System and method for optimal scheduling of production and resources in a plurality of blow molding machines

Abstract

The application discloses a kind of multi-blowing machine collaborative production and resource optimal scheduling system and method, it is related to industrial manufacturing technical field.The method includes: obtaining global scheduling information including order, resource and environmental parameter;Real-time acquisition of the parison temperature field and material rheological property of each blowing machine and other multivariate working condition information;Based on the above information, integrated decision is made through collaborative scheduling optimization model, and collaborative scheduling instruction set containing production task allocation and process parameter instruction adapted to the real-time working condition of each machine is generated;Finally, it is issued and driven to execute blowing machine.The system corresponds to include global information perception, multivariate working condition acquisition, collaborative scheduling decision and instruction distribution execution and other modules.The application realizes integrated global optimization of task, process and resource by deeply integrating core process state perception into scheduling decision, effectively improves the comprehensive performance, quality consistency and dynamic adaptability of multi-machine collaborative production.

Description

Technical Field

[0001] This invention relates to the field of industrial manufacturing and automation control technology, and in particular to a system and method for collaborative production and optimal resource scheduling of multiple blow molding machines. Background Technology

[0002] In modern plastics manufacturing, blow molding machines are core equipment. As production scales up, multiple blow molding machines are often deployed and operate simultaneously in a workshop. How to efficiently and economically coordinate these machines and optimize the allocation of production tasks and the utilization of resources (such as energy, molds, and raw materials) becomes the key to improving overall production efficiency and effectiveness.

[0003] Currently, there are some related technologies for multi-machine collaboration. For example, patent document CN120572708A proposes an energy utilization and recovery system for injection molding workshops based on multi-machine bus interconnection. Its core is to realize direct energy recovery and utilization between multiple injection molding machines through hardware interconnection. Its scheduling objective is singular, mainly focusing on optimizing the timing of equipment actions around energy matching, without comprehensively considering multi-dimensional factors such as production orders, real-time process status of equipment, and mold life for global production scheduling.

[0004] For example, patent document CN120962948A discloses a centralized material supply system for a central material supply system. This system features a multi-machine collaborative management module, capable of task allocation and material distribution based on equipment status and order priority. However, the core of this solution revolves around scheduling the logistics link of "material supply," with decisions based primarily on the equipment's on / off status and material requirements. It fails to deeply perceive and utilize core process status information affecting blow molding quality (such as parison temperature field uniformity and material flowability). Therefore, its scheduling scheme cannot adapt to the actual process capacity of each machine, making it difficult to achieve true global resource optimization while ensuring consistent product quality across multiple workstations.

[0005] In the process of realizing this invention, the inventors discovered that the prior art has at least the following problems: existing multi-machine scheduling schemes mostly start from a single dimension (such as energy, logistics) or only consider the macroscopic state of the equipment, lacking the perception and utilization of the real-time and multi-variable state of the core process of the blow molding machine, resulting in the scheduling scheme being disconnected from the actual process capacity of the equipment, making it impossible to achieve integrated collaborative decision-making on production task allocation, process parameter optimization and multi-dimensional resource constraints (orders, energy, equipment health), and making it difficult to achieve optimal comprehensive efficiency in complex dynamic production environments.

[0006] The embodiments of the present invention are improvements made to solve the above problems. Summary of the Invention

[0007] The purpose of this invention is to provide a collaborative production and optimal resource scheduling system and method for multiple blow molding machines. By implementing this invention, the problem of existing scheduling schemes being disconnected from actual production due to a lack of awareness of the real-time process status of equipment, and thus failing to achieve global integrated collaborative optimization of production tasks, process parameters, and multi-dimensional resources can be solved.

[0008] To achieve the aforementioned objective, in a first aspect, embodiments of the present invention provide a method for collaborative production and optimal resource scheduling of multiple blow molding machines, applied to a production system including at least two blow molding machines, comprising: Obtain global scheduling information, which includes at least the set of production orders to be processed, the status of available production resources, and external environment parameters. Real-time acquisition of multivariate operating condition information for each blow molding machine. The multivariate operating condition information includes at least the parison temperature field information and material rheological property information within the current production cycle of the corresponding blow molding machine, in order to characterize the real-time process status that directly affects the blow molding quality. Based on global scheduling information and real-time acquired multivariate operating condition information, an integrated decision is made through a preset collaborative scheduling optimization model to generate a collaborative scheduling instruction set. The collaborative scheduling optimization model is configured to: under the premise of meeting production resource and order delivery constraints, take multivariate operating condition information as the key input, optimize the production task allocation and process parameter combination that can improve the overall efficiency of the system and adapt to the real-time process status of each blow molding machine. The collaborative scheduling instruction set includes: production task instructions assigned to each blow molding machine, and process parameter instructions matched for each production task, which are dynamically determined based on the current multivariate operating conditions of the corresponding blow molding machine. The collaborative scheduling instruction set is sent to the corresponding blow molding machine for execution, so as to drive the at least two blow molding machines to carry out collaborative production.

[0009] Preferably, based on global scheduling information and real-time acquired multivariate operating condition information, an integrated decision is made through a pre-defined collaborative scheduling optimization model, specifically including: With the goal of optimizing the overall system performance index, an optimization problem model is constructed that includes production resource constraints, process feasibility constraints, and order delivery constraints. Among them, the process feasibility constraint is the range of process parameters that each blow molding machine can safely and stably execute under the current working conditions, determined based on multivariate working condition information. The optimization problem model is solved using an intelligent optimization algorithm, and the cooperative scheduling instruction set is output.

[0010] Preferably, the overall system performance index is a multi-objective weighted function that takes into account the on-time delivery rate of orders, the overall utilization rate of equipment, the total energy consumption of production, and the consistency of product quality in the same batch.

[0011] Preferably, the process parameter instructions matched for each production task include at least the blowing pressure timing instruction; the blowing pressure timing instruction is dynamically generated by an adaptive control model based on the preform temperature field information and material rheological properties information, and is used to drive the blowing head of the corresponding blow molding machine to perform synchronous blowing action.

[0012] Preferably, the adaptive control model adopts a hybrid architecture that combines feedforward compensation and feedback correction. It is configured to adjust the pressure curve parameters and action timing start points of each blowing head according to the billet temperature field information and material rheological properties information, so as to drive all blowing heads to act synchronously according to a unified target time reference.

[0013] Preferably, the collaborative scheduling optimization model is a decision model based on deep reinforcement learning, which obtains scheduling strategies through interactive learning with the environment; the state space of the environment includes the global scheduling information and historical and real-time multivariate operating condition information, the action space is the possible task allocation and process parameter combination, and the reward function is designed based on the system comprehensive performance index.

[0014] Secondly, a collaborative production and optimal resource scheduling system for implementing any of the above methods is also provided, deployed in a production workshop containing at least two blow molding machines, the system comprising: The global information perception module is configured to acquire the global scheduling information; The multivariable operating condition acquisition module is connected to each blow molding machine and is configured to collect multivariable operating condition information of each blow molding machine in real time. The collaborative scheduling decision engine, which is communicatively connected to the global information perception module and the multivariate operating condition acquisition module, is configured to: run the collaborative scheduling optimization model and generate a collaborative scheduling instruction set based on the input global scheduling information and multivariate operating condition information; The instruction distribution and execution module communicates with the collaborative scheduling decision engine and the control systems of each blow molding machine, and is configured to issue collaborative scheduling instruction sets and drive the blow molding machines to execute them.

[0015] Preferably, the multivariable operating condition acquisition module includes modules deployed on each blow molding machine: Infrared thermal imager array for non-contact detection of the surface temperature field of a blank; Melt pressure sensor and screw speed sensor are used to obtain the raw data required to calculate the rheological properties of materials; The collaborative scheduling decision engine or the local controller of the blow molding machine is configured to calculate the apparent viscosity of the material based on the melt pressure and screw speed.

[0016] Preferably, the system further includes a digital twin simulation module; the digital twin simulation module performs forward-looking simulation and performance evaluation of the scheduling scheme based on the collaborative scheduling instruction set, equipment model and historical data, and feeds the evaluation results back to the collaborative scheduling decision engine for instruction correction or model learning.

[0017] Thirdly, a production workshop is provided, including at least two blow molding machines and the aforementioned collaborative production and optimal resource scheduling system.

[0018] Compared with the prior art, one of the above technical solutions has the following advantages or beneficial effects: 1. Achieved deep integration of process perception and scheduling: By collecting core process variables that directly affect product quality, such as "preform temperature field" and "material rheological properties" in real time, and using them as key inputs for scheduling decisions, the scheduling scheme can closely match the actual process capacity of each piece of equipment, fundamentally solving the problem of disconnect between traditional scheduling and the production process.

[0019] 2. Integrated collaborative optimization was achieved: A collaborative scheduling optimization model with "multi-variable working condition information" as the link was constructed. Production task allocation, dynamic optimization of process parameters, and multi-dimensional resource constraints (orders, energy, molds) were incorporated into a unified framework for solution. This enabled end-to-end integrated decision-making from workshop-level scheduling to equipment-level execution, improving the global optimality of the system.

[0020] 3. Significantly improved overall efficiency: By considering multiple objectives such as "on-time order delivery rate", "equipment utilization rate", "total production energy consumption" and "product quality consistency", the scheduling results are not only more efficient, but also improved in terms of energy consumption and quality stability, thereby enhancing the flexibility and competitiveness of the production system.

[0021] 4. Possesses dynamic adaptation and self-learning capabilities: By introducing intelligent algorithms such as reinforcement learning, or combining with digital twin simulation modules, the system can cope with dynamic disturbances such as equipment failures and order insertions, and continuously learn and optimize scheduling strategies from production data, possessing good robustness and evolutionary capabilities.

[0022] Furthermore, the above summary does not enumerate all the features required for embodiments of the present invention, and other combinations of these feature groups may also constitute embodiments of the present invention. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention or the background art, the accompanying drawings used in the embodiments of the present invention or the background art will be described below.

[0024] Figure 1This is a schematic diagram of the architecture of a multi-blow molding machine collaborative production and optimal resource scheduling system provided in an embodiment of the present invention.

[0025] Figure 2 This is a flowchart of a method for collaborative production and optimal resource scheduling of multiple blow molding machines provided in an embodiment of the present invention.

[0026] Figure 3 This is a schematic diagram of the internal processing logic of the collaborative scheduling decision engine provided in an embodiment of the present invention. Detailed Implementation

[0027] To make the technical means, creative features, objectives and effects of the embodiments of the present invention easier to understand, the embodiments of the present invention are further described below in conjunction with the figures and specific embodiments. It should be understood that the specific embodiments described herein are merely for explaining the embodiments of the present invention and are not intended to limit the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0028] To better understand the embodiments of the present invention, please refer to Figure 1 As shown, this illustrates the architecture of a collaborative production and optimal resource scheduling system 100 provided in an embodiment of the present invention. The system 100 is deployed in a production workshop containing multiple blow molding machines (as shown in the figure, blow molding machines 200a, 200b, ..., 200n), and mainly includes: a global information perception module 110, a multivariate working condition acquisition module 120, a collaborative scheduling decision engine 130, an instruction distribution and execution module 140, and an optional digital twin simulation module 150.

[0029] The global information perception module 110 is used to acquire global scheduling information. This information can come from a Manufacturing Execution System (MES), an Enterprise Resource Planning (ERP) system, or a manual input interface, and includes at least: a set of production orders to be processed (such as product model, quantity, delivery date, and priority), the status of available production resources (such as the availability and lifespan of each mold, raw material inventory, and real-time energy prices), and external environmental parameters (such as workshop temperature and humidity). In one feasible implementation, this module receives order data and resource status data from the MES system through an application programming interface (API).

[0030] The multivariable operating condition acquisition module 120 is communicatively connected to each blow molding machine 200 to collect multivariable operating condition information characterizing its core process state in real time. In one feasible embodiment, the module 120 includes a sensor group deployed on each blow molding machine: an infrared thermal imager array 121 for detecting and generating a two-dimensional temperature field distribution image of the preform surface (i.e., preform temperature field information); a melt pressure sensor 122 and a screw speed sensor 123 for reading real-time data of melt pressure and screw speed during plasticization, based on which the apparent viscosity of the material (i.e., material rheological property information) can be calculated. In this embodiment, the calculation task can be completed by the blow molding machine's local controller or the cloud / edge collaborative scheduling decision engine 130.

[0031] The collaborative scheduling decision engine 130 is the core of the system. It acquires global scheduling information from module 110 and real-time multivariate operating condition information from module 120, runs the preset collaborative scheduling optimization model 131, makes integrated decisions, and generates a collaborative scheduling instruction set.

[0032] In one feasible implementation, see Figure 3 As shown, the internal processing logic of the decision engine 130 includes: Constraint Construction Unit 1311: Based on global scheduling information, it constructs production resource constraints (e.g., mold A can only be used by machine 1 or 2) and order delivery constraints (e.g., order X must be completed before 18:00 today). Crucially, it constructs process feasibility constraints based on real-time multivariate operating condition information. For example, if the current upper part temperature of the blank in machine 200a is too low (determined through temperature field information), then its feasible upper limit for blowing pressure needs to be lowered, and the blowing start point needs to be slightly delayed, forming a feasible domain for process parameters.

[0033] Optimization solution unit 1312: With the goal of maximizing the overall system efficiency index (such as formula (1): overall efficiency = 0.4 * order on-time rate + 0.3 * equipment OEE - 0.2 * unit energy consumption + 0.1 * quality consistency coefficient), the above constraints are integrated into a mixed integer programming problem. Subsequently, the problem is solved using a genetic algorithm (as a type of intelligent optimization algorithm), and a set of Pareto optimal or satisfactory solutions are output, namely, the collaborative scheduling instruction set. This instruction set specifies that "order O1 is assigned to machine 200a and process parameter combination P1 (including a specific blowing pressure curve)" and "order O2 is assigned to machine 200b and process parameter combination P2" etc.

[0034] In this embodiment, due to the need to handle discrete (task allocation) and continuous (process parameter) variables and achieve multi-objective optimization, which requires a high degree of creativity, the optimization solution unit 1312 adopts a parallel intelligent optimization method—genetic algorithm. Through this approach, the system achieves integrated task and process collaborative scheduling based on real-time process status.

[0035] The instruction distribution and execution module 140 receives the collaborative scheduling instruction set from the engine 130, parses and converts it into specific instructions (such as G-codes and PLC control instructions) that can be recognized by the control systems of each blow molding machine, and distributes them through the industrial network. After receiving the instructions, the blow molding machine 200a's local controller (or integrated adaptive blowing control unit) will execute production according to the process parameters (such as blowing pressure timing instructions) in the instructions. In this embodiment, through the above closed loop, the system realizes collaborative production from cloud decision-making to edge execution, improving the accuracy of multi-machine production collaboration and resource utilization efficiency.

[0036] In another embodiment, the collaborative scheduling optimization model 131 can be implemented using a decision model based on deep reinforcement learning, which is particularly suitable for scenarios with complex production environments and frequent dynamic disturbances. This embodiment uses the Proximal Policy Optimization (PPO) algorithm as an example for illustration, but those skilled in the art should understand that other deep reinforcement learning algorithms (such as Deep Deterministic Policy Gradient (DDPG), Asynchronous Advantage Actor-Commentator (A3C), etc.) can also be used to implement this model.

[0037] 1. Model framework and element definition: The reinforcement learning model learns by interacting with an environment. In this application, this environment can be a high-fidelity production simulation system built by the digital twin simulation module 150, or it can learn online in an actual production system.

[0038] Agent: The reinforcement learning model running in the collaborative scheduling decision engine 130.

[0039] State (s_t): At each decision time t (such as when a new order arrives or when the equipment status changes significantly), state s_t is a high-dimensional vector that includes at least: Global scheduling information section: characteristics of all incomplete orders (such as type, remaining quantity, remaining delivery time, priority).

[0040] Multivariable operating condition information section: the parison temperature field uniformity index (TU_j), material apparent viscosity (V_j), and their changing trends for all blow molding machines in the most recent cycle.

[0041] Resources and Environment Section: Availability of each mold, current and projected future energy prices, and workshop calendar time.

[0042] Action (a_t): The action a_t output by the agent is a draft of a complete set of cooperative scheduling instructions. For ease of model processing, it can be designed as a structured output: Discrete action part: a two-dimensional matrix of order number × machine number, generated using the Gumbel-Softmax trick or Pointer Network, indicating which machine each order is assigned to (or "not assigned").

[0043] Continuous action section: A vector that outputs a set of normalized process parameter adjustments for each assigned "order-machine" pair, such as the offset of the blowing pressure reference value and the offset of the blowing start time.

[0044] Reward (r_t): The reward function guides the model's learning objective. The reward function r_t is designed as the instantaneous increment of the system's overall performance index or a negative deviation from the ideal value. For example:

[0045] Wherein, ΔT_{On Time} is the expected reduction in order delays due to this decision; ΔU_{Utilization} is the improvement in equipment utilization; ΔC_{Energy Consumption} is the change in energy consumption costs; ΔQ_{Quality} is the change in quality score estimated based on the proximity of the allocated process parameters to the ideal window; and P_{Penalty} is the large negative reward for violating hard constraints (such as mold conflicts).

[0046] 2. Model Structure: The model can adopt an actor-critic architecture.

[0047] Actor Network: A deep neural network that takes a state s_t as input and outputs the probability distribution (for the discrete part) and specific values ​​(for the continuous part) of actions a_t. Its parameters determine the scheduling policy π(a_t|s_t).

[0048] Critic Network: Another deep neural network that takes a state s_t as input and evaluates the long-term expected cumulative reward of that state (the state value function V(s_t)) to guide the updates of the actor network.

[0049] 3. Training process (taking PPO as an example): Data collection: In the digital twin environment, the agent of the current policy interacts with the simulation environment to generate a large amount of trajectory data (s_t, a_t, r_t, s_{t+1}).

[0050] Advantage estimation: Using methods such as generalized advantage estimation (GAE), the advantage value A_t of each action is calculated based on the commentator network output V(s) and the actual reward r obtained, representing the degree of superiority or inferiority of the action relative to the average level.

[0051] Policy Update (PPO Core): The actor network parameters are updated by maximizing the pruning objective function of PPO.

[0052] Where ratio_t = π_θ(a_t|s_t) / π_{θ_old}(a_t|s_t) is the probability ratio of the new and old policies. The clip operation prevents excessive policy changes in a single update, ensuring training stability.

[0053] Value function update: The critic network parameters are updated by minimizing the mean squared error between the critic network's predicted value V(s_t) and the actual reward.

[0054] Iteration: Repeat the above steps until the policy converges, that is, a high cumulative reward can be stably obtained in the simulation environment.

[0055] 4. Application Example: Assume the trained model has been deployed. When the system status s_t shows: Unit 1 (M1) has an excellent temperature field (TU_1=0.98), Unit 2 (M2) has a slight hot spot (HS_2=True), and a high-priority urgent order arrives, the model (actor network) might output action a_t based on s_t: assign the urgent order to M1 and output a set of positive process parameters for M1 (high pressure, on-time air blowing); output a set of more conservative process parameters for the existing regular order on M2 (slightly reduced pressure, slightly delayed air blowing) to compensate for its hot spot. The critic network evaluates that this decision will yield a high expected reward (because it ensures the quality and delivery of the urgent order). Subsequently, the instruction distribution and execution module 140 parses this a_t into specific instructions and issues them.

[0056] In this embodiment, a deep reinforcement learning model is adopted, which enables the system to learn complex and near-optimal scheduling strategies automatically from the interaction with the (simulated or real) environment without relying on precise manual modeling. It is particularly good at handling complex production optimization problems with randomness, dynamic disturbances and high-dimensional state spaces, demonstrating excellent adaptability and self-learning ability.

[0057] In another feasible implementation, the system further includes a digital twin simulation module 150. This module rapidly simulates the scheduling instruction set generated by the collaborative scheduling decision engine 130 based on high-fidelity equipment models, process models, and historical data. It predicts production cycle time, energy consumption, and expected quality after executing the plan, and feeds back the simulation evaluation results to the decision engine 130. The decision engine 130 can then fine-tune the plan before issuing instructions (instruction correction) or use the simulation data as samples to train an optimization / learning model (model learning). In this embodiment, the introduction of the digital twin module reduces trial-and-error costs and improves the predictability and reliability of the scheduling plan.

[0058] See Figure 2 As shown, the method for collaborative production and optimal resource scheduling of multiple blow molding machines executed by the above system includes the following steps: S210: System initialization, connecting and synchronizing all data sources and devices.

[0059] S220: Obtain global scheduling information.

[0060] S230: Real-time acquisition of multivariable operating condition information for each blow molding machine.

[0061] S240: Based on the information from S220 and S230, an integrated decision is made through a collaborative scheduling optimization model to generate a collaborative scheduling instruction set.

[0062] S250: Sends the collaborative scheduling instruction set to the corresponding blow molding machine.

[0063] S260: The blow molding machine executes instructions to complete collaborative production.

[0064] S270: Collect production execution result data (such as actual completion time, measured energy consumption, and product quality data).

[0065] S280: Based on the deviation between the execution result and the expectation, update and optimize the model parameters or knowledge base to achieve closed-loop learning and optimization.

[0066] In another feasible implementation, the "process parameter instructions matched for each production task" are specifically explained. Taking the blowing pressure timing instruction as an example, its generation relies on a low-level adaptive control model. This model acquires the temperature field information of the current preform (e.g., high temperature in the center and low temperature at the edges detected by an infrared thermal imager) and the apparent viscosity information of the material. Based on a pre-trained feedforward lookup table (LUT), the model outputs a preliminary blowing pressure curve to compensate for temperature unevenness. Simultaneously, a feedback correction loop reads the wall thickness uniformity deviation of the product from the previous cycle and adjusts the parameters of the feedforward model online using the recursive least squares method. Finally, the model dynamically generates a final blowing pressure curve P(t) adapted to the current operating conditions and a precise blowing start point t_s, ensuring synchronous blowing across multiple workstations. This process parameter instruction is issued as part of the scheduling instruction set, realizing the combination of upper-level scheduling and low-level precision process control, ensuring the executability of the scheduling scheme at the equipment level and high-quality output.

[0067] In a specific application scenario, assume a production workshop has three blow molding machines (M1, M2, M3). Currently, there are two urgent orders (O1: 10,000 Type A bottles, delivery time 8 hours; O2: 5,000 Type B bottles, delivery time 12 hours) and one regular order (O3: 20,000 Type A bottles, delivery time 24 hours). The mold library contains corresponding molds (Mold_A, Mold_B), and the current electricity price is peak (price P_high). The multivariate operating condition acquisition module reports in real time: the parison temperature field uniformity index of M1 is 0.95 (excellent), and the material viscosity is stable; the temperature field uniformity index of M2 is 0.85 (good), but local hot spots are detected; the temperature field uniformity index of M3 is 0.70 (poor), and the material viscosity shows a fluctuating trend.

[0068] The collaborative scheduling decision engine performs the following integrated decision-making: Analysis constraints: O1 and O2 have tight delivery times; peak electricity prices are high; M1 has the best process condition, and M3 has the worst; Mold_A needs to be shared between M1 and M3.

[0069] Optimization Solution: The engine optimizes to "maximize order on-time performance, minimize peak energy consumption, and maximize expected quality." Traditional scheduling may only allocate tasks according to the order in which devices are idle (e.g., M1 does O1, M2 does O2, M3 does O3).

[0070] Output of the present invention: Task allocation: Assign the urgent order O1 with high precision requirements to M1, which has the best process status. Assign another urgent order O2 to M2, which is in good condition, but dynamically generate a set of conservative process parameters for it (such as slightly lower blowing pressure and longer holding time) to compensate for its local hot spots and ensure quality.

[0071] Resources and Scheduling: Regular order O3 is assigned to M3, which is in a worse state, but its production time is scheduled to avoid peak electricity price periods, delaying production until off-peak electricity price periods. Simultaneously, the system triggers a preventative maintenance reminder for M3 in advance and plans to use a set of strongly compensated process parameters for its off-peak production.

[0072] Process instructions: The blowing pressure curve generated for M1 is aggressive and efficient (P1(t)), the curve generated for M2 is robust and conservative (P2(t)), and the curve generated for M3 is strongly compensated and adapted to valley production (P3(t)). These instructions are all issued as part of the scheduling instruction set.

[0073] In this embodiment, by matching low-priority tasks with equipment in poor condition and moving them to off-peak production, the overall energy consumption cost is significantly reduced. Furthermore, by assigning appropriate process parameters to equipment in different states, the consistency of output quality for all orders (especially urgent orders) is ensured, reflecting the optimization of global resources.

[0074] In another specific application scenario, see Figure 1 and Figure 2 This section details the specific workflow of the collaborative scheduling decision engine 130 within a single decision cycle.

[0075] When the system reaches step S240, the decision engine 130 is activated. (See also...) Figure 3 Its internal operations are as follows: 1. Data Input: Constraint building unit 1311 receives ( Figure 1 The input to the E module comes from structured data from the global information perception module 110, such as: order set O={O1,O2,O3}, where each order O_i contains attributes: product type Type_i, quantity Qty_i, latest delivery time DDL_i, and priority weight W_i.

[0076] Simultaneously, the constraint construction unit 1311 reads the real-time state vectors of each machine from the multivariable operating condition acquisition module 120. For example, for machine M_j, its state vector S_j can be quantized as: S_j={TU_j,V_j,HS_j}, where TU_j is the uniformity index of the billet temperature field (0-1), V_j is the apparent viscosity of the material (Pa·s), and HS_j is a Boolean value indicating whether there are local hot spots.

[0077] The status of available resources is also input, such as the mold set D={Mold_A,Mold_B} and its current machine and lifespan.

[0078] 2. Constraint Construction: Based on the above inputs, the constraint construction unit 1311 transforms the business rules into strict mathematical constraints.

[0079] Resource allocation constraints (binary decision variable x_{ij}=1 indicates that order O_i is allocated to machine M_j):

[0080] Process feasibility constraints (core innovation, dynamically generated based on multivariate operating condition information S_j): Define a dynamic feasible process parameter domain Φ_j for each machine M_j and the types of products it may produce. For example, for the blowing pressure P and the starting point t_s:

[0081]

[0082] Here, P_{max}(S_j) and t_{s,min}(S_j) are functions of S_j. For example, if HS_j=True (hot spot), then the value of P_{max}(S_j) is decreased, and the value of t_{s,min}(S_j) is increased. This ensures that any order assigned to machine M_j must have process parameters that are safe and executable under the current operating condition S_j.

[0083] Order delivery constraints: Let p_{ij} be the estimated processing time of order O_i on machine M_j, which is a function of the order quantity Qty_i and the process parameters (affecting cycle time) assigned to that order. Let b_i be the start time of order O_i. Then we have:

[0084] Energy cost constraints: Let c(t) be the electricity price function at time t, and e_{ij}(t) be the power curve (determined by process parameters) when order O_i is produced on machine M_j. Then the total energy cost C_{energy} is:

[0085] 3. Optimization Solution: Unit 1312 integrates the above constraints and constructs a comprehensive system performance index F as the objective function. F is the weighted sum of multiple sub-objectives:

[0086] Wherein, T_{on-time rate} is calculated based on the satisfaction of delivery constraints.

[0087] U_{equipment utilization rate} is the ratio of total effective processing time to total available time.

[0088] Q_{quality expectation} is a quality consistency score estimated based on the degree to which the assigned process parameters fall within its optimal window Φ_j^* (a subset of Φ_j), and it directly depends on the multivariate operating condition information S_j.

[0089] Ultimately, the collaborative scheduling problem is formalized into a complex mixed-integer dynamic optimization problem: under the condition of satisfying all the above constraints, the overall efficiency index F is maximized by adjusting the discrete variables x_{ij}, b_i and continuous variables (process parameters P, t_s, etc.).

[0090] The optimization unit 1312 invokes an intelligent optimization algorithm (such as a genetic algorithm) to solve the problem. The algorithm maintains a population, where each individual (chromosome) encodes a complete scheduling scheme (i.e., the values ​​of all decision variables). It iterative evolution is achieved through operations such as selection (retaining individuals with high F-values), crossover (exchanging partial assignments or parameters of different individuals), and mutation (randomly changing a certain assignment or parameter). Ultimately, the algorithm converges to one or a set of Pareto optimal solutions.

[0091] 4. Instruction Generation: The optimal solution output by the solver is decoded. Engine 130 determines the final scheduling scheme based on this and generates a specific, machine-readable instruction set. This instruction set not only includes the task allocation of "when, where, and what to produce," but also a set of "tailor-made" process parameters for each task-machine pair.

[0092] For example, a final instruction might be: {“Target”:“M1”,“Order”:“O1”,“Start_Time”:t1,“Process_Profile”:{“Pressure_Curve”:“P1(t)”,“Blow_Start”:t_s1,…}}, where P1(t) and t_s1 are the specific process parameters obtained through optimization and adapted to the current operating condition S_1 of M1.

[0093] In a specific dynamic production scenario, after the system has been running for 2 hours according to the above embodiment, the multivariate working condition acquisition module 120 suddenly detected an abnormal drop in the melt pressure sensor reading of blow molding machine M2. Combined with vibration data analysis, the collaborative scheduling decision engine 130 judged that the plasticizing unit of M2 may have a sudden failure and it is expected to require 1 hour of maintenance.

[0094] 1. Dynamic Response: The decision engine 130 responds immediately to this unknown disturbance. It obtains the completion progress of all current orders and the latest operating status of each machine.

[0095] 2. Rapid Re-decision: The engine initiates emergency rescheduling optimization with the goal of "minimizing the impact of failures". Based on the real-time operating conditions of M1 and M3 (M1 is still running efficiently, and M3 has completed preheating and is ready to enter the valley production), and the remaining amount of unfulfilled order O2 on M2, rapid simulation and optimization are performed.

[0096] 3. Generate new instructions: The engine may decide on a new solution within seconds: split the remaining O2 orders on M2, with one part being produced by M1, which is close to completing its current task, during the gap (dynamically generating a set of rapid production parameters adapted to its current operating conditions), and the other part still being completed by the repaired M2 (but its subsequent process parameters are adjusted to compensate for the state after maintenance). At the same time, the off-peak production plan of M3 is dynamically adjusted to ensure overall capacity.

[0097] 4. Instruction Issuance: The new set of dispatch instructions is immediately issued to M1 and M3, and M2 receives the shutdown and maintenance instructions.

[0098] In this embodiment, when faced with the uncertain and unknown situation of "sudden equipment failure", the system of the present invention can quickly detect the abnormality based on real-time multivariate operating condition information and use its optimization model to perform dynamic rescheduling, minimizing the impact of the failure and avoiding the production line stagnation or large-scale order delays that may be caused by traditional static scheduling schemes, demonstrating the system's strong robustness and dynamic adaptability.

[0099] In another feasible implementation, the scheduling system of the present invention further integrates predictive capabilities. Assume that the multivariate operating condition acquisition module 120, through long-term monitoring, discovers that the apparent viscosity of the material in blow molding machine M3 exhibits a slow increase with continuous production time (e.g., viscosity increases by 5% every 4 hours of production). The model in the collaborative scheduling decision engine 130 learns this pattern.

[0100] 1. Predictive Judgment: When the system plans to allocate a long batch order to M3, the engine predicts based on historical operating data that the viscosity of M3 will exceed the optimal process window in the later stages of production of that batch.

[0101] 2. Preventive Scheduling: Therefore, during the initial decision-making process, the engine not only allocates tasks but also generates a time-varying sequence of process parameter instructions. For example, the standard parameter set P3_std(t) is used for the first 4 hours, the fine-tuning parameter set P3_adj1(t) is used from 4 to 8 hours, and the strongly compensated parameter set P3_adj2(t) is used after 8 hours.

[0102] 3. Parameter preloading: This parameter sequence is preloaded into the M3's local controller and automatically switches during production based on time or production count.

[0103] 4. Comparison of Results: Without this solution, M3 may experience uneven product wall thickness in the later stages of a batch due to increased viscosity, requiring production stoppages for adjustments or resulting in a large number of defective products. With this solution, M3 can seamlessly adapt to gradual changes in its condition, maintaining high-quality output throughout long-batch production, thus improving equipment utilization and first-pass yield.

[0104] This embodiment demonstrates how the proposed solution utilizes historical multivariate operating condition data for trend learning and prediction, and based on this, performs forward-looking and refined process parameter scheduling, eliminating problems before they occur, thus showcasing an advanced form of intelligent scheduling system.

[0105] It should be noted that the sensors, algorithms (genetic algorithms, reinforcement learning, recursive least squares), and communication methods (API, industrial networks) mentioned in the above embodiments are all common knowledge or existing technologies in this technical field. The focus of the embodiments of the present invention is to combine and apply them in a specific way to solve specific technical problems.

[0106] It should be understood that the terms "one embodiment," "an embodiment," "a feasible implementation," or "some implementations" used throughout the specification mean that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of the present invention. Therefore, "one embodiment," "an embodiment," "a feasible implementation," or "some implementations" appearing throughout the specification do not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. Those skilled in the art should also recognize that the embodiments described in the specification are optional embodiments, and the actions and modules involved are not necessarily essential to the embodiments of the present invention.

[0107] The above description is merely a specific embodiment of the present invention, but the protection scope of the embodiments of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the embodiments of the present invention should be included within the protection scope of the embodiments of the present invention. Therefore, the protection scope of the embodiments of the present invention should be determined by the protection scope of the claims.

Claims

1. A method for optimal scheduling of production and resources in coordination of multiple blow molding machines, characterized in that, Applied to a production system comprising at least two blow molding machines, the method includes: Obtain global scheduling information, which includes at least the set of production orders to be processed, the status of available production resources, and external environment parameters; Real-time acquisition of multivariate operating condition information for each blow molding machine, including at least the parison temperature field information and material rheological properties information within the current production cycle of the corresponding blow molding machine, to characterize the real-time process status that directly affects the blow molding quality. Based on the global scheduling information and the real-time acquired multivariate operating condition information, an integrated decision is made through a preset collaborative scheduling optimization model to generate a collaborative scheduling instruction set. The collaborative scheduling optimization model is configured to: under the premise of meeting the constraints of production resources and order delivery, take the multivariate working condition information as the key input, optimize the decision-making of production task allocation and process parameter combination that can improve the overall efficiency of the system and adapt to the real-time process status of each blow molding machine; The collaborative scheduling instruction set includes: production task instructions assigned to each blow molding machine, and process parameter instructions matched for each production task and dynamically determined based on the current multivariate operating condition information of the corresponding blow molding machine. The collaborative scheduling instruction set is sent to the corresponding blow molding machine for execution, so as to drive the at least two blow molding machines to carry out collaborative production.

2. The method of claim 1, wherein, The integrated decision-making based on the global scheduling information and the real-time acquired multivariate operating condition information, through a preset collaborative scheduling optimization model, specifically includes: With the goal of optimizing the overall system performance index, an optimization problem model is constructed that includes constraints on production resources, process feasibility, and order delivery. The process feasibility constraint is the range of process parameters that each blow molding machine can safely and stably execute under the current operating conditions, determined based on the multivariate operating condition information. The optimization problem model is solved using an intelligent optimization algorithm, and the cooperative scheduling instruction set is output.

3. The method according to claim 2, characterized in that, The system's overall performance index is a multi-objective weighted function that takes into account on-time order delivery rate, equipment utilization rate, total production energy consumption, and the consistency of product quality in the same batch.

4. The method according to claim 1, characterized in that, The process parameter instructions matched for each production task include at least the blowing pressure timing instructions. The blowing pressure timing command is dynamically generated by an adaptive control model based on the preform temperature field information and material rheological properties information, and is used to drive the blowing head of the corresponding blow molding machine to perform synchronous blowing action.

5. The method according to claim 4, characterized in that, The adaptive control model adopts a hybrid architecture that combines feedforward compensation and feedback correction. It is configured to adjust the pressure curve parameters and action timing start points of each blowing head according to the billet temperature field information and material rheological properties information, so as to drive all blowing heads to act synchronously according to a unified target time reference.

6. The method according to claim 1, characterized in that, The collaborative scheduling optimization model is a decision model based on deep reinforcement learning, which obtains scheduling strategies through interaction with the environment. The state space of the environment includes the global scheduling information and the historical and real-time multivariate operating condition information. The action space is the possible task allocation and process parameter combination. The reward function is designed based on the comprehensive performance index of the system.

7. A collaborative production and optimal resource scheduling system for implementing the method as described in any one of claims 1-6, characterized in that, Deployed in a production workshop containing at least two blow molding machines, the system includes: The global information perception module is configured to acquire the global scheduling information; The multivariable operating condition acquisition module is connected to each blow molding machine and is configured to acquire the multivariable operating condition information of each blow molding machine in real time. The collaborative scheduling decision engine, which is communicatively connected to the global information perception module and the multivariate operating condition acquisition module, is configured to: run the collaborative scheduling optimization model and generate the collaborative scheduling instruction set based on the input global scheduling information and multivariate operating condition information; The instruction distribution and execution module is communicatively connected to the collaborative scheduling decision engine and the control systems of each blow molding machine, and is configured to issue the collaborative scheduling instruction set and drive the blow molding machine to execute it.

8. The system according to claim 7, characterized in that, The multivariable operating condition acquisition module includes modules deployed on each blow molding machine: Infrared thermal imager array for non-contact detection of the surface temperature field of a blank; Melt pressure sensor and screw speed sensor are used to obtain the raw data required to calculate the rheological properties of materials; The collaborative scheduling decision engine or the local controller of the blow molding machine is configured to calculate the apparent viscosity of the material based on the melt pressure and screw speed.

9. The system according to claim 7, characterized in that, The system also includes a digital twin simulation module; the digital twin simulation module performs forward-looking simulation and performance evaluation of the scheduling scheme based on the collaborative scheduling instruction set, equipment model and historical data, and feeds the evaluation results back to the collaborative scheduling decision engine for instruction correction or model learning.

Images