Anti-control stop operation method for production plan of injection molding machine

Abstract

The invention belongs to the field of industrial Internet of Things, and relates to an injection molding machine production plan anti-control stop operation method, which comprises the following steps: a plurality of sensors are installed on equipment, and the sensors collect corresponding parameters of the equipment in an operation state and upload the parameters through the Internet of Things; neural network model training is carried out through equipment parameters and operation parameters collected by a sensor, and the state of the equipment is predicted through time sequence data through a neural network model; the Internet of Things collects sensor data and performs state analysis through the neural network model to obtain a current state; when the current state accords with a preset threshold value, continuing to operate; when the current state does not accord with a preset threshold value, early warning is carried out; when early warning is carried out and a fault occurs, analysis is carried out through the central control large model, and a processing strategy is provided; and when the central control large model receives early warning information of the Internet of Things, double termination operation is carried out through the control equipment embedded control system and the mechanical arm.

Description

Technical Field

[0001] This invention belongs to the field of industrial Internet of Things, and in particular is a method for reverse control of the production plan to stop the operation of an injection molding machine. Background Technology

[0002] There is no existing technology that effectively combines video, IoT technology, sensors, and analysis models to provide a solution, and then integrates embedded terminal technology to implement a solution that stops operation based on the field conditions.

[0003] In existing technologies, the monitoring system for injection molding machines basically includes an MES module, an injection molding machine control module, a data acquisition module, and a display device. The MES module is connected to the injection molding machine control module via a data cable, the injection molding machine control module is connected to the data acquisition module via a data cable, and the data acquisition module is connected to the display device via a data cable.

[0004] However, after comprehensive analysis, the aforementioned existing technologies still have at least the following drawbacks: 1. Among the existing technologies mentioned above, none of them effectively combine IoT technology, sensors, and analysis models to provide a solution, and then combine it with embedded terminal technology to stop operation according to the field conditions.

[0005] 2. The existing technologies mentioned above cannot avoid misoperation, which can lead to unnecessary resource loss and plan delays, nor can they provide automated solutions.

[0006] 3. In the above-mentioned existing technologies, it is not possible to stop the injection molding machine and related controls after performing automated analysis.

[0007] 4. In the above-mentioned existing technologies, the operations of each part are strongly correlated and require the participation of personnel, resulting in a large demand for manpower, insufficient degree of freedom, and low efficiency. Summary of the Invention

[0008] This invention proposes a method for reverse control of production planning and shutdown of injection molding machines.

[0009] A method for reverse-controlling the production plan and stopping operation of an injection molding machine, comprising equipment, a robotic arm, an Internet of Things (IoT), a central control model, and a neural network model, including the following: Several sensors are installed on the device, which collect corresponding parameters of the device's operating status and upload them through the Internet of Things; The device parameters and operating parameters collected by sensors are used to train a neural network model, and the device status is predicted from time series data using the neural network model. The Internet of Things (IoT) collects sensor data and performs state analysis through the neural network model to obtain the current state. When the current state meets a preset threshold, the state is normal and operation continues. When the current state does not meet the preset threshold, the state is abnormal and an early warning is issued. When an early warning is issued or a malfunction occurs, the central control system's big data model analyzes the situation and provides handling strategies, including adjusting production plans, contacting maintenance, suspending equipment, and stopping production. When the central control large model receives an early warning message from the Internet of Things, it performs a dual termination operation through the embedded control system of the control device and the robotic arm.

[0010] Preferably, the sensors include temperature sensors, pressure sensors, flow sensors, robotic arm switch sensors, and mold switch sensors. The sensors collect continuous data and construct time-series data.

[0011] Preferably, the central control large model performs analysis and provides processing strategies, including: The information is synchronized to the production planning management system, and the production plan is updated accordingly, including postponement, suspension, and priority change. Synchronize information to the maintenance system for repair processing, and send corresponding maintenance instructions according to the type of problem; Send control commands to the equipment, including pausing the equipment and stopping production.

[0012] Preferably, when the central control large model receives a warning message, it performs a dual termination operation through the embedded control system of the control device and the robotic arm, including the following: The corresponding switch in the embedded control system of the control device is disconnected; The central control system uses the coordinate positioning of the robotic arm to plan the path and controls the robotic arm to perform actions. Physical buttons are used to terminate the operation.

[0013] Preferably, the central control large model performs path planning through the coordinate positioning of the robotic arm and controls the robotic arm to perform actions by pressing physical buttons to terminate the operation, including the following: Establish a 3D coordinate system and obtain the position coordinates of the buttons; Obtain the position coordinates of the robotic arm and calculate the difference between them and the position coordinates of the button. Calculate and obtain the joint angles of each joint of the robotic arm, and generate motion trajectories based on the coordinates and joint angles.

[0014] Preferably, when the robotic arm terminates the operation, the following process is also included: The central control large model sends a stop command, including the target button's number information, and retrieves the target button's standard coordinates and corresponding operating force from a pre-set database; The position of the target button is calibrated using a visual sensor to confirm whether there is a deviation between its actual position and coordinate position. If there is a deviation, it is calibrated according to the actual position. The current joint angle of the robotic arm is collected by the joint encoder installed on the robotic arm, and the Cartesian space position and attitude of the robotic arm are calculated by combining the DH parameters. Calculate and obtain the joint angles of each joint of the robotic arm, and generate motion trajectories based on the coordinates and joint angles; The robotic arm controller sends instructions to control the robotic arm to execute the planned motion trajectory and activates the force sensor feedback. When the contact force reaches the preset feedback force threshold, it is held for a preset time and then reset.

[0015] Preferably, the calculation of the joint angles of each joint of the robotic arm and the generation of motion trajectories based on the coordinates and joint angles include the following: The joint angles of each joint in the robotic arm are obtained by using the Newton-Raphson iterative formula. Based on the coordinate position of the robotic arm as the starting coordinate and the coordinate position of the target button as the target coordinate, a smooth trajectory is generated using trapezoidal velocity planning.

[0016] Preferably, the method of predicting the state of the device from time series data using the neural network model includes the following: based on several parameters acquired by the sensors, parameter training is performed through a central control model to obtain time series data; then, the neural network model is trained using the time series data; after the neural network model is trained to obtain data features, data prediction is achieved using the time series data.

[0017] Preferably, the parameter training of the central control large model includes the following: initial training of the central control large model through manual operation; then learning and fine-tuning the neural network model by combining historical time parameters, and analyzing the fine-tuned neural network model to obtain prediction results. If the prediction results achieve the expected effect, the current neural network model meets the requirements; otherwise, if the prediction results do not achieve the expected effect, the model training and fine-tuning are repeated.

[0018] Preferably, the neural network model includes any one or a combination of LSTM, RNN, and Transformer models; the neural network model includes an embedding layer, a mask layer, a hidden layer, and an attention layer, wherein the embedding layer is the input layer of the neural network model, and the input of the input layer is the sensor input parameters, including the data stream.

[0019] The present invention has the following beneficial effects: 1. This invention collects equipment operation data in real time through sensors and quickly analyzes the status through a neural network model. When the equipment malfunctions or issues a warning, the central control model can immediately trigger an emergency shutdown process, avoiding the escalation of the fault due to the lag in traditional manual response.

[0020] 2. The robotic arm reverse control uses precise coordinate positioning, path planning and force control to ensure the accuracy of emergency stop operations, avoid equipment damage or shutdown failure due to operational deviations, and minimize the impact of failures on production. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating the steps of a method for controlling the shutdown of an injection molding machine production plan according to the present invention. Figure 2 This is a schematic diagram of the working structure and state of a method for reverse control of production planning to stop operation of an injection molding machine according to the present invention; Figure 3 This is a system operation status diagram of a production planning reverse control shutdown method for injection molding machines according to the present invention. Detailed Implementation

[0022] To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions in the embodiments of this invention will be clearly described below in conjunction with the examples.

[0023] like Figure 1 As shown, this invention proposes a method for reverse control of injection molding machine production planning to stop operation, including the following: S1: Install sensors or reuse relevant sensors on smart devices. These sensors are installed inside the devices and send relevant parameters to the Internet of Things (IoT). Specific installation: Install a temperature sensor at the mold to collect the cavity temperature. A pressure sensor is installed at the screw to collect the injection pressure. A vibration sensor is installed at the motor to collect vibration acceleration. ,unit ; S2: Video and IoT technologies monitor the operating status of production equipment. Sensors upload relevant parameters via the Internet of Things, such as a sampling frequency of 1Hz, equipment parameters, rated temperature of 180℃, rated pressure of 10MPa, and operating parameters, in real time. The goal of training a neural network model is to predict device status using time-series data. ,in ; in, The device status at time t is 0 - normal, 1 - warning, 2 - fault; This includes sensor data from the past 30 time steps, such as temperature and pressure. These are the model parameters; Attn-LSTM is a long short-term memory network that incorporates an attention mechanism to enhance the weights of data at key time steps.

[0024] The sensors transmit data to the Internet of Things (IoT) in real time, where the cloud analyzes the status and issues warnings through models; on-site, relevant alarms are triggered, with the alarm content provided by the model, prompting relevant actions. S3: Warning or fault situation occurs, such as And last for 5 seconds, or The system calls upon the central control model for analysis and provides processing strategies. These strategies include suspending or stopping equipment operation, notifying the production planning management system (e.g., adjusting the original production plan of 500 units to prioritize maintenance before resuming production), and making relevant status changes to the current plan. Simultaneously, based on relevant issues, such as... Notify the maintenance system to carry out emergency handling; Among them: equipment suspension applies to the early warning status. Or a moderate fault, meaning the equipment parameters are outside the normal range but do not reach the level of "seriously threatening equipment safety or product quality".

[0025] For example: a pressure sensor detects pressure. However, it does not exceed the safety limit, such as 16MPa, or the temperature T briefly exceeds the normal range, such as the normal temperature of 180℃ and the current temperature of 190℃, and the duration is short, such as 5-10 time steps.

[0026] Pausing the equipment at this time can prevent the parameters from deteriorating further, and at the same time allow time for maintenance or parameter adjustment, such as reducing the injection speed or adjusting the temperature setting, without having to completely terminate the production plan.

[0027] Production halt applies to fault conditions. Or a serious warning, meaning that the equipment parameters are seriously out of control, threatening equipment safety, personnel safety or product quality, and cannot be restored to normal through temporary adjustments.

[0028] For example: pressure Far exceeding the safety threshold of 16MPa, mold temperature Temperatures far exceeding the normal upper limit of 250°C, or vibration acceleration It may cause damage to the mechanical structure.

[0029] At this point, production must be forcibly stopped to prevent the malfunction from escalating, such as screw breakage or mold rupture, and emergency repairs must be initiated.

[0030] S4: The relevant switches of the embedded control system of the central control large model directly control the equipment to disconnect or terminate operation; S5: The central control model controls the robotic arm to operate physical buttons, such as the emergency stop button, to terminate the operation. The robotic arm operation process is as follows: The central control model uses a 3D vision camera (Intel RealSense D435) installed at the end of the robotic arm to capture images of the button area, uses the YOLOv8 algorithm to identify the emergency stop button, and outputs its coordinates in the camera coordinate system. After conversion to the robot arm's base coordinate system, it becomes Then, the inverse kinematics function is called to calculate the target angles of the six joints. The robotic arm is controlled to move to the target position at a speed of 50 mm / s. After contact, a pressure of 30 N is applied and held for 0.5 s. After the confirmation button is triggered, it returns to its original position.

[0031] Among them, the DH parameters (Denavit-Hartenberg parameters) are core parameters used to describe the coordinate system relationship of each joint of the robotic arm and to calculate the position and attitude of the robotic arm's end effector. They are used to establish a standardized parameter system for the joint coordinate system of a serial 6-axis robotic arm, and are based on four key parameters: link length. Linkage torsion angle Linkage offset Joint angle It can accurately describe the relative position and attitude relationship between two adjacent joints or links of a robotic arm, and ultimately construct a complete coordinate transformation link from the base coordinate system of the robotic arm to the coordinate system of the end effector.

[0032] Step 1: Collect joint angles (corresponding to the 6 joints of a 6-axis robotic arm), the current rotation angle of each joint is obtained in real time through the joint encoder; Step 2: Substitute the DH parameters These are the fixed parameters of the robotic arm when it leaves the factory. For variables acquired in real time, a "DH transformation matrix" is constructed for adjacent joints. A 4×4 matrix containing translation and rotation information; Step 3: Multiply the DH transformation matrices of the 6 joints sequentially. This yields the total transformation matrix T from the robot arm's base coordinate system to the end effector's coordinate system. Step 4: Extract the Cartesian space coordinates of the robotic arm's end effector from the total transformation matrix T. The corresponding three-dimensional position and attitude of the button, such as roll angle, pitch angle, and yaw angle, are used to ensure that the button is pressed vertically at the end.

[0033] The preset feedback force threshold is 30N, and the preset time is 0.5s.

[0034] The central control big model is an integrated decision-making system used to call the state data output by the trained neural network model and combine it with the production plan to generate processing strategies; the neural network model is a dedicated state prediction model that outputs the equipment status—normal / early warning / fault—based on sensor time series data.

[0035] As an improvement to the above solution, the intelligent device includes an injection molding machine, a screw, a mold, a motor, a sensor, and a controller; The sensors are multiple and are allocated according to actual needs, usually one for each type. If the expected effect cannot be achieved, the number of corresponding sensors is increased. The multiple sensors are installed in the mold, screw, and inside the injection molding machine.

[0036] As an improvement to the above solution, the sensors include temperature sensors, pressure sensors, flow sensors, switch sensors related to robotic arm operation, and mold switch sensors, and their sensor parameters are continuous, time-series data.

[0037] As an improvement to the above solution, the central control large model trains parameters using sensor parameters obtained from sensors to acquire time-series data. This time-series data is then used to train a neural network model. The trained neural network model includes LSTM models with attention mechanisms, RNN models, Transformer models, and improved models. After obtaining data features, data prediction is achieved using the time-series data. The prediction accuracy is measured by the accuracy rate. Assessment, Objectives .

[0038] As an improvement to the above solution, the central control big model is a machine learning model, which uses a deep learning model and employs one or more of attention-enhanced neural networks, deep neural networks, and GAN models.

[0039] Deep learning models employ one or more of the following: attention-enhanced neural networks, convolutional neural networks, and GAN models.

[0040] As an improvement to the above scheme, the parameter training includes: The model was initially trained manually. Three months of historical data, including 200 fault samples, were used to allow the model to acquire certain features. Then, the model is fine-tuned and learned using historical time parameters. After using the model for a period of time, the relevant parameters are analyzed to obtain certain analytical results. If the expected results are achieved, the accuracy rate will be [not specified]. If the parameter predictions are accurate, then the model can be used; otherwise, the model will not be accurate. The model is then fine-tuned through training.

[0041] As an improvement to the above solution, the sensor obtains data parameters, uploads them to the Internet of Things (IoT), analyzes the data parameters through a model, obtains the device operating status through feature extraction and related features, analyzes the results in the cloud, and transmits them to the production system through the IoT and the operation of physical buttons and related switches by the robotic arm. Based on different parameters, the central control model is called to give different control actions. If the operation is normal ( If the operation is abnormal, then continue working. or If so, the production planning management system will be adjusted, and information will be sent to the maintenance system for maintenance processing; The maintenance system handles issues through manual verification or directly through the Internet of Things (IoT).

[0042] As an improvement to the above solution, the central control big model analyzes sensor data parameters, issues alarms according to relevant rules, processes the alarm results using different time steps to avoid misoperation, and sends the corresponding alarm information to the central control big model for processing. The central control system determines, based on the situation, whether to notify the maintenance system, directly issue a stop command to the embedded control system of the equipment, or directly instruct the central control system to perform relevant processing, such as terminating the operation of physical buttons, or instructing the robotic arm actuator to stop the relevant components of the injection molding machine through its operation button. The algorithm flow for the robotic arm to perform the termination operation is as follows: 1) Receive the stop command from the central control large model, including the target button ID; 2) Retrieve the standard coordinates of the key from the database. and operating force ; 3) Collect the current joint angle of the robotic arm Calculate the position in Cartesian space; 4) The motion trajectory is generated using trapezoidal velocity planning, and the total travel time t ; 5) Send motion commands and simultaneously activate force sensor feedback. When the contact force reaches... Hold for 0.5 seconds and then reset.

[0043] The core parameter of the robotic arm control algorithm is: position loop proportional gain. Integral time Differential time Force ring proportional gain This ensures a smooth operation without overshoot.

[0044] As an improvement to the above scheme, the model is a neural network model, specifically an LSTM model, an RNN model, or a Transformer model. The neural network model has an embedding layer, a mask layer, a hidden layer, and an attention layer. The embedding layer is the input layer of the neural network model, and the input of the input layer is the sensor input parameters, which is a data stream. The neural network model input layer uses a time step method for input. Depending on the sensor, the time step of the model input layer is 10-100, and the length of each time step is one time series data. The neural network model uses an LSTM or RNN network, or a combination of both. The output layer is processed to predict the output at that time step. The output layer network is a softmax layer, or, depending on the situation, a fully connected layer network is used for prediction, or a combination of multiple layers in the network structure; the output layer is an LSTM prediction layer that predicts the output.

[0045] like Figure 2 , Figure 3 As shown, the working process of this invention is as follows: During use, sensors are installed or related sensors are reused on the smart device, which includes an injection molding machine. The sensors reuse related sensors and include multiple sensors installed inside the device. The sensors send relevant parameters to the Internet of Things. Video and IoT technologies are used to monitor the operating status of production equipment. Sensors will upload relevant parameters through the Internet of Things. Based on the equipment parameters and operating parameters, a neural network model will be trained. The purpose of the training is to predict the equipment status through time series data. The sensors transmit data to the Internet of Things (IoT) in real time, where the cloud analyzes the status and issues warnings through models; on-site, relevant alarms are triggered, with the alarm content provided by the model, prompting relevant actions. In the event of an early warning or malfunction, the central control system's large-scale model is invoked for analysis, and a handling strategy is provided. The handling strategy includes suspending or stopping the equipment, notifying the production planning management system to make relevant status changes to the current plan, and simultaneously notifying the maintenance system to carry out emergency handling based on the relevant issues. The central control system directly controls the relevant switches of the embedded control system of the equipment to disconnect or terminate operation; the central control system controls the operation of the physical buttons of the robotic arm to terminate the operation.

[0046] Example 2: Based on an improved neural network model The neural network model in this embodiment 2 is an LSTM model that incorporates an attention mechanism. The input layer of the improved neural network model is the sensor input parameters, such as T, P, and F. The input layer input is a data stream, which is input in the form of time steps. Depending on the sensor, the time step of the model input layer is 30, and the length of each time step is 1 second of time series data. The neural network model uses a network structure including an LSTM layer with 64 neurons; an attention layer that calculates weights for 30 time steps; a fully connected layer with 3 neurons, corresponding to 3 states; an output layer activated by softmax that predicts the output result and state probability distribution for that time step. The output layer network is a softmax layer, or, depending on the situation, a fully connected layer network is used for prediction, or a combination of multiple layers; the output layer is an LSTM prediction layer that predicts the output.

[0047] Example 3: Based on an improved deep learning model The difference between this embodiment and Embodiment 1 is as follows: The deep learning model in this embodiment employs one or more of the following: attention-enhanced neural networks, convolutional neural networks, and GAN models.

[0048] Example 4: A method for reverse control of injection molding machine production plan to stop operation, comprising the following steps: S1: Install sensors on smart devices and reuse related sensors. The smart devices include injection molding machines. S2: The sensors include multiple sensors installed inside the device, and the sensors send relevant parameters to the Internet of Things; S3: Video and IoT technologies monitor the operating status of production equipment. Sensors will upload relevant parameters through the Internet of Things. Based on the equipment parameters and operating parameters, a neural network model will be trained. The purpose of the training is to predict the equipment status through time series data. S4: The sensor transmits data to the Internet of Things in real time, and the cloud uses models to analyze the status and issue warnings. S5: Relevant alarms will be triggered on-site, with the alarm content provided by the model, and relevant actions will be taken accordingly; S6: In the event of a warning or malfunction, the central control model is invoked for analysis, and a handling strategy is provided. The handling strategy includes suspending or stopping the equipment, notifying the production planning management system to make relevant status changes to the current plan, and notifying the maintenance system to carry out emergency handling based on the relevant issues. S7: The relevant switches of the embedded control system of the central control large model directly control the equipment to disconnect or terminate operation; S8: The robotic arm operates physical buttons via the central control large model to perform a termination operation. The core algorithm of the robotic arm's reverse control is inverse kinematics solving. For a 6-DOF robotic arm, a numerical iteration method is used: given the end-effector pose... Using the Newton-Raphson iterative formula Calculate the joint angles, where This is the pseudoinverse of the Jacobian matrix. Let be the pose error vector, and the iteration termination condition be... .

[0049] The software implementation of the robotic arm's reverse control is as follows: A control node is developed based on the ROS system, including a tf coordinate transformation module, a Moveit! motion planning module, and a force control feedback module. The motion planning module uses the RRT algorithm to plan a collision-free path, and the force control module achieves force tracking control through a ROS control loop (500Hz). The code snippet is as follows: cpp voidforceControlCallback(constWrenchStamped::ConstPtr&msg) { floatcurrent_force=msg->wrench.force.z; floaterror=target_force-current_force; floatcmd=Kp*error+Kd*(error-last_error); last_error = error; publishVelocityCommand(cmd); } As an improvement to the above scheme, the model analysis and processing includes inputting relevant features through time series data and manual operation, and training a neural network model, such as an LSTM model. After a period of training, a model is obtained. The model has an embedding layer, an output layer, a hidden layer, and an attention layer. The output predicted device state sequence has a length of 10-100.

[0050] As an improvement to the above scheme, the model training adopts a machine learning model, which can be a deep learning model, or one or more of attention-enhanced neural networks, deep neural networks, and GAN models.

[0051] As an improvement to the above solution, the model is processed by an embedded system. The input of the embedded system is data from multiple sensors, and the output is the status. The output of the model is operation and early warning. Depending on the situation, the early warning is divided into different levels, and the corresponding solution for each level is called according to the early warning level.

[0052] As an improvement to the above solution, the model includes analysis and early warning rules, which are divided into different levels. Based on the rules, the embedded system and machine learning model are called to directly output the stop state or directly call the model center. The early warning rules are divided into several levels: severe, medium, and early warning. Different levels provide different solutions.

[0053] Example 5: Application in the injection molding production of automobile bumpers This embodiment, focusing on the injection molding production scenario of automobile bumpers, details the implementation process of the injection molding machine production plan reverse control shutdown method: 1. Equipment and sensor deployment: Two temperature sensors are installed in the mold cavity of a bumper-specific injection molding machine, model HTF1600, to collect data. Temperature range: 0-300℃; a pressure sensor is installed at the screw head to collect data. Range 0-20MPa; hydraulic system, equipped with a flow sensor to collect data. Range 0-50 L / min; motor shaft, equipped with a vibration sensor to collect data. Range 0-10 m / s²; Install sensors; Reuse the current sensor of the original controller of the equipment to collect data. , range 0-50A.

[0054] The robotic arm is a 6-axis collaborative robot, model: UR10e, with a 3D vision sensor at the end effector, a resolution of 1280×720, and a force sensor with a range of 0-100N.

[0055] 2. Data Collection and Upload: The sensor collects data at a frequency of 1Hz and uploads it to a cloud server via an industrial Ethernet IoT module (ME3616). The data format is as follows: ,in This is a timestamp. Robotic arm joint data, including position and velocity, is uploaded in real-time via EtherCAT.

[0056] 3. Model Training: Using production data from the past 6 months, including 120 fault records such as abnormal mold temperature and pressure overload, an LSTM model with an attention mechanism was trained: Input: 7D sensor data from the past 30 time steps ( ) Output: The device status at any given time, 0 - Normal, 1 - Warning, 2 - Fault Training objective: loss function accuracy Model Deployment: The trained model is deployed on a cloud AI platform (GPU: NVIDIA A10).

[0057] In one feasible embodiment, the accuracy .

[0058] 4. Real-time monitoring and early warning: During the production process, the cloud-based model receives sensor data every second and predicts the state for the next 5 seconds: like The system is operating normally, and the Production Planning and Management System (MES) maintains the original plan, producing 500 bumpers that day.

[0059] like Warning, such as forecast The pressure will reach 18MPa in 5 seconds, exceeding the threshold of 16MPa: the central control model triggers an early warning, and the on-site HMI displays "Pressure warning, it is recommended to reduce speed", while notifying the MES system to prepare to adjust the subsequent production sequence.

[0060] 5. Fault Handling and Reverse Control: When detected Faults, such as and : The central control unit immediately sends a stop command to the injection molding machine's embedded control system PLC (S7-1200) to cut off the power to the hydraulic pump.

[0061] Simultaneously initiate the robotic arm's reverse control process: a) The visual sensor captures images of the control panel, identifies the location of the emergency stop button, and obtains the coordinates in the base coordinate system after coordinate transformation. ; b) Call Moveit! to plan the movement from the current position. The trajectory to the target location is planned in less than 0.5 seconds. c) The robotic arm moves at a speed of 100 mm / s, and switches to force control mode when it approaches, with a target force of 30 N. d) After contact, hold for 0.5 seconds, the confirmation button is triggered, a feedback signal is sent through the PLC, and then the device returns to the safe position.

[0062] The MES system is notified: "Equipment malfunction, the current batch is suspended, the plan is adjusted to prioritize repair, and subsequent batches will be postponed."

[0063] Send fault details to the maintenance system: "Mold temperature abnormal." "Vibration exceeds the standard. It is recommended to check the positioning of the heating element and the mold." The maintenance system automatically generates a work order and assigns it to the maintenance team.

[0064] 6. Recovery and Plan Adjustments: After replacing the heating element and repositioning the mold, the maintenance system reported "repaired" upon completion of the repair. The central control unit removed the stop command, the MES system updated the production plan, and the remaining 450 pieces were completed in two shifts. The injection molding machine then restarted production.

[0065] This embodiment achieves rapid fault response in the automotive bumper production process through precise sensor deployment, advanced predictive models, and automated countermeasures, reducing downtime compared to the traditional manual inspection mode.

[0066] It should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for controlling the shutdown of an injection molding machine's production plan, characterized in that, This includes injection molding machine equipment, robotic arms, the Internet of Things, central control system models, and neural network models, encompassing the following: Several sensors are installed on the device, which collect corresponding parameters of the device's operating status and upload them through the Internet of Things; The device parameters and operating parameters collected by sensors are used to train a neural network model, and the device status is predicted from time series data using the neural network model. The Internet of Things (IoT) collects sensor data and performs state analysis through the neural network model to obtain the current state. If the current state meets the preset threshold, the state is normal and operation continues; if the current state does not meet the preset threshold, the state is abnormal and a warning is issued. When an early warning is issued or a malfunction occurs, the central control system's big data model analyzes the situation and provides handling strategies, including adjusting production plans, contacting maintenance, suspending equipment, and stopping production. When the central control large model receives an early warning message from the Internet of Things, it performs a dual termination operation through the embedded control system of the control device and the robotic arm.

2. The injection molding machine production plan reverse control stop operation method according to claim 1, characterized in that, The sensors include temperature sensors, pressure sensors, flow sensors, robotic arm switch sensors, and mold switch sensors. The sensors collect continuous data and construct time-series data.

3. The injection molding machine production plan reverse control stop operation method according to claim 1, characterized in that, The central control model analyzes and provides processing strategies, including: The information is synchronized to the production planning management system, and the production plan is updated accordingly, including postponement, suspension, and priority change. Synchronize information to the maintenance system for repair processing, and send corresponding maintenance instructions according to the type of problem; Send control commands to the equipment, including pausing the equipment and stopping production.

4. The injection molding machine production plan reverse control stop operation method according to claim 1, characterized in that, When the central control model receives an early warning message from the Internet of Things, it performs a dual termination operation through the embedded control system of the control device and the robotic arm, including the following: The central control system sends instructions to the embedded control system of the control equipment, and the embedded control system of the control equipment controls the corresponding switch in the injection molding machine to disconnect. The central control system uses the coordinate positioning of the robotic arm to plan the path and controls the robotic arm to perform actions. Physical buttons are used to terminate the operation.

5. The injection molding machine production plan reverse control stop operation method according to claim 1, characterized in that, The central control model uses the coordinate positioning of the robotic arm for path planning and controls the robotic arm to execute actions via physical buttons to terminate the operation, including the following: Establish a 3D coordinate system and obtain the position coordinates of the buttons; Obtain the position coordinates of the robotic arm and calculate the difference between them and the position coordinates of the button. Calculate and obtain the joint angles of each joint of the robotic arm, and generate motion trajectories based on the coordinates and joint angles.

6. The injection molding machine production plan reverse control stop operation method according to claim 5, characterized in that, When the robotic arm terminates, the specific process includes the following: The central control large model sends a stop command, including the target button's number information, and retrieves the target button's standard coordinates and corresponding operating force from a pre-set database; The position of the target button is calibrated using a visual sensor to confirm whether there is a deviation between its actual position and coordinate position. If there is a deviation, it is calibrated according to the actual position. The current joint angle of the robotic arm is collected by the joint encoder installed on the robotic arm, and the Cartesian space position and attitude of the robotic arm are calculated by combining the DH parameters. Calculate and obtain the joint angles of each joint of the robotic arm, and generate motion trajectories based on the coordinates and joint angles; The robotic arm controller sends instructions to control the robotic arm to execute the planned motion trajectory and activates the force sensor feedback. When the contact force reaches the preset feedback force threshold, it is held for a preset time and then reset.

7. A method for reverse-controlling the production plan to stop operation of an injection molding machine according to any one of claims 5 and 6, characterized in that, The calculation obtains the joint angles of each joint of the robotic arm, and generates a motion trajectory based on the coordinates and joint angles, including the following: Calculate and obtain the joint angles of each joint in the robotic arm; Based on the coordinate position of the robotic arm as the starting coordinate and the coordinate position of the target button as the target coordinate, a smooth trajectory is generated using trapezoidal velocity planning.

8. The injection molding machine production plan reverse control stop operation method according to claim 1, characterized in that, The neural network model predicts the state of a device from time-series data, including the following: based on several parameters acquired by sensors, parameters are trained using a central control model to obtain time-series data; then, the neural network model is trained using the time-series data; after the neural network model is trained to obtain data features, data prediction is achieved using the time-series data.

9. A method for reverse control of injection molding machine production planning and shutdown according to claim 8, characterized in that, The parameter training of the central control large model includes the following: the central control large model is initially trained by manual operation; then the neural network model is learned and fine-tuned by combining historical time parameters, and the fine-tuned neural network model is analyzed to obtain the prediction results. If the prediction results achieve the expected effect, the current neural network model meets the requirements; otherwise, if the prediction results do not achieve the expected effect, the model is retrained and fine-tuned.

10. The injection molding machine production plan reverse control stop operation method according to claim 1, characterized in that, The neural network model includes any one or a combination of LSTM, RNN, and Transformer models; the neural network model includes an embedding layer, a mask layer, a hidden layer, and an attention layer. The embedding layer is the input layer of the neural network model, and the input of the input layer is the sensor input parameters, including the data stream.

Images