature matrix containing specific noise characteristics. This is achieved by executing pure text quantization deduction logic based on reverse tracing of known outputs. Specifically, when this pure text quantization deduction logic is executed at the bottom central processing unit, it refers to converting the text-based field process rule constraints into discrete 0-1 digital logic vectors through a one-hot encoding mechanism. These vectors are then used as boundary penalty terms and superimposed on the standard variance loss function formed by the known impedance measured output. Gradient descent is then executed at the floating-point arithmetic unit. During calculation, the partial derivatives of the reconstruction loss function with respect to the weights of each network connection are calculated, thereby adaptively translating the physicochemical rules and textual constraints into the underlying floating-point gradient update step size, directly changing the numerical state of the weight matrix. The initial asset feature matrix is ​​then substituted into the hierarchical cascaded topology network for weight fine-tuning. After continuous calculation and comparison of 50 time-series sampling points, if the variance envelope of the residual fluctuations at each point is stably maintained within a range of less than 0.002, it is determined that the updated model has achieved physical fit with the current physical and chemical properties of the object. Based on this, the system locks the current adaptive calibration gain operator and generates a self-check pass status label, thereby ending the pre-deployment calibration and switching the system to online continuous monitoring stream, thus avoiding erroneous triggering of routing reconstruction decisions caused by initial benchmark drift.

[0042] Example 5: In the customized manufacturing system of high-frequency and high-speed printed circuit boards, in order to avoid the failure of the impedance prediction model due to the differences in the physicochemical properties of multiple batches of copper-clad laminate raw materials and the drift of the etching solution concentration inside the etching reactor, the high-frequency and high-speed printed circuit board impedance adaptive control system automatically triggers a closed offline model reconstruction procedure by the business management decision scheduling unit before the production batch switch. Before the procedure is executed, the statistical value of the discrete distribution of impedance quality generated at the end of the production line in the last 24 hours is retrieved through the data bus as the initial reference benchmark for calibrating the prediction accuracy of the differential impedance regression model. The impedance prediction analysis unit enters the calibration mode and injects three sets of process parameter feature tensors covering the entire process window into the differential impedance regression model according to the preset discretization gradient strategy. These tensors correspond to the absolute lower limit, statistical normal median, and absolute upper limit of the substrate dielectric constant change, respectively.

[0043] During the offline parameter reconstruction process, the impedance prediction analysis unit uses the root mean square error between the predicted output values ​​of each set of tensors and the preset physical impedance reference value as the optimization objective function. It iteratively corrects the weight matrices of each level within the differential impedance regression model using a heuristic search algorithm. When the prediction residuals of the model converge to a physical engineering accuracy boundary of less than 0.005 ohms in the three calibration scenarios (absolute lower limit, normal median, and absolute upper limit), the reconstruction program triggers termination logic and outputs a model parameter verification confirmation signal. The business management decision scheduling unit solidifies the verified model weight parameters into the non-volatile storage area of ​​the local computing unit and simultaneously calculates a set of adaptive deviation correction coefficients for the current specific batch of raw materials based on the current process environment factors. This coefficient uses the ratio of the measured dielectric loss factor of the current batch of copper-clad laminate to the reference loss factor as input weights. It performs real-time compensation calculations on the impedance prediction indicator factor output by the prediction network according to a preset linear proportional relationship. Specifically, the preset linear proportional relationship is expressed as the compensated... The impedance prediction indicator factor is equal to the algebraic product of the original factor and the adaptive deviation correction coefficient. The adaptive deviation correction coefficient is calculated by multiplying the ratio of the measured dielectric loss factor of the current batch of copper-clad laminate to the reference loss factor by a fixed proportional empirical constant. In this embodiment, the empirical constant is 0.88. This linear relationship establishes a proportional transmission chain from the intrinsic high-frequency loss fluctuation of the material to the final impedance deviation factor, ensuring that the physical measurement bias can be eliminated in real time in a deterministic algebraic gain form. Thus, the impedance prediction analysis unit realizes real-time perception of process interference based on the evolution trajectory of the model's timeliness deviation. The system performs a status evaluation on the impedance prediction indicator factor according to the preset route reconstruction threshold. When the evaluation value is outside the safety envelope, the system directly updates the underlying manufacturing route label and automatically diverts products that do not meet the impedance technical specifications of specific high-level orders to the low-frequency communication backplane production queue. This ensures that the high-frequency, high-speed production network maintains the determinism of impedance prediction calculation and the stability of production scheduling decisions even when experiencing batch switching or process equipment aging interference.

[0044] The embodiments of this application have been described above with reference to the accompanying drawings. Unless otherwise specified, the embodiments and features in the embodiments of this application can be combined with each other. This application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit of this application and the scope of protection of this invention, and all of these forms are within the protection scope of this application.

Claims

1. A high-frequency high-speed PCB impedance self-adaptive regulation control system, characterized in that the system include: Material and process characteristic aggregation unit, impedance prediction and analysis unit, and business management decision-making and scheduling unit; The material process feature aggregation unit is connected to the impedance prediction and analysis unit, and the impedance prediction and analysis unit is connected to the business management decision scheduling unit. The material process feature aggregation unit acquires the substrate factory feature data and line edge physical monitoring feature data of the current batch of materials processed for high-frequency and high-speed printed circuit boards, and completes tensor alignment according to the timestamp sequence to assemble and output the state feature tensor. The impedance prediction analysis unit receives the state feature tensor and calculates the impedance prediction indicator factor reflecting the degree of impedance deviation of the current batch of materials through the impedance regression model. When the impedance prediction indicator factor crosses the safety deviation threshold, the business management decision scheduling unit calculates the capital loss carried over to the current node by combining the real-time bill of materials cost data of the current batch of materials. When the capital loss is lower than the low economic diversion threshold, the underlying routing label of the manufacturing management system is automatically rewritten, changing the target order attribute of the current batch of materials from server motherboard order to low frequency communication backplane order, and allocating the current batch of materials to the low frequency printed circuit board production queue.

2. The high-frequency, high-speed PCB impedance adaptive control system according to claim 1, characterized in that, Under the condition that the business management decision-making and scheduling unit calculates the capital loss carried over to the current node, the system controls the compensation timing of sudden disturbances in the sensor network: the business management decision-making and scheduling unit extracts the impedance prediction indicator factor and marks the sudden jump of more than 15% as the transient disturbance. The business management decision scheduling unit sets the sampling timing deviation parameter based on the transient disturbance, modifies the sampling window timing of the next cycle sensor, and calls the historical rolling average of the current batch of materials at the data link end to complete the smoothing filter replacement.

3. The high-frequency high-speed PCB impedance adaptive regulation and control system according to claim 1, characterized in that, When the system is running and data is flowing, the system controls network interruption and downgrades scheduling: when the communication interruption is within 50μs, the business management decision scheduling unit automatically extracts the historical rolling average of the current batch of materials and completes smooth replacement; when the network interruption duration exceeds the 100ms threshold, the business management decision scheduling unit cuts off the continuous prediction state and reduces the dimension to the batch-level risk compliance judgment mode based on the static factory characteristics of raw materials.

4. The high-frequency high-speed PCB impedance self-adaptive regulation and control system according to claim 1, characterized in that, The process of acquiring line-edge physical monitoring feature data by the material process feature aggregation unit includes: the material process feature aggregation unit collects the dielectric layer thickness deviation data and etching linewidth fluctuation data of the material in the current processing batch; the material process feature aggregation unit acquires the copper foil roughness parameters of the material in the current processing batch, and combines the dielectric layer thickness deviation data, etching linewidth fluctuation data and copper foil roughness parameters into the line-edge physical monitoring feature data.

5. The high-frequency high-speed PCB impedance self-adaptive regulation and control system according to claim 1, characterized in that, The process of acquiring substrate factory characteristic data by the material process characteristic aggregation unit includes: the material process characteristic aggregation unit reads the original dielectric constant and original dielectric loss factor of the copper clad laminate of the current batch of materials; the material process characteristic aggregation unit retrieves the factory resin content and fiberglass cloth weaving density of the current batch of materials, and combines the original dielectric constant, original dielectric loss factor, factory resin content and fiberglass cloth weaving density into the substrate factory characteristic data.

6. The high-frequency high-speed PCB impedance self-adaptive regulation and control system according to claim 1, characterized in that, The process of the impedance prediction analysis unit running the impedance regression model includes: the impedance prediction analysis unit inputs the substrate manufacturing characteristic data and the line-side physical monitoring characteristic data into a multi-layer nonlinear regression network; the impedance prediction analysis unit uses the multi-layer nonlinear regression network to determine the nonlinear mapping relationship between the substrate manufacturing characteristic data and the line-side physical monitoring characteristic data and the target impedance, and outputs the impedance prediction indicator factor.

7. The high-frequency high-speed PCB impedance self-adaptive regulation and control system according to claim 1, characterized in that, The process by which the business management decision-making and scheduling unit determines the safety deviation threshold includes: the business management decision-making and scheduling unit establishes a first safety deviation upper limit boundary and a first safety deviation lower limit boundary for the high-level server motherboard order as the safety deviation threshold; when the impedance prediction indicator factor exceeds the first safety deviation upper limit boundary or falls below the first safety deviation lower limit boundary, the business management decision-making and scheduling unit determines that the current batch of materials has generated impedance deviation scrap loss and triggers subsequent cost accounting.

8. The high-frequency high-speed PCB impedance self-adaptive regulation and control system according to claim 1, characterized in that, The process by which the business management decision-making and scheduling unit controls the reorganization and circulation status when the capital loss exceeds the low economic diversion threshold includes: when the capital loss is greater than or equal to the low economic diversion threshold, the business management decision-making and scheduling unit determines that the current batch of materials does not have the value of being downgraded and realized assets; the business management decision-making and scheduling unit synchronously issues a stop circulation instruction to the underlying scheduling route, suspends the circulation queue of the current batch of materials in place at the manufacturing management system end, and generates a special material scrap warning signal at the system end.

9. The high-frequency high-speed PCB impedance self-adaptive regulation and control system according to claim 1, characterized in that, The system also includes a data synchronization module: the data synchronization module is connected to both the material process feature aggregation unit and the business management decision scheduling unit; while the material process feature aggregation unit is acquiring feature data, the data synchronization module collects the real-time process status of work-in-process in the upstream multi-task sequence and transmits the real-time process status synchronously to the business management decision scheduling unit, so that the business management decision scheduling unit can mark the work-in-process asset carry-over amount on the multi-task sequence processing path when calculating the capital loss carried over at the current node.

10. A high-frequency high-speed PCB impedance self-adaptive regulation and control method, which is run in the high-frequency high-speed PCB impedance self-adaptive regulation and control system of claim 1, characterized in that, Includes the following steps: Step 1001: The material process feature aggregation unit acquires the substrate factory feature data and line-side physical monitoring feature data of the current batch of materials, and completes tensor alignment according to the timestamp sequence to assemble and output the state feature tensor. Step 1002: The impedance prediction analysis unit receives the state feature tensor and calculates the impedance prediction indicator factor through the impedance regression model. Step 1003: When the impedance prediction indicator factor crosses the safety deviation threshold, the business management decision scheduling unit calculates the capital loss carried over to the current node by combining the real-time bill of materials cost data of the current batch of materials. When the capital loss is lower than the low economic diversion threshold, the unit automatically rewrites the underlying routing label of the manufacturing management system, changes the target order attribute of the current batch of materials from server motherboard order to low frequency communication backplane order, and allocates the current batch of materials to the low frequency printed circuit board production queue.

Images