Flexible production line intelligent management system based on artificial intelligence
By using an AI-based intelligent management system for flexible production lines, machine learning and reinforcement learning technologies are employed to achieve efficient and flexible switching of flexible production lines. This solves the problems of low efficiency and high cost in traditional flexible production lines when frequently switching between multiple product types, thereby improving production efficiency and equipment utilization.
Patent Information
- Application Number
- CN202511180202.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-08-22
- Publication Date
- 2026-01-16
AI Technical Summary
Traditional flexible production lines suffer from inefficiency and high costs in process conversion, equipment adjustment, and production sequence arrangement. In particular, when frequently switching between multiple product types, the lack of intelligent management and real-time data analysis leads to low production efficiency and equipment utilization.
An AI-based intelligent management system for flexible production lines is adopted, including a data acquisition and preprocessing module, a feature extraction module, a strategy optimization module, and an early warning and dynamic adjustment module. Utilizing machine learning and reinforcement learning technologies, it automatically generates production changeover plans, optimizes equipment parameters and production sequences, and monitors and adjusts the production process in real time.
It enables efficient and flexible switching of production lines, reduces production costs, improves production efficiency and equipment utilization, and ensures the continuity and stability of production.
Smart Images
Figure CN121348979A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application belongs to the technical field of production line management, and particularly relates to an intelligent management system for a flexible production line based on artificial intelligence. BACKGROUND
[0002] In today's manufacturing industry, flexible production lines have become a key technology to cope with the demand for multi-variety and small-batch production. Traditional rigid production line design is only suitable for mass production and is difficult to adapt to the rapid changes in market demand. In order to realize flexible production of different products, manufacturing enterprises generally use flexible production lines to quickly switch different products in the production process. However, traditional flexible production lines still face many challenges in process conversion, equipment adjustment and production sequence arrangement, especially when switching frequently between different products, production efficiency is often affected and production cost is difficult to control.
[0003] In the operation of a flexible production line, different products produced often have different process requirements and equipment parameters. For example, different products may require different processing temperatures, speeds, pressures, etc. Traditional production lines usually rely on manual setting and adjustment of equipment parameters. This not only takes time and effort, but also is prone to errors due to human factors, which can affect production efficiency and product quality. In addition, the production line needs a long conversion time when switching products, which reduces the utilization rate of the production line and increases production costs.
[0004] The existing flexible production line management mode lacks comprehensive analysis of real-time production data and intelligent monitoring of equipment status, and cannot fully optimize equipment switching and scheduling in the production process. How to shorten the production switching time, reduce the production cost and improve the production flexibility is a key problem faced by the current flexible production line.
[0005] The existing flexible production line management systems on the current market usually rely on pre-set control programs and parameter adjustments, but these systems have limited adjustment speed and lack sufficient intelligence. Especially when dealing with frequent switching and small-batch multi-variety production, existing systems still have great deficiencies in automation and response to complex environments. In addition, traditional flexible production lines rely on fixed production plans and cannot be dynamically adjusted according to real-time orders and equipment status, resulting in insufficient flexibility. SUMMARY
[0006] To solve the above technical problems, the present application provides an intelligent management system for a flexible production line based on artificial intelligence to solve the problems existing in the prior art.
[0007] To achieve the above purpose, the present application provides an intelligent management system for a flexible production line based on artificial intelligence, comprising:
[0008] The data acquisition and preprocessing module, the feature extraction module, the strategy optimization module and the early warning and dynamic adjustment module are connected in sequence.
[0009] The data acquisition and preprocessing module is configured to acquire production line data and perform preprocessing.
[0010] The feature extraction module is configured to perform feature extraction on the preprocessed production line data to obtain a time series feature vector.
[0011] The strategy optimization module is configured to obtain an optimal action according to the time series feature vector.
[0012] The early warning and dynamic adjustment module is configured to generate an early warning signal according to a preset threshold and real-time acquired production line data, and adjust the production line according to the early warning signal.
[0013] The data acquisition and preprocessing module, the feature extraction module, the strategy optimization module and the early warning and dynamic adjustment module are connected in sequence.
[0014] Optionally, the data acquisition and preprocessing module comprises a data acquisition unit and a preprocessing unit.
[0015] The data acquisition unit is configured to acquire production line data, and the production line data comprises equipment running state data, production process parameters and order information.
[0016] The preprocessing unit performs data cleaning on the production line data, processes the cleaned data based on a correlation analysis statistical method to obtain key features, and performs normalization processing on the key features.
[0017] Optionally, the feature extraction module is configured to construct a feature extraction model, input the normalized key feature data into the feature extraction model to obtain a time series feature vector.
[0018] Optionally, the feature extraction module constructs a feature extraction model based on a long short-term memory network and a convolutional neural network, wherein the convolutional neural network model adopts Resnet50.
[0019] Optionally, the strategy optimization module constructs an optimization model, inputs the time series feature vector into the optimization model as an input of the optimization model to obtain an optimal action, and the optimization model is obtained based on a deep Q network.
[0020] Optionally, a reward function of the optimization model is obtained based on production efficiency, equipment utilization rate, order fulfillment rate and equipment load; a state space of the optimization model comprises equipment state, production task, order demand and product information; and an action space comprises equipment parameter adjustment, production order adjustment and task allocation operation.
[0021] Optionally, it may also include a constraint module, which includes a constraint construction unit and a constraint integration unit;
[0022] The constraint construction unit is used to construct physical constraints;
[0023] The constraint integration unit is used to integrate the physical constraints into the decision-making process of the optimization model.
[0024] Optionally, the physical constraints in the constraint construction unit include equipment performance constraints, production process constraints, resource constraints, equipment safety constraints, and environmental constraints.
[0025] Compared with the prior art, the present invention has the following advantages and technical effects:
[0026] This invention discloses an intelligent management system for flexible production lines based on artificial intelligence. Through AI technology, combined with machine learning algorithms and reinforcement learning techniques, the system can automatically generate production changeover plans and operation instructions, automatically adjust equipment parameters according to changes in production orders, optimize the production sequence, and dynamically detect potential problems during equipment changeover, providing early warnings and adjustments to ensure the continuity and efficiency of the production line. Attached Figure Description
[0027] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:
[0028] Figure 1 This is a schematic diagram of the feature extraction module according to an embodiment of the present invention;
[0029] Figure 2 This is a system schematic diagram according to an embodiment of the present invention. Detailed Implementation
[0030] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.
[0031] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.
[0032] Example 1
[0033] like Figures 1-2 As shown, this embodiment provides an intelligent management system for flexible production lines based on artificial intelligence, including:
[0034] The module includes data acquisition and preprocessing, feature extraction, strategy optimization, and early warning and dynamic adjustment.
[0035] The data acquisition and preprocessing module is used to acquire production line data and perform preprocessing.
[0036] Furthermore, the data acquisition and preprocessing module includes a data acquisition unit and a preprocessing unit;
[0037] The data acquisition unit is used to acquire production line data, which includes equipment operating status data, production process parameters, and order information.
[0038] After cleaning the production line data, the preprocessing unit processes the cleaned data using statistical methods based on correlation analysis to obtain key features; and then normalizes the key features.
[0039] Specifically, each piece of equipment and production process on a flexible production line is related to the production of a specific product. To achieve intelligent switchover management, the system first needs to collect data from the production line, including:
[0040] Equipment parameter data: Data is collected through sensors. Sensors installed on key components of the equipment (such as temperature sensors and pressure sensors) are connected to the data acquisition system via the Industrial Internet of Things (IIoT) to transmit the equipment's operating status data in real time.
[0041] Production process parameters: Process requirements during production (such as time, temperature, pressure, etc. for each step) are collected through the Production Management System (MES). These parameters are updated synchronously through the interface between the equipment and the MES system.
[0042] Order information: Order information is obtained through the ERP system, including product model, production quantity, production time, and product attribute data.
[0043] All data will be standardized into a vector representation, where each element x i This represents the state of a piece of equipment or a parameter of a production process. Data vector X = {x1, x2, ..., x...} n This constitutes the system's input dataset.
[0044] After data collection, the system preprocesses and organizes the collected data to facilitate the training and application of machine learning models. Specific steps include:
[0045] Data cleaning: Remove missing values, outliers, and noisy data, and use interpolation or data imputation techniques to handle missing values.
[0046] Feature extraction and selection: Important features are extracted from the raw data, and redundant features are removed. Feature selection uses statistical methods based on correlation analysis.
[0047] Data normalization: Normalize each feature to make them have similar scales, thereby improving the model training effect.
[0048] The feature extraction module is used to extract features from the preprocessed production line data to obtain a time-series feature vector;
[0049] Furthermore, the feature extraction module is used to construct a feature extraction model, inputting the normalized key feature data into the feature extraction model to obtain a temporal feature vector.
[0050] Furthermore, the feature extraction module constructs a feature extraction model based on a long short-term memory network and a convolutional neural network, wherein the convolutional neural network model adopts ResNet50.
[0051] Specifically, in the preprocessed data, we use Long Short-Term Memory (LSTM) networks and Convolutional Neural Networks (CNNs) for temporal feature extraction. LSTM effectively captures long-term dependencies in time-series data, while CNNs help extract local production pattern features. The feature vector formed by combining the two serves as a state representation of the production system, such as... Figure 1 As shown.
[0052] The convolutional neural network model uses the ResNet50 model. The core idea of LSTM is to use three gates (forget gate, input gate, and output gate) to control the flow of information, avoiding the vanishing gradient problem common in traditional RNNs during long sequence learning. Its update formula is:
[0053] f t =σ(W f [h t-1 ,x t ]+b f )
[0054] i t =σ(W i [h t-1 ,x t ]+b i )
[0055]
[0056] o t =σ(W o [h t-1 x t ]+b o )
[0057] h t =o t *tanh(C t );
[0058] Among them, f t For the Gate of Oblivion, i t For the input gate, C t In cellular state, o t For the output gate, h t This is a hidden state. LSTM networks effectively capture long-term dependencies in the production process through these gating mechanisms.
[0059] The strategy optimization module is used to obtain the optimal action based on the temporal feature vector;
[0060] Furthermore, the strategy optimization module constructs an optimization model, using the temporal feature vector as input to the optimization model to obtain the optimal action; wherein, the optimization model is obtained based on a deep Q-network.
[0061] Furthermore, the reward function of the optimization model is obtained based on production efficiency, equipment utilization rate, order fulfillment rate, and equipment load; the state space of the optimization model includes equipment status, production tasks, order requirements, and product information; the action space includes equipment parameter adjustment, production sequence adjustment, and task allocation operations.
[0062] Specifically, the Deep Q-Network (DQN) method from Deep Reinforcement Learning (DRL) is used to optimize the equipment changeover plan for a flexible production line. DQN is a deep reinforcement learning method based on the Q-learning algorithm. The Q-learning algorithm selects the optimal action by continuously updating the Q-value. In DQN, the Q-value is approximated by a deep neural network, avoiding the storage problem caused by the excessively large state space in traditional Q-learning. The goal of this model is to maximize the overall efficiency of the production line (including production efficiency, equipment utilization, and order fulfillment rate) and ensure the continuity and stability of the production changeover process.
[0063] The Q-value function in the Deep Q-Network (DQN) method:
[0064]
[0065] Where Q(s) t ,a t () indicates that at time t, the agent is in state s. t Take action a t The Q-value represents the quality of the state-action pair. E[·] represents the expected value, indicating the anticipated future reward. R t This indicates an immediate reward (immediate feedback), that is, in state st Next, take action a t The immediate reward received afterward. γ is a discount factor that controls the influence of future rewards. Typically, 0 ≤ γ ≤ 1; the smaller the discount factor, the more the agent focuses on short-term rewards; the larger the discount factor, the more the agent focuses on long-term rewards. at+1 Q(s t+1 ,a t+1 ): Indicates the state s in the next state. t+1 In this context, the maximum Q value that can be obtained by taking the optimal action is determined.
[0066] This section contains the following three parts:
[0067] State space S t : Represents the current state of the production line, including equipment status, production tasks, order requirements, and product information.
[0068] Action Space A t This includes adjusting equipment parameters, adjusting production sequence, and assigning tasks.
[0069] Reward function R t Used to measure the merits of each decision, it is usually related to production efficiency, product quality, and equipment load.
[0070] The decision-making process includes:
[0071] S1 Action Selection: The reinforcement learning model (DQN) evaluates the current state S t To select the optimal action A t In DQN, action selection is achieved by maximizing the Q-value:
[0072]
[0073] S2 Reward Calculation: After the action is executed, the reward R is calculated based on the feedback from the production line. t Rewards are typically based on factors such as production efficiency and equipment load.
[0074] R t =w1·E t +w2·U t +w3·Q t -w4·L t ;
[0075] Among them, E t For production efficiency, U t For equipment utilization, Q t For order fulfillment rate, L t For equipment load.
[0076] S3 Policy Optimization: Reinforcement learning algorithms continuously adjust decision-making strategies by optimizing the Q-value (DQN) to maximize cumulative rewards.
[0077] Q(S t A t )←Q(S t A t )+α[R t +γmaxQ(S t+1 A t+1 )-Q(S t A t )];
[0078] α is the learning rate, which represents the degree to which new information affects the update of the Q value. In the next state s t+1 In the context of the optimal action, the Q value is determined.
[0079] S4 Output: Selected Action A t The updated Q value is used to guide production line operations. It is used to optimize production changeover decisions.
[0080] The early warning and dynamic adjustment module is used to generate early warning signals based on preset thresholds and real-time production line data, and to adjust the production line according to the early warning signals;
[0081] Specifically, thresholds for equipment temperature, load, and current are set, and an early warning is issued when these thresholds are exceeded. The system monitors the production line's operating status in real time, performing fault detection and production plan adjustments to ensure production continuity and stability. After the early warning mechanism detects potential problems or anomalies, the system will automatically adjust production plans, equipment parameters, or task scheduling to eliminate problems, restore production efficiency, and ensure smooth production line operation.
[0082] The data acquisition and preprocessing module, the feature extraction module, the strategy optimization module, and the early warning and dynamic adjustment module are connected in sequence.
[0083] Furthermore, it also includes a constraint module, which comprises a constraint construction unit and a constraint integration unit;
[0084] The constraint construction unit is used to construct physical constraints;
[0085] The constraint integration unit is used to integrate the physical constraints into the decision-making process of the optimization model.
[0086] Furthermore, the physical constraints in the constraint construction unit include equipment performance constraints, production process constraints, resource constraints, equipment safety constraints, and environmental constraints.
[0087] For example, in a flexible production line intelligent management system, physical constraints are limitations imposed on the actual physical laws governing the production process and the performance of production equipment. During the decision-making process of the reinforcement learning model, physical constraints ensure that the intelligent system's decisions are not only optimal but also practically feasible and safe. Physical constraints typically include physical performance limitations of equipment, physical constraints of the production process, and physical laws governing material flow. The types of physical constraints set in this invention are as follows:
[0088] Equipment performance constraints:
[0089] Each piece of equipment has limited performance, such as maximum processing speed, maximum load capacity, and equipment failure rate. These limitations affect the scheduling of production tasks; for example, at a certain moment, the equipment may not be able to withstand excessive loads or operate at excessive speeds.
[0090] Production process constraints:
[0091] Some production steps may need to be performed sequentially and the order cannot be arbitrarily changed. For example, some processes can only be carried out after the preceding steps are completed. Constraints on material flow ensure that materials on the production line flow along reasonable paths, avoiding bottlenecks or unnecessary waiting.
[0092] Resource constraints:
[0093] Production line resources (such as electricity, cooling water, raw materials, etc.) are limited. Production process scheduling must take into account the rational allocation and use of resources.
[0094] Equipment safety constraints:
[0095] Every piece of equipment has its safe operating limits, such as maximum values for parameters like temperature, pressure, and current. Exceeding these safety limits can lead to equipment damage or production accidents.
[0096] Environmental constraints:
[0097] In some production processes, environmental factors (such as temperature and humidity) may affect equipment performance or product quality. The system needs to dynamically adjust the production process according to environmental changes.
[0098] The integration of physical constraints into reinforcement learning is as follows:
[0099] Representation of constraints in state space: Physical constraints are typically implemented by limiting certain parts of the state space. For example, when the device load exceeds a safety threshold, the corresponding state St is marked as infeasible, thus avoiding the selection of overload operation.
[0100] Representation of constraints in the action space: In the action space, physical constraints can restrict the selection of certain actions. For example, when the system is under excessive load on certain equipment, it cannot choose to adjust the equipment parameters or schedule the equipment for production tasks.
[0101] Incorporating constraints into the reward function: Physical constraints can also be implemented by modifying the reward function. For example, when the operating parameters of a device exceed the allowable range, the system will penalize the reward for that action, thus making the system more inclined to choose actions that conform to physical constraints.
[0102] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
Claims
1. An intelligent management system for flexible production lines based on artificial intelligence, characterized in that, include: The module includes data acquisition and preprocessing, feature extraction, strategy optimization, and early warning and dynamic adjustment. The data acquisition and preprocessing module is used to acquire production line data and perform preprocessing. The feature extraction module is used to extract features from the preprocessed production line data to obtain a time-series feature vector; The strategy optimization module is used to obtain the optimal action based on the temporal feature vector; The early warning and dynamic adjustment module is used to generate early warning signals based on preset thresholds and real-time production line data, and to adjust the production line according to the early warning signals; The data acquisition and preprocessing module, the feature extraction module, the strategy optimization module, and the early warning and dynamic adjustment module are connected in sequence.
2. The intelligent management system for flexible production lines based on artificial intelligence according to claim 1, characterized in that, The data acquisition and preprocessing module includes a data acquisition unit and a preprocessing unit; The data acquisition unit is used to acquire production line data, which includes equipment operating status data, production process parameters, and order information. After cleaning the production line data, the preprocessing unit processes the cleaned data using statistical methods based on correlation analysis to obtain key features. The key features are then normalized.
3. The intelligent management system for flexible production lines based on artificial intelligence according to claim 2, characterized in that, The feature extraction module is used to construct a feature extraction model. The normalized key feature data is input into the feature extraction model to obtain a time-series feature vector.
4. The intelligent management system for flexible production lines based on artificial intelligence according to claim 3, characterized in that, The feature extraction module is based on a long short-term memory network and a convolutional neural network to construct a feature extraction model, wherein the convolutional neural network model adopts ResNet50.
5. The intelligent management system for flexible production lines based on artificial intelligence according to claim 1, characterized in that, The strategy optimization module constructs an optimization model, using the temporal feature vector as input to the optimization model to obtain the optimal action; wherein, the optimization model is obtained based on a deep Q-network.
6. The intelligent management system for flexible production lines based on artificial intelligence according to claim 5, characterized in that, The reward function of the optimization model is obtained based on production efficiency, equipment utilization rate, order fulfillment rate, and equipment load; the state space of the optimization model includes equipment status, production tasks, order requirements, and product information; the action space includes equipment parameter adjustment, production sequence adjustment, and task allocation operations.
7. The intelligent management system for flexible production lines based on artificial intelligence according to claim 1, characterized in that, It also includes a constraint module, which comprises a constraint construction unit and a constraint integration unit; The constraint construction unit is used to construct physical constraints; The constraint integration unit is used to integrate the physical constraints into the decision-making process of the optimization model.
8. The intelligent management system for flexible production lines based on artificial intelligence according to claim 7, characterized in that, The physical constraints in the constraint construction unit include equipment performance constraints, production process constraints, resource constraints, equipment safety constraints, and environmental constraints.