A supply chain collaborative management and early warning system for inkjet printing of PC boards

Abstract

This invention relates to the fields of supply chain collaborative management, risk warning, and AI scheduling technology, specifically a supply chain collaborative management and early warning system for PC board inkjet processing. It includes an external supply chain status perception unit for aggregating external status information and generating an external status vector; an internal logistics bottleneck monitoring unit for monitoring the inbound quality control (IQC) channel and generating IQC congestion alarm signals and internal material delay times; a collaborative vulnerability quantification unit for quantifying the collaborative vulnerability index and generating collaborative vulnerability status signals; a bimodal adaptive decision-making unit for generating scheduling strategy instructions; and a resilience strategy execution unit for executing material redundancy actions and resource redundancy actions to reduce the collaborative vulnerability index. This invention solves physical bottlenecks and proactively restores human-machine trust, ensuring the system automatically stabilizes after a crisis is resolved.

Description

Technical Field

[0001] This invention relates to the fields of supply chain collaborative management, risk warning and AI scheduling technology, specifically a supply chain collaborative management and early warning system for PC board inkjet processing. Background Technology

[0002] In complex manufacturing environments such as modern PC board inkjet processing, the stability of the supply chain and the intelligence of production scheduling are crucial.

[0003] Existing AI scheduling systems generally pursue a single goal of efficiency or cost optimization, lacking resilience to external shocks. Furthermore, these systems suffer from fatal blind spots, such as the inability to distinguish between material receipt status and actual production delays caused by congestion in the IQC (Initial Quality Control) inspection channel. This information asymmetry undermines human-machine trust, leading production line managers to adopt self-protective behaviors such as buffering strategies, thereby contaminating the data needed for AI decision-making. This vicious cycle of blind spots and lack of trust can easily trigger systemic collapse. Therefore, how to build a collaborative management system that can bridge the blind spots, quantify and manage human-machine trust, and adaptively switch between efficiency and resilience has become a pressing technical problem to be solved in this field. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a supply chain collaborative management and early warning system for inkjet printing of PC boards. Specifically, the technical solution of this invention includes:

[0005] The external supply chain status sensing unit is used to aggregate external status information and generate an external status vector.

[0006] The internal logistics bottleneck monitoring unit is used to monitor the inbound quality inspection (IQC) channel and generate IQC congestion alarm signals and material internal delay time.

[0007] The collaborative vulnerability quantification unit is used to quantify the collaborative vulnerability index based on the external state vector, the IQC congestion alarm signal, and the internal delay time of the material; the collaborative vulnerability quantification unit is also used to generate a collaborative vulnerability state signal based on the collaborative vulnerability index and a preset risk threshold;

[0008] A bimodal adaptive decision-making unit is used to switch between a conventional optimization mode and a resilience-first mode based on the cooperative vulnerability state signal, and to generate scheduling strategy instructions.

[0009] The resilience strategy execution unit is used to execute material redundancy actions and resource redundancy actions according to the resilience instructions in the scheduling strategy instructions; the execution result of the resilience strategy execution unit is fed back to the collaborative vulnerability quantification unit to reduce the collaborative vulnerability index.

[0010] Preferably, the external supply chain status awareness unit is used to receive and process in real time yield fluctuation data from single-source suppliers, logistics tracking data of short-shelf-life materials, and market intelligence that may trigger asymmetric shocks, in order to generate the external status vector.

[0011] Preferably, the internal logistics bottleneck monitoring unit is used to specifically monitor the inbound quality inspection (IQC) channel, and to monitor the queue length of the IQC channel, the average waiting time of materials, and the instantaneous influx of general materials.

[0012] Preferably, the internal logistics bottleneck monitoring unit is also used for:

[0013] Based on the queue length of the IQC channel, historical average waiting time data, and the instantaneous influx of the general materials, the internal delay time of the materials is estimated by dynamically predicting using statistical or machine learning models.

[0014] Preferably, the collaborative vulnerability quantification unit is used to increase the collaborative vulnerability index through at least one of the following decision logics:

[0015] When the external supply chain status sensing unit shows that the materials have been signed for, and the internal logistics bottleneck monitoring unit issues the IQC congestion alarm signal, it is determined that there is a sensing deviation.

[0016] When the internal delay time of the material exceeds the current safety stock buffer of the production line, or approaches the boundary of a catastrophic shutdown, the risk exposure is determined to be aggravated.

[0017] Monitor the frequency or magnitude of manual inventory correction operations in the manufacturing execution system or enterprise resource planning system; cross-compare the frequency or magnitude with usage data collected by automated equipment; when abnormal discrepancies are identified, determine that a buffer game is in progress.

[0018] Preferably, the process of switching modes in the dual-modal adaptive decision unit is as follows:

[0019] When the collaborative vulnerability status signal is at a safe level, the system enters the normal optimization mode, with the decision objective being to optimize production line utilization and cost.

[0020] When the collaborative vulnerability status signal is at a warning or danger level, the system enters the resilience priority mode, with the decision objective being to reduce the collaborative vulnerability index.

[0021] Preferably, in the resilience-first mode, the bimodal adaptive decision unit dynamically modifies the reward function of the deep reinforcement learning architecture, taking the decrease in the collaborative vulnerability index as the main positive reward, and reducing the weight of the production line utilization rate and cost in the reward function.

[0022] Preferably, when the resilience strategy execution unit receives the resilience instruction, it performs at least one of the following actions:

[0023] The material redundancy action is: issuing an excess of emergency material orders;

[0024] The resource redundancy action is to allocate a dedicated IQC green channel for the urgent material order to bypass congestion.

[0025] Preferably, the collaborative vulnerability quantification unit is also used for:

[0026] Monitor the reduction in internal delay time of the material caused by the IQC green channel;

[0027] The monitoring showed a reduction in the aforementioned buffered game behavior;

[0028] The collaborative vulnerability index is reduced based on the shortened delay time and the reduced buffering game behavior; the collaborative vulnerability state signal is restored to the security level based on the reduction of the collaborative vulnerability index.

[0029] Compared with the prior art, the present invention has the following beneficial effects:

[0030] 1. This system sets up an internal logistics bottleneck monitoring unit to specifically monitor the queue length and general material inflow of the inbound quality inspection channel, dynamically predict the internal delay time of materials, effectively bridge the key perception blind spot between the receipt of materials and their readiness for online processing, and improve the data foundation for decision-making.

[0031] 2. This system innovatively constructs a collaborative vulnerability quantification model, which identifies and quantifies buffer game behavior caused by a lack of trust by monitoring operations such as manual inventory adjustments; this is the first time that implicit human factors risks have been transformed into manageable engineering indicators, preventing the vicious cycle of data pollution.

[0032] 3. This system breaks through the limitations of traditional AI scheduling that focuses solely on efficiency; it establishes a dual-modal adaptive decision-making unit that can dynamically switch between the conventional mode that pursues the optimal cost and the resilience-first mode that ensures survival, based on the real-time collaborative vulnerability index, thus achieving a balance between efficiency and resilience.

[0033] 4. This system constructs a social-technical hybrid closed loop from risk identification and decision-making to execution feedback; in resilient mode, it performs highly visible redundant actions such as opening a dedicated quality inspection channel, which not only solves the physical bottleneck, but also proactively restores human-machine trust, ensuring that the system automatically stabilizes after the crisis is over. Attached Figure Description

[0034] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0035] Figure 1 This is a structural diagram of the system of the present invention. Detailed Implementation

[0036] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0037] Example 1:

[0038] Please see Figure 1 A supply chain collaborative management and early warning system for inkjet printing of PC boards, comprising:

[0039] The external supply chain status sensing unit is used to aggregate external status information and generate an external status vector.

[0040] The internal logistics bottleneck monitoring unit is used to monitor the inbound quality inspection (IQC) channel and generate IQC congestion alarm signals and material internal delay time.

[0041] The collaborative vulnerability quantification unit is used to quantify the collaborative vulnerability index based on the external state vector, the IQC congestion alarm signal, and the internal delay time of the material; the collaborative vulnerability quantification unit is also used to generate a collaborative vulnerability state signal based on the collaborative vulnerability index and a preset risk threshold;

[0042] A bimodal adaptive decision-making unit is used to switch between a conventional optimization mode and a resilience-first mode based on the cooperative vulnerability state signal, and to generate scheduling strategy instructions.

[0043] The resilience strategy execution unit is used to execute material redundancy actions and resource redundancy actions according to the resilience instructions in the scheduling strategy instructions; the execution result of the resilience strategy execution unit is fed back to the collaborative vulnerability quantification unit to reduce the collaborative vulnerability index.

[0044] The system is constructed as a complete closed loop from risk identification to decision switching and then to execution feedback; its design purpose is to solve the risk of systemic collapse caused by the neglect of perception blind spots and human-machine trust in existing AI scheduling systems.

[0045] The core purpose of the external supply chain status sensing unit is to aggregate all status information related to supply chain stability from outside the factory. Its input is multi-source heterogeneous data streams, such as the yield data of supplier production batches and the logistics tracking data of key materials. This unit parses, cleans and timestamps this information to achieve standardized processing, and its output is an external status vector. This vector is transmitted to the collaborative vulnerability quantification unit as a unified representation of the external environment status.

[0046] The internal logistics bottleneck monitoring unit is designed to address the critical blind spot where material receipt does not equate to readiness for production. It is specifically configured to monitor the critical internal bottleneck: the inbound quality control (IQC) channel. The core logic of this unit determines whether external shocks are unintentionally clogging the IQC channel for critical specialty inks. This unit outputs two key signals:

[0047] A qualitative IQC congestion alarm signal; the generation of this qualitative judgment signal is based on the classification threshold set by those skilled in the art according to the physical design capacity of the IQC channel, the statistics of historical average waiting time, and the experience of experts; the generation of this qualitative signal is achieved through a preset rule engine: when the IQC real-time queue length detected by the unit exceeds 80% of its physical design capacity, and the predicted waiting time of key materials, such as special inks, exceeds 2 standard deviations of its historical average waiting time, the unit generates an IQC congestion alarm signal;

[0048] A quantitative internal delay time for materials; this estimate is dynamically predicted by this unit based on real-time monitored IQC queue length, historical average waiting time data, and parameters such as the instantaneous influx of current common materials, through statistical or machine learning models.

[0049] These two signals are sent directly to the co-vulnerability quantization unit as the core input for calculating vulnerability;

[0050] The purpose of the collaborative vulnerability quantification unit is to quantify the safe distance of the system from a shutdown in real time. The collaborative vulnerability index is a key engineering indicator defined in this invention, used to characterize the comprehensive quantitative result of physical bottleneck risk and human-machine trust collapse risk. The unit receives external state vectors, IQC congestion alarm signals, and internal material delay time as inputs. Its processing logic is to comprehensively evaluate these inputs. For example, when signed-out and IQC congestion occur simultaneously, or when the internal delay time exceeds the safety stock buffer, the unit will increase the collaborative vulnerability index.

[0051] The unit further compares this index with a preset risk threshold. It should be understood that the threshold is set by those skilled in the art based on sensitivity analysis of the system failure boundary and backtesting of historical data on the buffer game behavior of production line managers. Specifically, the backtesting process involves inputting historically actual line stoppage event data and production line manager buffer game data within the corresponding time window into a classification model, such as a support vector machine or decision tree, and training it through cross-validation to obtain the index critical value that can provide the earliest and most accurate warning of systemic collapse. This critical value is set as the preset risk threshold. The aim is to ensure that the system provides sufficient reaction time before reaching the irreversible failure boundary. When the comparison result exceeds the threshold, the unit outputs a cooperative vulnerability state signal.

[0052] The bimodal adaptive decision-making unit, as the core decision-making hub of the system, aims to achieve dynamic switching between efficiency and survival. Its input is a collaborative vulnerability state signal, and its decision logic is strictly controlled by this signal: when the signal is safe, the system is in normal optimization mode, and the decision objective is set to pursue the optimal production line utilization rate and cost; when the signal is a warning or danger, the system is forced to switch to a resilience-first mode. The essence of this switching mechanism is to dynamically switch the decision objective from the traditional efficiency maximization to the collapse probability minimization, which is a fundamental breakthrough from the single optimization objective in existing technologies. The output of this unit is the scheduling strategy instruction.

[0053] The resilience strategy execution unit aims to resolutely execute suboptimal strategies designed to restore trust and ensure survival. Its input is the scheduling strategy instruction. When a resilience instruction is received, the unit is activated, bypassing traditional MRP rules, and executes material redundancy and resource redundancy actions.

[0054] The execution results of the resilience strategy execution unit are fed back to the collaborative vulnerability quantification unit; the internal logistics bottleneck monitoring unit will detect that the internal material delay time caused by resource redundancy actions is drastically shortened; at the same time, the collaborative vulnerability quantification unit will detect that the buffer game behavior caused by trust reconstruction is reduced; these two feedbacks are captured by the collaborative vulnerability quantification unit, and its internal logic determines that the risk has been eliminated, resulting in a significant reduction in the collaborative vulnerability index, so that the system automatically recovers to a safe state.

[0055] Through the complete closed-loop collaboration of the above five units, this invention achieves adaptive control from risk identification to decision switching and then to execution feedback. Its essential difference from existing technologies lies in the construction of a social-technical hybrid feedback system. For the first time, it transforms the implicit human factor of human-machine trust into quantifiable and manageable engineering indicators through the monitoring of perception bias and buffer game, and solves the perception blind spot of AI through the internal logistics bottleneck monitoring unit. When systemic risks surge, the system can proactively sacrifice short-term OEE and costs to execute highly visible redundancy strategies. This action restores human-machine trust, breaks the vicious cycle of data pollution, and ensures the long-term survival of the system under extreme shocks.

[0056] Example 2:

[0057] The external supply chain status awareness unit is used to receive and process yield fluctuation data from single-source suppliers, logistics tracking data of short-shelf-life materials, and market intelligence that may trigger asymmetric shocks in real time, in order to generate the external status vector.

[0058] The input data sources for the external supply chain status perception unit have been specifically defined; the purpose is to achieve a deep and predictive perception of supply chain risks, rather than being limited to perceiving states that have already occurred.

[0059] This unit is configured to aggregate and process multiple critical, high-risk heterogeneous data streams in real time; yield fluctuation data from single-source suppliers: this data is used to detect fundamental supply instability; for example, for specialty inks supplied by only one supplier, this unit continuously accesses that supplier's production quality control data; logistics tracking data for short-shelf-life materials: this data is used to detect the physical condition risks of materials in transit; for example, specialty conductive inks are extremely sensitive to timeliness and temperature control, this unit receives their real-time GPS and temperature control data, and pays special attention to the "received" status node; market intelligence that may trigger asymmetric shocks: this data is used to warn of impending internal bottleneck risks; for example, a competitor's surprise release of this intelligence usually indicates panic buying of general-purpose materials in the market; this unit captures such intelligence to warn of the risk of these general-purpose materials crowding out the plant's IQC channel;

[0060] This unit generates an external state vector with richer information dimensions by fusing the above three types of data.

[0061] By integrating the aforementioned specific high-risk data sources, this invention greatly enhances the predictability of the system; it enables the system not only to perceive the logistics status that has already occurred, but also to perceive the impending supply risks and potential internal bottlenecks; this configuration provides a more forward-looking risk input for subsequent collaborative vulnerability quantification units, significantly improving the timeliness and accuracy of early warnings.

[0062] Example 3:

[0063] The internal logistics bottleneck monitoring unit is used to specifically monitor the inbound quality inspection (IQC) channel, and to monitor the queue length of the IQC channel, the average waiting time of materials, and the instantaneous influx of general materials;

[0064] The internal logistics bottleneck monitoring unit is also used for:

[0065] Based on the queue length of the IQC channel, historical average waiting time data, and the instantaneous influx of the general materials, the internal delay time of the materials is estimated by dynamically predicting using statistical or machine learning models.

[0066] The working method of the internal logistics bottleneck monitoring unit is further described in detail, and key technical means to solve the perception blind spot are explained.

[0067] The purpose of this unit is to accurately and dynamically quantify the actual time difference between the receipt of materials and their readiness for production.

[0068] This unit is specifically configured to monitor the Inbound Quality Control (IQC) channel. It acquires various key, dynamically changing parameters in real time by connecting to the IQC's Warehouse Management System (WMS) or on-site sensors. These parameters include the queue length of the IQC channel (i.e., how many batches of materials are currently awaiting inspection); the average waiting time for materials, especially for short-shelf-life materials such as specialty inks; and a key, non-obvious monitoring indicator: the instantaneous influx of general-purpose materials. To further improve the accuracy of the predictive model, this unit is also configured to receive signals from the upstream Warehouse Management System (WMS), such as sorting delay alarms or AGV scheduling congestion alarms. These internal fluctuations in upstream logistics are common leading indicators of instantaneous congestion in the IQC channel, and this model uses these leading indicators as additional feature inputs. The instantaneous influx of general-purpose materials refers to the degree to which non-critical materials, caused by external shocks, crowd out IQC resources.

[0069] The core function of this unit is to dynamically estimate the internal delay time of materials based on the aforementioned monitoring values. This estimation does not use a fixed time standard, but is achieved through a statistical model or machine learning model. The construction of this model relies on regression analysis and training on historical operating data of the IQC channel. The input of this model is the aforementioned real-time queue length, instantaneous inflow of general materials, and historical average waiting time data. The processing logic of this model is that it learns how long a new signed-in material actually needs to be released by IQC and delivered to the production line under different combinations of general material inflow and queue length. Its output, the internal delay time of materials, is a dynamic and forward-looking prediction value, which is transmitted to the collaborative vulnerability quantification unit.

[0070] This invention achieves precise, dynamic, and forward-looking perception of internal bottlenecks; it identifies the instantaneous influx of general-purpose materials as a key trigger for IQC bottlenecks and uses it as a monitoring target; it uses predictive models to estimate internal delay times; this combination enables the system to distinguish between normal busy periods and fatal delays caused by general-purpose material overload for specialty inks, thus providing highly accurate and valuable risk input for the collaborative vulnerability quantification unit. Example

[0071] The collaborative vulnerability quantification unit is used to improve the collaborative vulnerability index through at least one of the following decision logics:

[0072] When the external supply chain status sensing unit shows that the materials have been signed for, and the internal logistics bottleneck monitoring unit issues the IQC congestion alarm signal, it is determined that there is a sensing deviation.

[0073] When the internal delay time of the material exceeds the current safety stock buffer of the production line, or approaches the boundary of a catastrophic shutdown, the risk exposure is determined to be aggravated.

[0074] Monitor the frequency or magnitude of manual inventory correction operations in the manufacturing execution system or enterprise resource planning system; cross-compare the frequency or magnitude with usage data collected by automated equipment; when abnormal discrepancies are identified, determine that a buffer game is in progress.

[0075] The unit provides three specific and operational judgment logics for calculating the collaborative vulnerability index. The unit aims to combine traditionally unmanageable implicit human factors with physical risks for unified quantification. Its inputs are data from external units, internal units, and the MES / ERP system. The processing logic involves executing the following three judgments in parallel; triggering any one of them will lead to an increase in the collaborative vulnerability index. The collaborative vulnerability index is defined as a normalized risk value within the range [0,1]. The unit calculates this index using a weighted summation model, where different risk weights are assigned to judgments regarding perceptual bias, risk exposure, and human factor game theory. For example, human factor game theory may be given the highest weight because it represents the systemic risks of trust collapse and data contamination.

[0076] Regarding the determination of perceptual bias:

[0077] The purpose of this logic is to quantify the damage to human-machine trust caused by information asymmetry. The unit receives two inputs simultaneously: the material receipt status from the external unit and the IQC congestion alarm signal from the internal unit. When both signals are true, the system determines that there is a fatal perception bias. This bias is considered a major disruption to human-machine collaboration, and the system significantly increases the collaboration vulnerability index accordingly.

[0078] Regarding the determination of risk exposure:

[0079] The purpose of this logic is to quantify the imminent nature of physical risks. The unit receives two inputs: the internal material delay time, which is a dynamic estimate from the internal unit, and the current safety stock buffer of the production line from the ERP system. When the unit determines that the 8-hour delay time has significantly exceeded the 6-hour buffer, or when the delay time is approaching the catastrophic shutdown boundary, the system determines that the risk exposure has intensified and increases the collaborative vulnerability index.

[0080] Regarding the determination of human-factor game theory:

[0081] The purpose of this logic is to quantify data contamination behavior resulting from a breakdown of trust. This unit is configured to proactively monitor buffer game behavior by production line managers in MES or ERP systems. Buffer game refers to self-protective behaviors taken by production line managers to cope with perceived risks, leading to distortion of AI input data. This unit continuously monitors the frequency or magnitude of manual inventory correction operations and simultaneously cross-compares these manual correction frequencies or magnitudes with usage data collected by automated equipment. Once significant and abnormal discrepancies are identified, the system determines that buffer game is occurring. To improve the robustness of monitoring, this unit is also configured to monitor other suspicious digital game signals, such as the frequency of non-standard material requisition requests or the number of times emergency work orders are manually created. This unit comprehensively compares the differences between these signals and usage data collected by automated equipment to more comprehensively identify buffer game behavior caused by a lack of trust. This behavior indicates that trust has collapsed and the AI ​​input data has been contaminated, and the system accordingly drastically increases the collaborative vulnerability index.

[0082] This invention creatively constructs a multi-dimensional collaborative vulnerability quantification model; it combines traditionally unmanageable implicit human factors with quantifiable physical risks; this quantification method enables the system to not only perceive physical bottlenecks, but also the dynamic changes in human-machine trust, thus providing a solid and quantifiable basis for subsequent resilience-first decisions.

[0083] Example 5:

[0084] The process of switching modes in the dual-modal adaptive decision unit is as follows:

[0085] When the collaborative vulnerability status signal is at a safe level, the system enters the normal optimization mode, with the decision objective being to optimize production line utilization and cost.

[0086] When the collaborative vulnerability status signal is at the warning or danger level, the resilience priority mode is entered, with the decision objective being to reduce the collaborative vulnerability index;

[0087] In the resilience-first mode, the bimodal adaptive decision unit dynamically modifies the reward function of the deep reinforcement learning architecture, taking the decrease in the collaborative vulnerability index as the main positive reward, and reducing the weight of the production line utilization rate and cost in the reward function.

[0088] The core working mechanism of the bimodal adaptive decision unit is described in detail, especially when it is implemented based on a deep reinforcement learning architecture; a clear and engineerable mechanism is provided to achieve a strategic shift from pursuing efficiency to pursuing survival.

[0089] The decision objective function or reward function of this unit is strictly controlled by the collaborative vulnerability state signal from the collaborative vulnerability quantification unit:

[0090] When the input collaborative vulnerability status signal is at a safe level, the system activates the regular optimization mode. In this mode, its decision objective is to optimize production line utilization and reduce overall costs. For example, it generates regular optimization instructions to maintain a 6-hour safety stock at the limit.

[0091] When the input collaborative vulnerability status signal is triggered as a warning or danger level, the system forcibly switches to resilience priority mode; in this mode, its decision objective is dynamically modified to: reduce the collaborative vulnerability index as the primary decision objective;

[0092] In the DRL architecture, this dynamic modification of decision objectives has a sophisticated and engineerable implementation method, namely the vulnerability-decision coupling control mechanism.

[0093] In the resilience-first mode, the state signal of the danger level will serve as a key control parameter, triggering dynamic switching of the reward function;

[0094] The specific switching logic is as follows: the system sets the decrease in the collaborative vulnerability index as the main positive reward for the DRL agent; at the same time, it significantly reduces the weight of production line utilization and overall cost in the reward function; production line utilization and cost are even regarded as the price that must be paid to achieve the primary goal of reducing the collaborative vulnerability index.

[0095] This modification forces the DRL policy network to search for action sequences that can reduce the cooperative vulnerability index most quickly and effectively at all costs, and output them as scheduling policy instructions.

[0096] This invention provides a clear and engineerable mechanism to achieve a strategic shift from pursuing efficiency to pursuing survival. By dynamically modifying the reward function, this invention provides a concrete DRL implementation path for the complex and nonlinear decision of sacrificing short-term OEE / cost in exchange for long-term survival. This enables AI systems to overcome the limitations of traditional optimization algorithms and to proactively execute antifragile redundancy strategies when a systemic risk of collapse is detected.

[0097] Example 6:

[0098] Upon receiving the resilience instruction, the resilience strategy execution unit performs at least one of the following actions:

[0099] The material redundancy action is: issuing an excess of emergency material orders;

[0100] The resource redundancy action is to allocate a dedicated IQC green channel for the urgent material order to bypass congestion;

[0101] The collaborative vulnerability quantification unit is also used for:

[0102] Monitor the reduction in internal delay time of the material caused by the IQC green channel;

[0103] The monitoring showed a reduction in the aforementioned buffered game behavior;

[0104] The collaborative vulnerability index is reduced based on the shortened delay time and the reduced buffering game behavior; the collaborative vulnerability state signal is restored to the security level based on the reduction of the collaborative vulnerability index.

[0105] The implementation of the resilience-feedback closed loop is described in detail, that is, how the system can automatically recover to a safe state by executing specific redundant actions and monitoring the effects of these actions;

[0106] When the resilience strategy execution unit receives a resilience instruction from the decision-making unit, it is configured to bypass the traditional MRP rigid rules and perform the following two highly visible redundant actions:

[0107] Material redundancy action, that is, the unit immediately places an emergency material order that exceeds the standard MRP economic batch logic, even if it causes a temporary surge in procurement costs;

[0108] Resource redundancy action, that is, the unit simultaneously and proactively increases IQC resource redundancy, forcibly allocating a dedicated IQC green channel for the aforementioned urgent material orders, so that they can bypass the congestion caused by general materials;

[0109] The system forms a feedback loop; the collaborative vulnerability quantification unit continues to operate to monitor the reverse effects of the aforementioned high-visibility actions. This monitoring includes two dimensions:

[0110] This unit, through input from the internal logistics bottleneck monitoring unit, detected that the internal delay time of the materials was drastically shortened due to the activation of the IQC green channel; this directly reduced the risk of judgment logic two.

[0111] The production line manager clearly perceived that the AI ​​system not only recognized the IQC bottleneck but also took strong measures to resolve it; this sense of being seen and resolved rebuilt the production line manager's trust in the AI; the rebuilding of trust, in turn, inhibited the production line manager's buffering game behavior; the co-vulnerability quantification unit detected a reduction in the buffering game behavior.

[0112] Based on the reduction in delay time and the decrease in buffering game behavior, the collaborative vulnerability quantification unit determines through its internal algorithm that the system risk has significantly decreased, thereby reducing the collaborative vulnerability index; the collaborative vulnerability state signal also recovers to the security level based on the reduction in the collaborative vulnerability index; this security signal is fed back to the bimodal adaptive decision-making unit, causing it to automatically switch back to the conventional optimization mode. Thus, a complete closed loop of risk response - trust repair - state stabilization is completed.

[0113] This demonstrates a complete socio-technical hybrid closed loop. The green channel is a purely technical solution used to address physical bottlenecks. The combination of emergency orders and the green channel is a social solution, whose high visibility essentially signals to human managers that the system has perceived the risk and is strongly intervening, aiming to restore trust. This allows the system to confirm in a closed loop that trust has been restored and that the physical risk has been eliminated. This execution-perception-confirmation closed loop ensures that the system can automatically and safely revert to an efficient, routine optimization mode after the crisis is resolved, avoiding unnecessary long-term operation of the system in a resilient mode and achieving a dynamic balance between efficiency and survival.

[0114] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A supply chain collaborative management and early warning system for PC board inkjet processing, characterized in that, The method comprises the following steps: An external supply chain state perception unit is used to aggregate external state information and generate an external state vector; An internal logistics bottleneck monitoring unit is used to monitor an incoming quality control (IQC) channel and generate an IQC congestion alarm signal and an internal material delay time; A collaborative vulnerability quantification unit is used to quantify a collaborative vulnerability index according to the external state vector, the IQC congestion alarm signal and the internal material delay time; the collaborative vulnerability quantification unit is also used to generate a collaborative vulnerability state signal based on the collaborative vulnerability index and a preset risk threshold; A dual-mode adaptive decision unit is used to switch between a regular optimization mode and a resilience priority mode according to the collaborative vulnerability state signal and generate a scheduling strategy instruction; A resilience strategy execution unit is used to execute material redundancy actions and resource redundancy actions according to resilience instructions in the scheduling strategy instruction; the execution result of the resilience strategy execution unit is fed back to the collaborative vulnerability quantification unit to reduce the collaborative vulnerability index; The collaborative vulnerability quantification unit is used to improve the collaborative vulnerability index through at least one of the following decision logics: When the external supply chain state perception unit shows that a material has been signed for, and the internal logistics bottleneck monitoring unit issues the IQC congestion alarm signal, it is determined that there is a perception deviation; When the internal material delay time exceeds the current safety stock buffer of a production line or approaches a disastrous stop-line boundary, it is determined that the risk exposure is aggravated; The frequency or amplitude of manual inventory correction operations in a manufacturing execution system or an enterprise resource planning system is monitored; the frequency or amplitude is cross-compared with usage data collected by an automated device; when an abnormal difference is identified, it is determined that a buffer game is occurring.

2. The supply chain collaborative management and early warning system for PC board inkjet processing according to claim 1, characterized in that, The external supply chain state perception unit is used to receive and process yield fluctuation data of a single-source supplier, logistics tracking data of a short-shelf-life material, and market intelligence that may cause an asymmetric impact in real time to generate the external state vector.

3. The supply chain collaborative management and early warning system for PC board inkjet processing according to claim 1, characterized in that, The internal logistics bottleneck monitoring unit is used to monitor the IQC channel and monitor the queue length of the IQC channel, the average waiting time of a material, and the instantaneous inflow of a general material.

4. The supply chain collaborative management and early warning system for PC board inkjet processing according to claim 3, characterized in that, The internal logistics bottleneck monitoring unit is also used to: Based on the queue length of the IQC channel, historical average waiting time data and the instantaneous inflow of the general material, a statistical model or a machine learning model is used for dynamic prediction to estimate the internal material delay time.

5. The supply chain collaborative management and early warning system for PC board inkjet processing according to claim 1, characterized in that, The process of switching modes of the dual-mode adaptive decision unit is as follows: When the collaborative vulnerability state signal is at a safe level, the regular optimization mode is entered to optimize the production line utilization rate and cost as the decision target; When the collaborative vulnerability state signal is at a pre-warning or danger level, the resilience priority mode is entered to reduce the collaborative vulnerability index as the decision target.

6. The supply chain collaborative management and early warning system for PC board inkjet processing according to claim 5, characterized in that, In the resilience priority mode, the dual-mode adaptive decision unit dynamically modifies the reward function of the deep reinforcement learning architecture, taking the decrease of the collaborative vulnerability index as the main positive reward, and reducing the weight of the line operation rate and cost in the reward function.

7. The supply chain collaborative management and early warning system for PC board inkjet processing according to claim 1, characterized in that, The resilience policy execution unit, upon receiving the resilience instruction, performs at least one of the following actions: The material redundancy action is to place an emergency material order in excess; The resource redundancy action is to allocate a dedicated IQC green channel for the emergency material order to bypass the congestion.

8. The supply chain collaborative management and early warning system for PC board inkjet processing according to claim 7, characterized in that, The collaborative vulnerability quantification unit is also used to: Monitor the shortening of the internal delay time of the material due to the IQC green channel; Monitor the reduction of the buffer game behavior; And according to the shortening of the delay time and the reduction of the buffer game behavior, reduce the collaborative vulnerability index; the collaborative vulnerability state signal is restored to the safety level based on the reduction of the collaborative vulnerability index.

Images