Injection molding machine and manipulator collaborative in-mold ejection synchronous control system and method

Abstract

The invention relates to an injection molding machine and manipulator cooperative in-mold ejection synchronous control system and method, and relates to the technical field of injection molding automation, and the system comprises a bidirectional communication module which is configured to establish a bidirectional data link between an injection molding machine ejector pin device and a manipulator control system, wherein the signal output end of the injection molding machine ejector pin encoder is directly connected to the signal input end of the manipulator control system and is used for transmitting real-time position data, real-time speed data and operation state parameters of an ejector pin. Real-time data interaction between the ejector pin device of the injection molding machine and the manipulator control system is achieved through the two-way communication module, accurate synchronization of ejector pin and manipulator actions is ensured, the problem of interference or waiting under traditional fixed time sequence control is avoided, and the dynamic position following module and the self-adaptive speed matching module work cooperatively, so that the dynamic position tracking precision is improved. The flexibility and response speed of collaborative operation are further improved, and invalid motion is reduced.

Description

Technical Field

[0001] This application relates to the field of injection molding automation technology, and in particular to an in-mold ejection synchronous control system and method for the collaboration of an injection molding machine and a robotic arm. Background Technology

[0002] In automated injection molding production, the collaborative operation of the injection molding machine and the robotic arm is a key element in achieving efficient and precise production. In traditional injection molding processes, the collaboration between the robotic arm and the injection molding machine typically relies on a pre-set, fixed timing control. That is, after the injection molding machine completes the product molding, it triggers the robotic arm to perform a part-picking action via an I / O signal. The robotic arm moves to the mold at a fixed speed or along a preset trajectory to pick up the product, and then the injection molding machine's ejector pins eject the product according to an independent program. While this control method achieves a degree of automation, its inherent limitations become increasingly apparent when facing the demands of high-precision and high-efficiency production.

[0003] In existing technologies, the movements of the robotic arm and the ejector pin device of the injection molding machine lack a real-time synchronization mechanism. During ejection, the product position is prone to shift, and fixed timing control cannot adapt to dynamic changes in ejector pin speed, leading to interference with the robotic arm's part removal or excessively long waiting times. Simultaneously, traditional I / O signal interaction suffers from delays, making it difficult to achieve precise matching between ejector pin position and robotic arm movements. Furthermore, in the confined space within the mold, asynchronous movements can cause equipment collisions or product deformation. These problems are particularly prominent in high-speed injection molding scenarios, severely impacting production efficiency and product quality. To address these issues, we propose a collaborative in-mold ejection synchronization control system and method for the injection molding machine and robotic arm. Summary of the Invention

[0004] The purpose of this application is to provide a synchronous control system and method for in-mold ejection in collaboration between an injection molding machine and a robotic arm. This technical solution solves the problems mentioned above, such as the lack of real-time synchronization between the movement of the robotic arm and the ejector pin of the injection molding machine, the difficulty of adapting to speed changes with fixed timing control, the delay in IO signal interaction, the small in-mold space which easily leads to equipment collision or product deformation, and the difficulty in meeting the high-precision and high-efficiency production requirements during high-speed injection molding.

[0005] Firstly, the in-mold ejection synchronous control system for the coordinated operation of an injection molding machine and a robotic arm provided in this application adopts the following technical solution: A synchronous control system for in-mold ejection in collaboration between an injection molding machine and a robotic arm includes: The two-way communication module is used to establish a two-way data link between the injection molding machine ejector device and the robot control system, and to collect key data of the ejector device in real time and transmit the key data synchronously to the downstream module. The dynamic position following module is used to receive key data through the bidirectional data link, calculate the real-time position deviation value, and generate the robot's motion speed command and acceleration command by combining the feedforward compensation amount and PID adjustment amount. The adaptive speed matching module is configured to identify the current motion stage of the ejector pin and switch the follow control mode of the robot arm according to different motion stages. The predictive path planning module is configured to store and analyze historical motion data of the ejector pin, build a dynamic model of ejector pin motion, predict the future position change trend of the ejector pin based on the model and the current state, and optimize the path planning of the robot arm. The safety redundancy module is configured to set the ejector pin position deviation threshold and the virtual safety space boundary of the mold cavity, and to perform early warning, pause or avoidance actions based on the position deviation and space intrusion.

[0006] Preferably, the bidirectional data link uses the real-time Ethernet protocol for data exchange, with a transmission period of less than 10 milliseconds; The key data of the ejector device is collected in real time through the ejector encoder. The signal output end of the ejector encoder is directly connected to the signal input end of the robot control system to transmit the key data of the ejector device collected in real time. The key data includes: position data, speed data and operating status parameters. The position data includes: the absolute position and relative displacement of the ejector pin; The velocity data includes: instantaneous velocity and acceleration; The operating status parameters include: status codes that indicate whether the ejector pin is in a startup, running, stopped, or faulty state.

[0007] Preferably, the dynamic position following module calculates the real-time position deviation value using the following method: The relative position rules between the ejector pin and the robot are pre-configured in the robot control system; The dynamic position following module receives key data from the ejector pin device in real time; the robot control system obtains the real-time actual position of the robot through its own position feedback device. Based on the collected key data and the real-time actual position of the robot, the difference between the two actual positions is calculated through a closed-loop control algorithm to obtain the real-time position deviation value; and the speed difference is calculated at the same time. The feedforward compensation amount is calculated based on the received real-time velocity data and real-time acceleration data of the ejector pin; The PID control is obtained by processing the real-time position deviation value through proportional, integral and derivative components; the final generated robot motion speed command and acceleration command are used to control the robot servo driver, so that the robot end effector suction cup, gripper and ejector pin end maintain a preset synchronous motion relationship and constant spatial distance.

[0008] Preferably, in the adaptive speed matching module, identifying the movement phase of the ejector pin specifically includes: Analyze the real-time speed data and rate of change of the pin. When the rate of change of speed is positive and exceeds the acceleration judgment threshold, it is determined to be the acceleration stage. When the absolute value of the rate of change of speed is lower than the constant speed judgment threshold, it is determined to be the constant speed stage. When the rate of change of speed is negative and lower than the deceleration judgment threshold, it is determined to be the deceleration stage. The specific steps for switching the robotic arm's follow control mode are as follows: During the pin acceleration phase, the robotic arm initiates trapezoidal acceleration based on preset initial acceleration and maximum speed parameters; During the constant speed phase of the ejector pin, the speed command value of the robot arm is set to be equal to the current linear speed value of the ejector pin; During the ejector deceleration phase, the robot arm calculates the deceleration point in advance based on the predicted ejector stopping position and the preset safety margin time, and then decelerates to the part removal completion position.

[0009] Preferably, in the predictive path planning module, the dynamic model of the thimble motion is constructed using a system identification method based on historical data. The dynamic model of the thimble motion includes at least fitting functions or state-space equations for the thimble displacement-time relationship, velocity-time relationship, and acceleration-time relationship. The prediction of the future position change trend of the ejector pin refers to extrapolating and calculating the expected position sequence of the ejector pin within several future control cycles using the current position and speed of the ejector pin and the dynamic model of the ejector pin motion. Optimizing the motion path planning of a robotic arm refers to recalculating the spatial trajectory and corresponding time-velocity curve of the robotic arm's end effector from its current position to the target grasping position based on the expected position sequence and using optimal time or optimal energy consumption as criteria.

[0010] Preferably, the pin position deviation threshold in the safety redundancy module includes a first threshold and a second threshold, wherein the first threshold is set to 80%-90% of the allowable position deviation range, and the second threshold is set to 110%-120% of the allowable position deviation range; The virtual safety space boundary is pre-set in the robot control system based on the three-dimensional CAD model data of the mold, and the boundary maintains a safety distance of not less than 5mm from the actual mold surface and moving parts; The safety redundancy module operates independently of the main control loop and is equipped with an independent watchdog circuit and software monitoring thread to ensure that safety protection actions can still be performed when the system main control fails.

[0011] Preferably, the bidirectional communication module further includes a position detection replacement unit; the position detection replacement unit is configured as follows: Scheme A, using a high-precision laser displacement sensor, which is installed on a fixed base, emits a laser beam to irradiate the end of the ejector pin or related components, and obtains the real-time position data of the ejector pin by measuring the reflected light signal in a non-contact manner; Option B utilizes a pressure sensor installed in the ejector cylinder of the injection molding machine to monitor the hydraulic pressure in real time, and combines it with a pre-calibrated cylinder pressure-piston displacement relationship model to indirectly calculate and derive the real-time position data of the ejector pin. The position data acquired by the position detection replacement unit is transmitted to the robot control system via the bidirectional data link.

[0012] Preferably, the bidirectional communication module further includes a communication mode substitution unit; the communication mode substitution unit is configured as follows: Option C uses a Wi-Fi 6 wireless communication module and a 5G NR industrial module that conform to the IEEE 802.11ax standard to replace the wired Ethernet physical connection, thereby realizing wireless data transmission between the injection molding machine and the robot control system. Option D involves deploying an OPCUA server on the injection molding machine control system side and an OPCUA client on the robot control system side. By encapsulating and exchanging ejector pin position, speed, and status parameter data according to the OPCUA information model specification, standardized data interaction can be achieved across different brands of equipment.

[0013] Preferably, the dynamic position following module further includes a control algorithm replacement unit; the control algorithm replacement unit is configured as follows: Scheme E uses a fuzzy control algorithm to replace the PID regulation. The fuzzy control algorithm takes the real-time position deviation value and its rate of change as input variables, and performs inference through a predefined fuzzy rule base and membership function to output a fuzzy quantity of the robot speed adjustment. After defuzzification processing, a precise speed command correction value is obtained. Scheme F adopts an adaptive control model based on machine learning. During operation, the adaptive control model continuously collects historical position deviation data, robot response data, and ejector pin motion data. It uses online learning algorithms to dynamically optimize the feedforward compensation amount and / or PID control parameters or fuzzy rule weights to adapt to different ejector pin motion characteristics and load changes.

[0014] Secondly, the in-mold ejection synchronization control method for the coordinated operation of an injection molding machine and a robotic arm provided in this application adopts the following technical solution: A method for synchronous control of in-mold ejection in collaboration between an injection molding machine and a robotic arm, used to realize the synchronous control system for in-mold ejection in collaboration between an injection molding machine and a robotic arm, includes the following steps: S1. Establish a bidirectional data link: Establish a bidirectional data communication connection between the injection molding machine ejector pin device and the robot control system; S2. Real-time data transmission and acquisition: Through the bidirectional data link, the robot control system acquires key data of the injection molding machine ejector pin in real time; S3, Dynamic Position Following Control: Calculate the real-time position deviation value based on the real-time position data of the ejector pin and the relative position relationship of the preset target, apply the closed-loop control algorithm, and generate and output the robot arm motion command by combining the feedforward compensation and PID control parameters. S4. Adaptive speed matching control: Monitors and identifies the movement phase of the ejector pin. During the acceleration phase, the robot arm is controlled to start accelerating according to the trapezoidal speed curve. During the constant speed phase, it maintains the same speed as the ejector pin. During the deceleration phase, it decelerates synchronously to the position where the part is picked up. S5. Predictive Path Planning: Store and analyze historical ejector pin motion data to build a dynamic model, predict the future motion trajectory of the ejector pin based on the current state, calculate the optimal picking path of the robot arm and plan it in advance; S6. Security Monitoring and Redundancy Protection: Set position deviation protection thresholds and virtual security space boundaries, monitor position deviation and space intrusion in real time, and execute early warning, pause, alarm or avoidance actions.

[0015] In summary, this application includes at least one of the following beneficial technical effects: This invention enables real-time data interaction between the injection molding machine's ejector pin device and the robot control system through a two-way communication module. This ensures precise synchronization of the ejector pin and robot's movements, avoiding interference or waiting problems inherent in traditional fixed-sequence control. The collaborative work of the dynamic position following module and the adaptive speed matching module allows the robot to adjust its trajectory and speed based on the real-time movement of the ejector pin, further enhancing the flexibility and response speed of collaborative operations. The predictive path planning module analyzes historical data to predict the future position of the ejector pin, optimizing the robot's part-picking path and reducing unnecessary movements. The safety redundancy module effectively prevents equipment collisions or product deformation, ensuring production safety and significantly improving the automation level and product quality of injection molding production. Attached Figure Description

[0016] Figure 1 This is a system framework diagram of the present invention; Figure 2 This is a flowchart of the method of the present invention. Detailed Implementation

[0017] The following is in conjunction with the appendix Figure 1 -Appendix Figure 2 This application will be described in further detail below.

[0018] Example 1: An in-mold ejection synchronous control system for the coordinated operation of an injection molding machine and a robotic arm, referring to... Figure 1 As shown, this includes: ensuring the robot arm maintains precise following throughout the movement of the ejector device by working in collaboration with multiple modules, while also taking into account path optimization and safety protection.

[0019] The system establishes a bidirectional real-time communication link through a bidirectional communication module, enabling the robot control system to directly acquire the raw motion data of the injection molding machine's ejector pin. This link uses a real-time Ethernet protocol (such as EtherCAT or PROFINETIRT) as the physical transmission layer, ensuring a data exchange cycle strictly less than 10 milliseconds, eliminating the latency issues of traditional indirect communication. The signal output of the injection molding machine's ejector pin encoder is directly connected to the robot controller's dedicated high-speed input port via shielded twisted-pair cable or fiber optic cable. The transmitted content includes the absolute position coordinates of the ejector pin (referenced to the mold's zero point), instantaneous linear velocity with millimeter-level accuracy, and operating status codes mapped by the status register (e.g., 0x01 for startup, 0x02 for operation, and 0x04 for fault shutdown). This direct-connection architecture avoids data packet loss and timing distortion caused by traditional PLC relay.

[0020] The dynamic position following module is the core algorithm carrier of synchronous control. The relative position rules between the ejector pin and the robot arm are pre-configured in the robot control system. For example: synchronous motion mode: the position difference between the two is ≤ ±0.5mm, and the speed difference is ≤ ±2mm / s; fixed-distance following mode: the center of the robot arm's suction cup is always maintained 10mm in front of the ejector pin tip (along the ejector pin's movement direction). Preset parameters such as the preset position difference threshold and safety distance are stored in the control system's registers as a reference for subsequent deviation calculations. The robot control system acquires the robot arm's real-time actual position through its own position feedback device (such as a robot arm joint encoder or linear grating). The dynamic position following module continuously receives real-time position stream data from the ejector pin. Based on the collected ejector pin position and the acquired robot arm position, it calculates the actual position difference ΔP using a closed-loop control algorithm. The calculation formula is: ΔP = Ejector pin real-time position - Robot arm real-time position - Preset safety distance; simultaneously, it calculates the speed difference ΔV, with the formula: ΔV = Ejector pin real-time speed - Robot arm real-time speed. The deviation value is input to the closed-loop controller, and its output is generated by a combination of feedforward compensation and PID control. The feedforward compensation is dynamically calculated based on the real-time acceleration signal of the ejector pin and is used to compensate for system inertial lag, specifically as follows: Based on the real-time velocity and acceleration changes of the ejector pin (calculated by differentiating the position data), identify the current motion stage of the ejector pin (acceleration stage, constant velocity stage, or deceleration stage). If it is in the acceleration phase: based on the acceleration value (a) of the thimble, using the kinematic formula (ΔP_feedforward=0.5×a×t) 2 (t is the prediction time) Calculate the feedforward compensation position and adjust the start-up speed of the robot in advance to avoid lag; If the speed is constant: use the current speed of the ejector pin as the feedforward speed command to ensure that the speed of the robot arm and the ejector pin are matched; If in the deceleration phase: Based on the deceleration acceleration (negative acceleration) of the ejector pin, pre-calculate the position change trend during the deceleration process, and generate the deceleration feedforward of the robot in advance to avoid overshoot.

[0021] The PID control variable responds quickly to deviations through the proportional element, eliminates static errors through the integral element, and suppresses overshoot through the derivative element. Specifically: Proportional (P) adjustment: For the real-time position deviation ΔP calculated in step 3, the adjustment amount (U_p=Kp×ΔP) is output through the proportional coefficient (Kp) to quickly respond to the current deviation; Integral (I) regulation: Integral accumulation (U_i=Ki×∫ΔPdt) is performed on persistent steady-state deviations (such as small offsets caused by mechanical backlash) to eliminate long-term errors; Differential (D) control: The control amount (U_d = Kd × dΔP / dt) is calculated based on the rate of change of deviation (dΔP / dt), the deviation trend is predicted, and overshoot is suppressed (e.g., to prevent the robot from exceeding the target position due to inertia).

[0022] Total PID regulation: The final PID correction is obtained by superimposing the regulation values ​​of P, I, and D (U_pid = U_p + U_i + U_d).

[0023] The feedforward compensation and PID correction are superimposed to obtain the overall adjustment command for the robot's movement. This overall adjustment command is then converted into a control signal for the robot's drive system (such as a servo motor pulse command or analog voltage), dynamically adjusting the robot's speed. The robot's drive system drives the robot's movement according to the control signal, adjusting its position and speed in real time. Each time the control system completes a sampling cycle (e.g., 1ms), it repeats the above steps, continuously acquiring new position data, calculating deviations, and updating the feedforward compensation and PID adjustment, forming a closed-loop control cycle of "acquisition-calculation-adjustment-execution-reacquisition" until the ejector pin completes its ejection action, ending the robot's work.

[0024] The adaptive speed matching module endows the system with the ability to intelligently identify the movement phase of the ejector pin. This module determines the movement phase by analyzing the rate of change of ejector pin speed (dv / dt) in real time: when dv / dt is continuously greater than 0.5 m / s... 2 A period exceeding 20ms is marked as an acceleration phase; the absolute value of dv / dt is less than 0.1m / s. 2 The velocity is considered constant if it lasts for 50 ms; dv / dt is less than -0.3 m / s. 2 This triggers the deceleration phase determination. For different phases, the robotic arm employs differentiated control strategies: in the initial stage of pin acceleration, the robotic arm follows a preset S-shaped curve (acceleration 2m / s²).2 → Maximum speed 1.5m / s) Start 100ms in advance to avoid initial lag; during the constant speed phase, lock the speed command with the linear velocity of the ejector pin to achieve "floating follow" with zero relative velocity; during the deceleration phase, predict the stopping position based on the ejector pin displacement model and start the smooth deceleration of the robot arm at a distance of 150mm from the target point to ensure that the part is picked up 50ms before the ejector pin comes to a complete stop.

[0025] The predictive path planning module constructs a dynamic model of thimble motion through machine learning. The specific implementation method of the dynamic model of thimble motion is as follows: The recursive least squares (RLS) algorithm is used for system identification of the pin motion model. The specific steps include: Data acquisition phase: Continuously acquire at least 500 sets of historical data for complete ejection cycles. Each set of data includes a timestamp (accuracy 1ms), absolute position of ejector pin (mm), real-time speed (m / s), hydraulic pressure of ejection cylinder (MPa), and mold temperature (°C). The data sampling frequency is consistent with the bidirectional communication cycle (≤10ms).

[0026] Model structure determined: Based on feature analysis of the collected data, the pin motion model adopts a second-order linear state-space equation: Where the state variable x(k) = [s(k), v(k)] T (Displacement, velocity), input u(k) is the cylinder pressure, output y(k) = s(k) (displacement), A, B, C are the parameter matrices to be identified, w(k) and v(k) are the process noise and measurement noise respectively (variance calibrated to 0.01 mm through experiments). 2 ).

[0027] Parameter identification process: The parameters of the above state-space equations are estimated using the RLS algorithm, and the forgetting factor λ = 0.98 is iteratively updated until the mean square error (MSE) between the model output and the actual displacement is ≤ 0.5 mm. 2 Complete the model calibration.

[0028] For high-precision scenarios, a dynamic model is constructed using a Long Short-Term Memory (LSTM) network. Network structure: The input layer contains 4 neurons (pressure, temperature, current displacement, current velocity), the hidden layer consists of 2 LSTM layers (32 neurons per layer), and the output layer contains 1 neuron (future displacement); Training process: The Adam optimizer was used with a learning rate of 0.001, a batch size of 32, and 500 training epochs. Overfitting was avoided by using early stopping (the validation set MSE stopped if it did not decrease for 10 consecutive epochs). Predictive execution: When the ejector pin starts, the current initial state (pressure p0, temperature T0, initial displacement s0, initial velocity v0) is input, and the model outputs the position sequence for the next N control cycles (N=30, corresponding to 300ms). Prediction error ≤ 1mm.

[0029] The system continuously collects historical motion data of the ejector pin, including displacement, velocity, hydraulic pressure, and ambient temperature. A recurrent neural network (RNN) is used to train and generate the displacement-time state-space equation for the ejector pin. Based on the dynamic characteristics of the ejector pin's motion, the single ejection cycle is typically 200-500 ms. Therefore, the time window length is set to 30 consecutive sampling points, corresponding to 300 ms of physical time. This is because the sampling period of the bidirectional communication module does not exceed 10 ms, meaning each sampling point is spaced 10 ms apart. The window sliding step is set to one sampling point to ensure the model can capture changes in motion trends in real time. The input features comprise four physical dimensions, forming a 30×4 two-dimensional input matrix. The real-time displacement of the ejector pin comes from the absolute position data of an encoder or laser sensor, in mm. The real-time velocity of the ejector pin is calculated using the first-order difference of displacement over time, with the formula v_k=(s_k-s_{k-1}) / 0.01s, in m / s. The hydraulic pressure of the ejector cylinder serves as a key parameter reflecting load changes, in MPa. The mold cavity temperature is an environmental parameter affecting the ejector pin's friction coefficient, in °C. The input features need to be standardized to eliminate dimensional differences. The standardization formula for displacement is s'=(s-s_min) / (s_max-s_min), where s_min is 0 mm and s_max is the maximum ejector pin stroke, such as 200 mm. The standardization formula for velocity is v'=v / v_max, where v_max is the maximum ejector pin velocity, such as 1.5 m / s. Similarly, pressure and temperature are normalized to the [0,1] interval according to their actual range.

[0030] Considering the long-term and short-term dependencies of the pin's motion, a Long Short-Term Memory (LSTM) network is used as the core structure to avoid the gradient vanishing problem. This network includes an input layer, two LSTM hidden layers, a fully connected layer, and an output layer. The input layer contains 4 neurons to receive a 30×4 two-dimensional input matrix; the first LSTM hidden layer contains 64 neurons, and the second LSTM hidden layer contains 32 neurons, both using the tanh activation function and returning a sequence set to True; the fully connected layer contains 16 neurons using the ReLU activation function; and the output layer contains 10 neurons to output the displacement predictions for the next 10 sampling points, corresponding to a physical time of 100ms. To prevent overfitting, a Dropout layer is added between the LSTM layer and the fully connected layer, with a dropout rate set to 0.2. A multi-step prediction mode is used to output the displacement sequence of the next 10 sampling points, rather than a single-point prediction, to improve the foresight of the path planning.

[0031] The training dataset was constructed by collecting at least 1000 complete top-out cycle data sets, covering different product models and load conditions. The dataset was divided into training, validation, and test sets in a 7:2:1 ratio. Each set of data contains a 30×4 input matrix and the actual displacement labels for the next 10 sampling points. The training process used the Adam optimizer, with an initial learning rate of 0.001. The learning rate decay strategy was to reduce it to 0.1 times if the validation set loss did not decrease after 3 rounds. The loss function used was mean squared error, calculated as follows: in For predicted displacement, yi represents the actual displacement; the batch size is set to 32; the maximum number of iterations is 500 rounds, and an early stopping method is used, stopping training when the MSE on the validation set shows no improvement for 10 consecutive rounds. After training, the prediction error on the test set must meet the following requirements: mean absolute error ≤ 0.5 mm, and prediction error within the 95% confidence interval ≤ 1 mm, to ensure prediction accuracy reaches ±1 mm.

[0032] Real-time prediction begins within 5ms after the ejector pin activation signal is triggered. When the status code switches to "Starting" (0x01), the system reads data from 30 sampling points in the current time window, covering features from the most recent 300ms, and inputs this data into the model for prediction. After the model outputs the displacement sequence for the next 100ms, the predictive path planning module generates the future trajectory of the ejector pin based on this sequence using B-spline curve fitting. It then calculates the spatial trajectory of the robot's end effector, such as the shortest path from the current position to the predicted grasping point, using time optimization as the criterion.

[0033] At the moment of each ejector pin activation, the module predicts the position sequence within the next 300ms based on current initial conditions (such as ejector cylinder pressure and mold temperature), with an accuracy of ±1mm. The robot arm reconstructs the part-picking path based on this prediction: aiming for optimal time, it uses a gradient descent algorithm to solve the B-spline space trajectory from the current position to the predicted gripping point, generating a seven-segment S-shaped velocity curve with the lowest energy consumption. For example, when the prediction indicates that the ejector pin will stop early, the system automatically shortens the robot arm's acceleration distance and increases the deceleration slope, reducing the part-picking cycle by 15%.

[0034] A multi-layered protection system is constructed using a safety redundancy module. The ejector pin position deviation threshold employs a dynamic calibration strategy: the first threshold is set at 80%-90% of the allowable deviation upper limit; when triggered, an early warning is sent to the HMI and the deviation data is recorded. The second threshold is set within the range of 110%-120%; if this limit is exceeded, the robot arm servo enable signal is immediately frozen and an audible and visual alarm is activated. The virtual safety space boundary is generated based on the mold CAD model: an STL file is imported into the robot arm control system, and a three-dimensional grid map is constructed by extending 5mm outward from the cavity surface using a voxelization algorithm. Each grid is marked as a prohibited intrusion area. The specific construction process includes: Preprocessing and format conversion of mold CAD model: Export the 3D CAD model of the mold to STL format, set the accuracy of the triangular facets to 0.1mm-0.5mm, and ensure the integrity of the key structure of the cavity; 3D raster modeling and coordinate mapping: The STL model is voxelized using a 0.5mm×0.5mm×0.5mm raster size. A calibration matrix is ​​used to achieve rigid mapping between the mold coordinate system and the robot motion coordinate system, ensuring that the raster coordinates are consistent with the actual motion coordinates. Boundary expansion and safe zone definition: The boundary contour is formed by expanding outward by 5mm ± 0.1mm along the normal direction of the inner wall of the cavity using the Offset algorithm. An additional 3mm is added to the movement trajectory of moving parts such as ejector pins and sliders as a dynamic buffer. Real-time intrusion detection logic: The control system queries the grid status corresponding to the position of the robot arm end effector at a frequency of 1kHz. When an intrusion into a "prohibited" grid is detected for three consecutive cycles, an emergency stop (STO) is immediately triggered and the intrusion coordinates are recorded.

[0035] Safety monitoring is performed by an independent hardware unit: the FPGA chip is used to process position signals in parallel, and the hardware comparator detects boundary intrusion in real time; the software layer deploys a dual watchdog mechanism, with the main watchdog monitoring the control cycle (timeout threshold 15ms) and the secondary watchdog verifying the data checksum. Any abnormality will trigger the Safety Torque Off (STO) function.

[0036] To address complex operating conditions, the system offers multiple technical solutions. In the signal acquisition stage, if the encoder fails, it can switch to a laser displacement sensor solution: using a 905nm wavelength laser source, measuring the distance to the reflective surface at the tip of the ejector pin at a 100kHz sampling frequency, and achieving an accuracy of ±0.01mm through Kalman filtering noise reduction. Alternatively, a hydraulic cylinder pressure sensing solution can be used: resonant MEMS pressure sensors are deployed at the inlet and outlet of the ejector cylinder, and position information is inferred by combining a pre-calibrated pressure-displacement transfer function (such as piecewise cubic polynomial fitting). The communication layer supports wireless redundancy: the Wi-Fi 6 module uses OFDMA technology to achieve low-latency transmission between multiple devices, and works with a 5G URLLC network to meet the 1ms air interface latency requirement; the OPCUA solution achieves cross-platform data interoperability through information model standardization (e.g., defining the [ejector pin speed] node as ns=1; i=2005).

[0037] The control algorithm layer provides an intelligent upgrade interface. The fuzzy control scheme is designed with a 49-rule base: taking position deviation E and deviation change rate EC as input, a seven-level fuzzy set {NB, NM, NS, ZO, PS, PM, PB} is defined. The output speed correction is defuzzified using the centroid method to generate precise instructions. The machine learning adaptive model uses incremental training: every 100 molds produced, the LSTM network weights are updated with the latest data, dynamically optimizing the PID gain parameters (e.g., Kp is adaptively adjusted from 0.6 to 0.8 to cope with increased friction caused by mold wear).

[0038] Example 2: A method for synchronous control of in-mold ejection in collaboration between an injection molding machine and a robotic arm, referring to... Figure 2 As shown, the in-mold ejection synchronous control system for realizing the collaboration between the injection molding machine and the robot includes: At the methodological level, the control flow begins in the bidirectional link establishment phase: the robot control system automatically identifies the EtherCAT node of the injection molding machine through the EDS file, configures the process data object (PDO) mapping, and completes the direct mapping from the encoder's raw data to the control memory. The real-time data acquisition phase employs a dual-buffering mechanism: a front-end buffer is used for reading by the control algorithm, while the back-end buffer continuously updates data via DMA to ensure no sampling omissions. In dynamic tracking control, the feedforward compensation is calculated based on the Newton-Euler equations. F ff =m·a ejector +c·v ejector In the formula, F ff This is the feedforward compensation force, i.e., the feedforward control force required by the robot arm, used to counteract system inertia and frictional hysteresis; m is the equivalent mass parameter, c is the viscous friction coefficient, and a... ejector For the real-time acceleration of the ejector pin, v ejectorTo achieve real-time speed control of the ejector pin, advance compensation for disturbance forces is implemented. During the speed matching phase, a finite state machine (FSM) is used for mode switching. The system immediately switches to the "synchronous deceleration" sub-state at the moment the ejector pin deceleration begins, and the deceleration start point of the robot arm is calculated according to the formula: In the formula, S decel v is the initial distance for deceleration. current v represents the current velocity of the thimble. end Let a be the target terminal velocity. max For the maximum permissible deceleration, t safe Allow for a safety margin of time.

[0039] Predictive planning begins trajectory pre-calculation 500ms before the thimble starts, employing Rolling Time Control (RHC) to update the path every 50ms. A safety monitoring thread scans the virtual boundary grid map at μs intervals, using a ray collision detection algorithm to determine if the robot tool's center point (TCP) has intruded into a restricted area. Upon detection, ISO13849-1PLe-level safety protection is immediately triggered.

[0040] During system implementation, depth parameter calibration is required: Under no-load conditions on the injection molding machine, the transfer function of the ejector system is obtained through step response testing; under load conditions, the influence of different product masses (50g-2000g) on ​​the ejector acceleration curve is recorded, and a mass-friction coefficient lookup table is established. During the debugging phase, a digital oscilloscope is used to synchronously capture the encoder signal and the actual position of the robot arm, verifying that the position deviation is consistently controlled within ±0.3mm, and the part-picking cycle is shortened by 23% compared to traditional trigger-based control. This collaborative system has been successfully applied to a precision injection molding production line for automotive lenses, achieving a 99.98% part-picking success rate at an ejector speed of 1.2m / s.

[0041] The embodiments described in this specific implementation are preferred embodiments of this application and are not intended to limit the scope of protection of this application. Identical components are represented by the same reference numerals. Therefore, all equivalent changes made to the structure, shape, and principle of this application should be covered within the scope of protection of this application.

Claims

1. An in-mold ejection synchronization control system of an injection molding machine and a robot, characterized by, The application relates to a control system for a needle device of an injection molding machine, comprising: a bidirectional communication module for establishing a bidirectional data link between the needle device and a manipulator control system and collecting key data of the needle device in real time and synchronously transmitting the key data to a downstream module; a dynamic position following module for receiving the key data through the bidirectional data link and calculating a real-time position deviation value, combining a feedforward compensation value and a PID adjustment value to generate a manipulator motion speed instruction and an acceleration instruction; an adaptive speed matching module configured to identify a current motion stage of the needle and switch a following control mode of the manipulator according to different motion stages; a predictive path planning module configured to store and analyze historical motion data of the needle, build a dynamic model of the needle motion, predict a future position change trend of the needle based on the model and a current state, and optimize path planning of the manipulator; a safety redundancy module configured to set a position deviation threshold value of the needle and a virtual safety space boundary of a mold cavity, and execute a warning, suspension or avoidance action according to a position deviation and space invasion.

2. The in-mold ejection synchronization control system of claim 1, wherein, The bidirectional data link adopts a real-time Ethernet protocol for data exchange, and a transmission period is less than 10 milliseconds; The key data of the needle device is collected in real time through a needle encoder, a signal output end of the needle encoder is directly connected to a signal input end of the manipulator control system, and the key data of the needle device collected in real time is transmitted, wherein the key data comprises position data, speed data and running state parameters; The position data comprises an absolute position and a relative displacement amount of the needle; The speed data comprises instantaneous speed and acceleration; The running state parameters comprise a state code representing that the needle is in a starting, running, stopping or fault state.

3. The in-mold ejection synchronization control system of claim 1, wherein, The dynamic position following module calculates the real-time position deviation value in the following manner: a relative position rule of the needle and the manipulator is preconfigured in the manipulator control system; the dynamic position following module receives the key data of the needle device collected in real time; the manipulator control system obtains a real-time actual position of the manipulator through a position feedback device of the manipulator control system; based on the collected key data and the real-time actual position of the manipulator, an actual position difference value of the two is calculated through a closed-loop control algorithm to obtain the real-time position deviation value; and a speed difference value is also calculated; the feedforward compensation value is calculated based on the received real-time speed data and real-time acceleration data of the needle; the PID adjustment value is obtained by processing the real-time position deviation value through a proportional link, an integral link and a differential link; and finally generated motion speed instructions and acceleration instructions of the manipulator are used to control a servo driver of the manipulator, so that a suction cup and a gripper of an end effector of the manipulator maintain a preset synchronous motion relationship and a constant spatial distance with a needle end.

4. The in-mold ejection synchronization control system of claim 1, wherein, In the adaptive speed matching module, the identification of the motion stage of the needle specifically comprises: analyzing the real-time speed data and a change rate thereof, and determining an acceleration stage when the change rate is positive and exceeds an acceleration determination threshold value, determining a uniform speed stage when an absolute value of the change rate is lower than a uniform speed determination threshold value, and determining a deceleration stage when the change rate is negative and lower than a deceleration determination threshold value; the switching of the following control mode of the manipulator specifically comprises: In the pin acceleration stage, the manipulator starts the trapezoidal acceleration according to the preset initial acceleration and maximum speed parameters; In the pin constant speed stage, the speed instruction value of the manipulator is set to be equal to the current pin linear speed value; In the pin deceleration stage, the manipulator calculates the deceleration point in advance and executes deceleration to the picking-up completion position according to the predicted pin stop position and the preset safety margin time.

5. The in-mold ejection synchronization control system of claim 1, wherein, In the predictive path planning module, the pin motion dynamic model is constructed by using a system identification method based on historical data, and the pin motion dynamic model at least includes a fitting function or a state space equation of the relationship between the pin displacement and time, the relationship between the speed and time, and the relationship between the acceleration and time; The predicted future position change trend of the pin refers to the extrapolation calculation of the expected position sequence of the pin in the future several control periods by using the current pin position, speed and the pin motion dynamic model; The optimization of the motion path planning of the manipulator refers to the recalculation of the spatial trajectory and the corresponding time-speed curve of the end effector of the manipulator from the current position to the target grasping position according to the expected position sequence and the time optimization or energy consumption optimization criterion.

6. The in-mold ejection synchronization control system of claim 1, wherein, The pin position deviation threshold in the safety redundancy module includes a first threshold and a second threshold, the first threshold is set to 80%-90% of the allowed position deviation range, and the second threshold is set to 110%-120% of the allowed position deviation range; The virtual safety space boundary is preset in the manipulator control system according to the three-dimensional CAD model data of the mold, and the boundary maintains a safety distance of not less than 5mm from the actual profile of the mold and the movable parts; The safety redundancy module is independent of the main control loop and is configured with an independent watchdog circuit and a software monitoring thread to ensure that safety protection actions can still be performed when the system main control is abnormal.

7. The in-mold ejection synchronization control system of claim 1, wherein, The bidirectional communication module further includes a position detection replacement unit; the position detection replacement unit is configured to: Scheme A: A high-precision laser displacement sensor is installed on a fixed base, emits a laser beam to irradiate the end of the pin or related components, and obtains real-time position data of the pin in a non-contact manner by measuring the reflected light signal; Scheme B: A pressure sensor installed on the ejection cylinder of the injection molding machine is used to monitor the hydraulic pressure in real time, and the real-time position data of the pin is indirectly calculated and derived by combining a pre-calibrated cylinder pressure-piston displacement relationship model; The position data obtained by the position detection replacement unit is transmitted to the manipulator control system through the bidirectional data link.

8. The in-mold ejection synchronization control system of claim 1, wherein, The bidirectional communication module further includes a communication mode replacement unit; the communication mode replacement unit is configured to: Scheme C: A Wi-Fi6 wireless communication module and a 5GNR industrial module conforming to the IEEE802.11ax standard are used to replace the wired Ethernet physical connection, realizing wireless data transmission between the injection molding machine and the manipulator control system; Scheme D: An OPCUA server is deployed on the injection molding machine control system side, an OPCUA client is deployed on the manipulator control system side, and the pin position, speed and state parameter data are encapsulated and exchanged according to the OPCUA information model specification, realizing standardized data interaction across different brands of equipment.

9. The in-mold ejection synchronization control system of claim 1, wherein, The dynamic position following module further comprises a control algorithm replacement unit; the control algorithm replacement unit is configured to: Scheme E, using a fuzzy control algorithm to replace the PID regulation, the fuzzy control algorithm takes the real-time position deviation value and its rate of change as input variables, infers through a pre-defined fuzzy rule base and membership function, outputs the fuzzy amount of manipulator speed adjustment, and then through defuzzification processing obtains the accurate speed command correction value; Scheme F, using a machine learning-based adaptive control model, the adaptive control model continuously collects historical position deviation data, manipulator response data and ejector pin movement data during operation, dynamically optimizes the feedforward compensation amount and / or PID control parameters or fuzzy rule weights using an online learning algorithm, to adapt to different ejector pin movement characteristics and load changes.

10. A method for synchronizing the in-mold ejection of an injection molding machine with a robot, characterized in that A method for implementing an in-mold ejection synchronization control system for a plastic injection machine and a manipulator as claimed in any one of claims 1-9, comprising the following steps: S1, constructing a bidirectional data link: establishing a bidirectional data communication connection between the ejector pin device of the plastic injection machine and the manipulator control system; S2, real-time data transmission and collection: through the bidirectional data link, the manipulator control system acquires real-time key data of the ejector pin of the plastic injection machine; S3, dynamic position following control: calculating the real-time position deviation value according to the real-time position data of the ejector pin and the preset target relative position relationship, applying a closed-loop control algorithm, combining the feedforward compensation amount and the PID control parameters to generate and output the manipulator movement command; S4, adaptive speed matching control: monitoring and identifying the ejector pin movement stage, controlling the manipulator to start acceleration according to the trapezoidal speed curve in the acceleration stage, maintaining the same speed as the ejector pin in the uniform speed stage, and synchronously decelerating to the pick-up completion position in the deceleration stage; S5, predictive path planning: storing and analyzing historical ejector pin movement data to construct a dynamic model, predicting the future movement trajectory of the ejector pin based on the current state, calculating the optimal pick-up path of the manipulator and planning in advance; S6, safety monitoring and redundancy protection: setting position deviation protection threshold and virtual safety space boundary, real-time monitoring of position deviation and space intrusion, executing warning, pause, alarm or avoidance actions.

Images