Distributed cooperative control system of barreled water cleaning and filling production line

Abstract

The invention discloses a distributed cooperative control system for a barreled water cleaning and filling production line. The distributed cooperative control system comprises a physical production line, a sensor network, an actuator network and a control system software platform, a control system software platform adopts a three-level distributed architecture of a cloud end, an edge end and an equipment end, and a causal digital twinborn body which is synchronous with a physical production line in real time is constructed. The system can perform dynamic optimization and collaboration from a global perspective, solves the contradiction between local optimization and global optimization in a distributed system, and comprehensively improves the production efficiency of a production line; the embedded causal atlas and the anti-factual reasoning engine enable the system to understand the internal logic relationship of the production process, and when an abnormality occurs, cause analysis can be carried out, and a correction scheme can be rapidly formulated and verified.

Description

Technical Field

[0001] This invention relates to the field of industrial automation control technology, specifically a distributed collaborative control system for a bottled water cleaning and filling production line. Background Technology

[0002] The bottled water production process mainly includes empty bottle collection, external inspection, internal washing, disinfection, filling, capping, light inspection, bagging and palletizing, etc. Currently, the mainstream control system architecture is mainly divided into two types: Centralized PLC control system: A main PLC is responsible for the logic control of the entire line. Sensors and actuators at each workstation are connected to the main PLC through fieldbus or I / O modules. This architecture is simple to program, but the main PLC becomes a single point of failure. Once it fails, it will cause the entire line to stop production. The control logic is tightly coupled, and any process modification or production line expansion requires reprogramming and debugging, resulting in poor flexibility. Segmented independent control system: Each major station (such as the cleaning station and the filling station) is equipped with an independent PLC or dedicated controller, and the stations are only roughly synchronized through simple start / stop signals or counting signals. Although this architecture improves local reliability, it has poor coordination and cannot be dynamically adjusted according to the real-time status of upstream and downstream.

[0003] As consumers demand higher quality bottled water and production costs increase, production lines need to become more flexible and intelligent. To this end, this application proposes a distributed collaborative control system for a bottled water cleaning and filling production line. Summary of the Invention

[0004] The purpose of this invention is to provide a distributed collaborative control system for a bottled water cleaning and filling production line to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a distributed collaborative control system for a bottled water cleaning and filling production line, comprising a physical production line, a sensor network, an actuator network, and a control system software platform; the control system software platform adopts a three-level distributed architecture of cloud, edge, and device, and constructs a causal digital twin that is synchronized with the physical production line in real time; The system includes: a causal digital twin module, which runs on the cloud or a factory server and is a high-fidelity dynamic simulation model driven by the fusion of first principles of physicochemistry and real-time operating data. It has embedded a structured causal map that characterizes the causal relationship of the production process. The strategic collaborative intelligent agent module, running in the cloud or on the factory server, is used to perform inferences based on global production goals in the causal digital twin, generate and issue high-level optimization instructions to each tactical collaborative intelligent agent module; Multiple tactical collaborative intelligent agent modules are deployed in the edge computing units corresponding to each production section to receive the high-order optimization instructions and generate a cross-section collaborative equipment control instruction set by combining the real-time status of the sections under their jurisdiction and adjacent sections. Multiple device control intelligent agent modules are deployed on each key execution device controller or adjacent edge node to receive and execute corresponding instructions in the collaborative device control instruction set; The causal diagnosis and self-healing engine module, running at the edge or in the cloud, is used to initiate counterfactual reasoning in the causal digital twin to locate the root cause when a production anomaly or quality risk is detected. It also simulates and generates dynamic workflow reorganization schemes and parameter adjustment strategies, which are then distributed to the relevant intelligent agent modules for execution.

[0006] Preferably, the structured causal graph is represented in the form of a directed graph, where nodes include equipment status parameters, process parameters, intermediate product quality attributes, and final product quality indicators, and edges represent the causal relationships and influence weights between nodes.

[0007] Preferably, the causal digital twin module has parameter self-calibration capability. By continuously comparing the sensor readings predicted by the twin with the actual sensor readings, when the deviation exceeds the adaptive threshold, the optimization algorithm is triggered to update the latent variable parameters in the twin that characterize the degradation of equipment performance or changes in material properties.

[0008] Preferably, the strategic collaborative intelligent agent module is trained using an algorithm based on multi-objective deep reinforcement learning, and its reward function integrates indicators such as production capacity, energy consumption, water consumption, quality pass rate, and equipment health.

[0009] Preferably, the tactical collaborative intelligent agent module is trained using a hierarchical course learning mechanism, with the training phases including: steady-state basic control training, target maintenance training under common disturbances, and cross-section collaborative recovery training under extreme anomalies.

[0010] As a preferred embodiment, the method for counterfactual reasoning in the causal diagnosis and self-healing engine module includes: fixing all conditions except for suspected root cause nodes in the causal digital twin, modifying the historical state of the root cause node and re-deducing the production process, verifying the root cause by observing changes in quality indicators and evaluating the effectiveness of different corrective measures.

[0011] Preferably, the sensor network includes a machine vision sensor for detecting the cleanliness of the container, an electrochemical sensor for detecting the concentration of the cleaning and disinfecting agent, an ultrasonic sensor for detecting the filling level, and a laser rangefinder for detecting the status of the sealing ring.

[0012] Compared with the prior art, the beneficial effects of the present invention are: By combining a three-tiered intelligent agent architecture of "strategy-tactics-equipment" with a causal digital twin, the system can dynamically optimize and coordinate from a global perspective, resolving the contradiction between local optimization and global optimization in a distributed system, and comprehensively improving the production efficiency of the production line. The embedded cause-effect graph and counterfactual reasoning engine enable the system to understand the inherent logical relationships of the production process, perform cause analysis when anomalies occur, and quickly formulate and verify corrective measures. The intelligent agent training mechanism based on curriculum learning, the two-way dynamic calibration of digital twins, and the cross-production line knowledge federation enable the entire control system to continuously adapt and optimize as equipment ages, processes are innovated, and products are changed. The distributed architecture avoids the risk of single point of failure. The failure of a single agent or node does not affect the overall operation. The modular agent design makes production line expansion and process changes more flexible. Only the corresponding agent modules need to be added or updated, which reduces the complexity and cost of system upgrades. Detailed Implementation

[0013] Exemplary embodiments will be described in detail below. The technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with 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.

[0014] This invention provides a technical solution: a distributed collaborative control system for a bottled water cleaning and filling production line, comprising a physical production line, a sensor network, an actuator network, and a control system software platform; the control system software platform adopts a three-level distributed architecture of cloud, edge, and device, and constructs a causal digital twin that is synchronized with the physical production line in real time; The system includes: a causal digital twin module, which runs on the cloud or factory server. It is a high-fidelity dynamic simulation model driven by the fusion of first principles of physicochemistry and real-time operating data, and has embedded a structured causal map characterizing the causal relationship of the production process. The strategic collaborative intelligent agent module, running in the cloud or on the factory server, is used to perform inferences in the causal digital twin based on global production goals, and to generate and issue high-level optimization instructions to each tactical collaborative intelligent agent module. Multiple tactical collaborative intelligent agent modules are deployed in the edge computing units corresponding to each production section to receive high-order optimization instructions and generate a set of collaborative equipment control instructions across sections by combining the real-time status of the sections under their jurisdiction and adjacent sections. Multiple device control intelligent agent modules are deployed on each key execution device controller or adjacent edge node to receive and execute corresponding instructions in the collaborative device control instruction set; The causal diagnosis and self-healing engine module runs at the edge or in the cloud. When a production anomaly or quality risk is detected, it initiates counterfactual reasoning in the causal digital twin to locate the root cause, and simulates and generates dynamic workflow reorganization schemes and parameter adjustment strategies, which are then distributed to the relevant intelligent agent modules for execution.

[0015] The structured causal graph is represented in the form of a directed graph. The nodes include equipment status parameters, process parameters, intermediate product quality attributes, and final product quality indicators. The edges represent the causal relationship and influence weight between the nodes.

[0016] Among them, the causal digital twin module has the ability to self-calibrate parameters. By continuously comparing the sensor readings predicted by the twin with the actual sensor readings, when the deviation exceeds the adaptive threshold, the optimization algorithm is triggered to update the hidden variable parameters in the twin that characterize the degradation of equipment performance or changes in material properties.

[0017] Among them, the strategic collaborative intelligent agent module is trained using an algorithm based on multi-objective deep reinforcement learning, and its reward function integrates indicators such as production capacity, energy consumption, water consumption, quality pass rate and equipment health.

[0018] The tactical collaborative intelligent agent module is trained using a hierarchical course learning mechanism. The training phases include: steady-state basic control training, target maintenance training under common disturbances, and cross-section collaborative recovery training under extreme anomalies.

[0019] The counterfactual reasoning method of the causal diagnosis and self-healing engine module includes: fixing all conditions except for suspected root cause nodes in the causal digital twin, modifying the historical state of the root cause node and re-deducing the production process, verifying the root cause by observing changes in quality indicators and evaluating the effectiveness of different corrective measures.

[0020] The sensor network includes a machine vision sensor for detecting the cleanliness of the container, an electrochemical sensor for detecting the concentration of the cleaning and disinfecting agent, an ultrasonic sensor for detecting the filling level, and a laser rangefinder for detecting the status of the sealing ring.

[0021] Example 1: This embodiment provides a distributed collaborative control system for a bottled water washing and filling production line. The physical production line includes, in sequence, an empty barrel loading section, a high-pressure spraying section, a disinfection and rinsing section, a sterile water draining section, a quantitative filling section, a capping and sealing section, a light inspection section, and a palletizing section. Each section is equipped with a rich network of sensors and actuators. The control system software platform adopts a three-tier architecture of cloud-edge-device: Cloud / Factory Server Layer: Deployed with causal digital twin module, strategic collaborative intelligent agent module, knowledge federation platform interface module, as well as global monitoring and human-machine interface; The causal digital twin module is a virtual mapping of the physical production line. Its model integrates principles such as cleaning, filling, disinfection, and microbial dynamics, and has a built-in causal graph. Based on the monthly production plan and real-time energy prices, the strategic collaborative intelligent agent module simulates production for the next few hours in the twin and outputs instructions such as "prioritize ensuring production capacity in the current period, allowing for a slight increase in energy consumption" or "switch to energy-saving mode and appropriately reduce line speed." Edge computing layer: Edge servers are deployed for each work section. Each edge server runs a tactical collaborative intelligent agent module. For example, the edge server in the cleaning area runs a "cleaning tactical intelligent agent". It receives strategic instructions and integrates the status of its own area and the adjacent drying section (such as the sensor detecting that a batch of recycling bins is particularly dirty). It makes millisecond-level decisions locally and generates collaborative strategies such as "extend the high-pressure spraying time of the current batch of bins by % and notify the drying section to adjust the air drying time accordingly", and decomposes them into specific equipment instructions. Equipment control layer: Deploy equipment control intelligent agent modules in the controllers or I / O stations of each key equipment, such as the intelligent agent in the filling valve controller, to receive precise filling volume instructions from the "filling tactical intelligent agent" and combine them with real-time flow feedback for PID control and anti-disturbance compensation; The causal diagnostics and self-healing engine module can be deployed on critical edge nodes or in the cloud, and it monitors all quality-related sensors and key nodes in the causal graph in real time.

[0022] The system workflow is as follows: Initialization and Synchronization: Upon system startup, the causal digital twin synchronizes the initial state of all equipment from the physical production line and establishes an initial virtual image; Strategic decision-making: The strategic collaborative intelligent agent in the twin performs deductions based on current order, inventory, and energy consumption data, generates high-order target instructions for the current production cycle, and broadcasts them to all tactical collaborative intelligent agents; Tactical coordination: Each tactical coordination agent makes distributed decisions based on higher-order objectives and local real-time data (including upstream and downstream status). For example, when the filling tactical agent predicts that its filling head needs preventive maintenance through its twin, it will negotiate in advance with the upstream draining and downstream capping tactical agents to fine-tune the timing, so that the maintenance window can be inserted without stopping the line. Device execution: The device controls the intelligent agent to precisely execute instructions and feeds back the execution results and sensor data to the edge layer and cloud in real time to update the digital twin; Dynamic calibration: The causal digital twin continuously compares the predicted values ​​with the actual values, automatically calibrating the model parameters to maintain their fidelity; Anomaly Handling: When the light inspection section reports multiple blurry barrel labels in succession, the cause-effect diagnosis and self-healing engine is triggered.

[0023] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A distributed collaborative control system for a bottled water cleaning and filling production line, comprising a physical production line, a sensor network, an actuator network, and a control system software platform; characterized in that, The control system software platform adopts a three-level distributed architecture of cloud, edge, and device, and has built a causal digital twin that is synchronized with the physical production line in real time. The system includes: a causal digital twin module, which runs on the cloud or a factory server and is a high-fidelity dynamic simulation model driven by the fusion of first principles of physicochemistry and real-time operating data. It has embedded a structured causal map that characterizes the causal relationship of the production process. The strategic collaborative intelligent agent module, running in the cloud or on the factory server, is used to perform inferences based on global production goals in the causal digital twin, generate and issue high-level optimization instructions to each tactical collaborative intelligent agent module; Multiple tactical collaborative intelligent agent modules are deployed in the edge computing units corresponding to each production section to receive the high-order optimization instructions and generate a cross-section collaborative equipment control instruction set by combining the real-time status of the sections under their jurisdiction and adjacent sections. Multiple device control intelligent agent modules are deployed on each key execution device controller or adjacent edge node to receive and execute corresponding instructions in the collaborative device control instruction set; The causal diagnosis and self-healing engine module, running at the edge or in the cloud, is used to initiate counterfactual reasoning in the causal digital twin to locate the root cause when a production anomaly or quality risk is detected. It also simulates and generates dynamic workflow reorganization schemes and parameter adjustment strategies, which are then distributed to the relevant intelligent agent modules for execution.

2. The distributed collaborative control system for a bottled water washing and filling production line according to claim 1, characterized in that: The structured causal graph is represented in the form of a directed graph. The nodes include equipment status parameters, process parameters, intermediate product quality attributes, and final product quality indicators. The edges represent the causal relationship and influence weight between nodes.

3. The distributed collaborative control system for a bottled water washing and filling production line according to claim 1, characterized in that: The causal digital twin module has parameter self-calibration capability. By continuously comparing the sensor readings predicted by the twin with the actual sensor readings, when the deviation exceeds the adaptive threshold, the optimization algorithm is triggered to update the hidden variable parameters in the twin that characterize the degradation of equipment performance or changes in material properties.

4. The distributed collaborative control system for a bottled water washing and filling production line according to claim 1, characterized in that: The strategic collaborative intelligent agent module is trained using an algorithm based on multi-objective deep reinforcement learning, and its reward function integrates indicators such as production capacity, energy consumption, water consumption, quality pass rate, and equipment health.

5. The distributed collaborative control system for a bottled water cleaning and filling production line according to claim 1, characterized in that: The tactical collaborative intelligent agent module is trained using a hierarchical course learning mechanism. The training phases include: steady-state basic control training, target maintenance training under common disturbances, and cross-section collaborative recovery training under extreme anomalies.

6. The distributed collaborative control system for a bottled water washing and filling production line according to claim 1, characterized in that: The method for counterfactual reasoning in the causal diagnosis and self-healing engine module includes: fixing all conditions except for suspected root cause nodes in the causal digital twin, modifying the historical state of the root cause node and re-deducing the production process, verifying the root cause by observing changes in quality indicators and evaluating the effectiveness of different corrective measures.

7. The distributed collaborative control system for a bottled water washing and filling production line according to claim 1, characterized in that: The sensor network includes a machine vision sensor for detecting the cleanliness of the container, an electrochemical sensor for detecting the concentration of the cleaning and disinfecting agent, an ultrasonic sensor for detecting the filling level, and a laser rangefinder for detecting the status of the sealing ring.