A smart monitoring and adaptive control system and method for resin flow in RTM process

Abstract

This invention relates to the field of composite material manufacturing technology, and more particularly to an intelligent monitoring and adaptive control system and method for resin flow in an RTM (Resin-to-Mortar) process. The system includes an image acquisition module, a multi-sensor array, a central processing unit, and an actuator. The image acquisition module acquires raw image sequences of resin flow through a local visualization window of the RTM mold. The multi-sensor array monitors the pressure and temperature at the injection port and vent. The central processing unit receives the raw image sequences and sensor data, identifies the resin flow front position using an image recognition algorithm, and predicts the flow state based on a resin flow process model. When the prediction deviates from a preset acceptable process window, adaptive control commands are generated according to control logic. The actuator adjusts the resin injection machine injection parameters according to the commands. This invention achieves transparent and intelligent closed-loop control of the RTM process, enabling real-time prevention and correction of flow defects, improving product quality consistency and production efficiency, and creating a traceable digital process archive.

Description

Technical Field

[0001] This invention relates to the field of composite material manufacturing technology, and in particular to an intelligent monitoring and adaptive control system and method for resin flow in RTM process. Background Technology

[0002] RTM (Resin-to-Mesh) is a closed-mold molding technology that involves injecting resin into a sealed mold cavity containing a fiber preform, allowing it to impregnate and cure to form a composite material product. The core of this process is the resin impregnation of the fibers, which occurs within the sealed mold and is invisible to the naked eye. Therefore, operators cannot directly observe the resin flow and fiber impregnation, resulting in extremely poor ability to anticipate and control defects such as dry spots and bubbles, making it difficult to guarantee product yield.

[0003] To achieve visualization of the manufacturing process, some existing technologies have been explored. For example, patent CN210821003U discloses a visual RTM molding mold, whose upper mold adopts a split design, including a frame and a glass plate, through which observation is achieved. However, this solution only provides a passive means relying on human visual observation, and cannot perform quantitative analysis of the flow state, let alone achieve automatic control. Furthermore, the glass material is fragile and carries the risk of seal failure due to a mismatch in thermal expansion coefficients with metals.

[0004] Patent CN120003071B improved the visualization mold by using PC board as the visualization material and optimizing the frame structure and sealing design, thereby increasing the mold's strength and lifespan. This patent solves the durability problem of visualization molds, but its core technology remains focused on the mold structure itself. While it provides the means to "see," it doesn't address "how to understand" or "how to automatically respond after understanding." The control of the entire dispensing process still relies on the operator's experience and manual intervention, failing to guarantee process consistency and reproducibility.

[0005] In summary, existing visualized RTM molds have failed to address the fundamental problem of how to convert visualized image information into executable control commands to achieve intelligent, adaptive closed-loop control of the process. To address this issue, this invention proposes an intelligent monitoring and adaptive control system and method for resin flow in RTM processes. This system and method transcends passive observation, automatically and proactively sensing and analyzing the resin flow state, and adaptively adjusting process parameters in real time to achieve intelligent monitoring and adaptive control of resin flow in RTM processes. Summary of the Invention

[0006] The purpose of this invention is to solve the problems existing in the prior art by proposing an intelligent monitoring and adaptive control system and method for resin flow in RTM process.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: an intelligent monitoring and adaptive control system and method for resin flow in an RTM process, wherein the system is used in conjunction with an RTM mold and a resin injection machine, and includes: An image acquisition module is configured to acquire raw images of the interior of the RTM mold cavity in real time to obtain a raw image sequence of the resin flow front. The multi-sensor array includes an inlet pressure sensor and an inlet temperature sensor disposed at the resin injection port of the RTM mold, and an outlet pressure sensor disposed at the vent port of the RTM mold. A central processing unit, which is communicatively connected to the image acquisition module and the multi-sensor array, is configured to: Based on the original image sequence acquired by the image acquisition module, the position of the resin flow front in the cavity is identified in real time through an image recognition algorithm; Based on the position of the resin flow front and the real-time data of the multi-sensor array, combined with the pre-stored resin flow process model and the preset qualified process window, it is determined whether the resin flow state deviates. When a deviation is detected, an adaptive control command is generated; An actuator, connected to the central processing unit, is configured to receive the adaptive control commands and adjust the injection parameters of the resin injection machine accordingly.

[0008] Preferably, the system is configured to be used in conjunction with an RTM mold having at least one local visualization window, the image acquisition module having an optical axis configured to be aligned with the local visualization window for observation of the cavity interior.

[0009] Preferably, the central processing unit determines whether the flow state deviates by comparing the time when the resin flow front, identified in real time, arrives at the preset observation point with the theoretical arrival time calculated by the resin flow process model based on real-time data; wherein, a time deviation exceeding a preset threshold is determined to be a deviation.

[0010] Preferably, the deviation specifically includes flow lag, flow advance, or asymmetric advancement of the resin flow front between different preset observation points; wherein: When a lag in the resin flow front is detected, the adaptive control command is to increase the injection pressure; When the resin flow front is detected to be ahead of schedule or asymmetrical, the adaptive control command is to reduce the injection pressure or injection flow rate, or to pause the injection.

[0011] Preferably, the central processing unit is further configured to: after generating the adaptive control command, verify whether the flow state deviates based on the injection parameter adjustment data fed back by the multi-sensor array, and decide whether to generate subsequent corrective adaptive control commands accordingly.

[0012] Preferably, the image acquisition module includes a high-resolution industrial camera and a high-brightness LED ring light source, the LED ring light source being arranged to surround the lens of the industrial camera.

[0013] Preferably, the central processing unit is further configured to bind the original image sequence, real-time data of the multi-sensor array, decision process data generated according to the control logic, and executed adaptive control instructions in each injection process with the corresponding product number and store them as a digital process archive for quality traceability and process optimization analysis.

[0014] Preferably, the actuator is a servo motor driver or a proportional valve controller.

[0015] Preferably, the central processing unit is further configured to: control the resin injection machine to stop injection when it is determined by the image recognition algorithm that the resin has completely filled the cavity, and when it is determined by the real-time data monitored by the outlet pressure sensor that the outlet pressure has reached and stabilized at a preset filling completion threshold.

[0016] A method for intelligent monitoring and adaptive control of resin flow in an RTM process, employing any one of the aforementioned intelligent monitoring and adaptive control systems for resin flow in an RTM process, includes the following steps: S10. System initialization: After the RTM mold is closed, the system is started and the pre-stored resin flow process model and preset qualified process window are loaded. S20. Data Synchronous Acquisition: Start the resin injection machine, synchronously trigger the image acquisition module to start continuously acquiring the original image sequence at the set acquisition frame rate, and synchronously start the multi-sensor array to record data. S30. Real-time analysis and decision-making: The central processing unit processes the raw image sequence received in real time, identifies the position of the resin flow front through the image recognition algorithm, and fuses the real-time data of the multi-sensor array. It then combines the resin flow process model to determine whether the current flow state deviates from the qualified process window. S40, Adaptive control execution: If step S30 determines that there is a deviation, the central processing unit generates a corresponding adaptive control command and sends it to the actuator, which drives the resin injection machine to adjust the injection parameters in real time. S50, Process completion judgment: Continue to continuously acquire and record the original image sequence of step S20, and repeat steps S30 to S40 until the process completion conditions are met, then control the resin injection machine to stop injection, and form and save the digital process file of this process. The process completion conditions are as follows: based on the original image sequence recognition results, it is determined that the resin has completely filled the cavity, and based on the real-time data monitored by the outlet pressure sensor, it is determined that the outlet pressure has stably reached the preset filling completion threshold.

[0017] Compared with existing technologies, the advantages of this invention are: 1. Intelligent closed-loop control has been achieved: the visualization observation has been upgraded to real-time perception and decision-making based on image recognition algorithms, realizing automatic and precise adjustment of process parameters and reducing reliance on human experience; 2. Defect prevention and quality improvement: It can predict and correct abnormal states such as flow lag and asymmetry in real time, effectively prevent the occurrence of defects such as dry spots and bubbles, and shift quality control from "post-inspection" to "in-process prevention", significantly reducing the scrap rate; 3. High process stability and consistency: The system can automatically compensate for interference caused by fluctuations in fiber preform permeability and changes in resin viscosity, ensuring the consistency of the process for different batches of products; 4. Full-process digitalization and traceability: The system records complete process data and forms digital process archives, providing complete data traceability for each product, which greatly facilitates quality analysis, problem investigation and continuous process optimization; 5. Strong system compatibility: This system can be used as an intelligent upgrade module and combined with existing high-performance visual RTM molds to achieve synergistic effects. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the overall structural framework of the present invention; Figure 2 This is a flowchart of the process steps of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Example 1, such as Figure 1 and Figure 2As shown, an intelligent monitoring and adaptive control system for resin flow in an RTM process is disclosed. The system is used in conjunction with an RTM mold and a resin injection machine, and includes: An image acquisition module is configured to acquire raw images of the interior of the RTM mold cavity in real time to obtain a raw image sequence of the resin flow front. The multi-sensor array includes an inlet pressure sensor and an inlet temperature sensor disposed at the resin injection port of the RTM mold, and an outlet pressure sensor disposed at the vent port of the RTM mold. The central processing unit is communicatively connected to the image acquisition module and the multi-sensor array. An actuator, connected to the central processing unit, is configured to receive the adaptive control commands and adjust the injection parameters of the resin injection machine accordingly.

[0021] The system is configured to be used in conjunction with an RTM mold having at least one local visualization window. The image acquisition module has an optical axis that is configured to be aligned with the local visualization window for observation of the interior of the cavity.

[0022] The image acquisition module includes a high-resolution industrial camera and a high-brightness LED ring light source, which is arranged to surround the lens of the industrial camera.

[0023] The actuator is a servo motor driver or a proportional valve controller.

[0024] In this embodiment, the system hardware mainly includes the following components: RTM molds: RTM molds with one or more localized visualization windows can be used. These visualization windows are preferably made of high-strength, heat-resistant transparent materials (such as polycarbonate (PC) sheets or tempered glass), and their design is optimized based on existing technologies to ensure their stability and sealing under high temperature and pressure, such as ensuring their sealing and durability under RTM process pressures (e.g., 0.1-1 MPa) and temperatures (e.g., room temperature - 180°C). The visualization windows should be located in critical resin flow paths, such as at the end of flow channels, in areas of abrupt thickness changes, and in corners prone to air stagnation.

[0025] Image Acquisition Module: This module serves as the system's vision sensor, including a high-resolution industrial camera (such as a CCD or CMOS camera with a resolution ≥ 2 megapixels and a frame rate ≥ 30fps) to ensure real-time capture of the dynamic details of resin flow. Simultaneously, it is equipped with a high-brightness white LED ring light source. The inner diameter of the LED ring light source is adapted to the industrial camera, and the LED ring light source is mounted at the front of the industrial camera lens. It has a color temperature of 6000K and continuously adjustable brightness, providing uniform, shadowless illumination to overcome the dim environment inside the RTM mold and reduce the interference of resin surface reflections on the original image quality. The industrial camera is securely mounted on the outside of the RTM mold, with its lens precisely aligned with the local visualization window to ensure clear imaging of the entire observation area. The industrial camera is connected to the central processing unit via a gigabit Ethernet cable, ensuring high-speed and stable transmission of the original image sequence.

[0026] Multi-sensor array: Inlet pressure sensor: Uses piezoelectric or strain gauge sensor, high-precision model, such as ±0.5%FS accuracy, directly installed on the pipeline connecting the resin injection machine outlet and the RTM mold resin injection port to monitor the resin pressure entering the RTM mold in real time; Inlet temperature sensor: A PT100 platinum resistance thermometer is used, also a high-precision model, such as an accuracy of ±0.1℃. It is installed in parallel with the inlet pressure sensor to ensure good thermal contact with the resin and to monitor the resin temperature entering the RTM mold in real time. Outlet pressure sensor: Same specifications as inlet pressure sensor, installed at the vent of RTM mold, used to monitor pressure changes at the end of the cavity; All sensor signals are connected to the analog input module of the central processing unit via shielded cables.

[0027] Central Processing Unit (CPU): This is the core control hub, employing an architecture combining an Industrial PC (IPC) and a Programmable Logic Controller (PLC). Its hardware includes a high-speed image acquisition card, analog input modules, analog output modules, and digital output modules.

[0028] Actuator: Depending on the type of resin injection molding machine connected, a servo motor driver (for injection molding machines with motor-driven metering pumps) or a proportional valve controller (for pneumatic or hydraulically driven injection molding machines) can be selected. The actuator is connected to the resin injection molding machine's own control system (such as a pressure / flow control unit), receives analog voltage or digital signals from the central processing unit, and converts them into precise adaptive control commands for the resin injection molding machine's pressure or flow regulation unit.

[0029] Resin injection machine: This is a conventional RTM injection device that needs to have an interface that can receive external control signals and adjust the injection pressure or flow rate in real time.

[0030] All components are connected via cables / fiber optic cables and industrial communication networks (such as Ethernet and EtherCAT) to form a unified intelligent control system.

[0031] Example 2: Based on Example 1, the central processing unit is configured as follows: Based on the original image sequence acquired by the image acquisition module, the position of the resin flow front in the cavity is identified in real time through an image recognition algorithm; Based on the position of the resin flow front and the real-time data of the multi-sensor array, combined with the pre-stored resin flow process model and the preset qualified process window, it is determined whether the resin flow state deviates. When a deviation is detected, an adaptive control command is generated. The central processing unit determines whether the flow state has deviated by comparing the time when the resin flow front reaches the preset observation point in real time with the theoretical arrival time calculated by the resin flow process model based on real-time data. A deviation is determined when the time deviation exceeds a preset threshold.

[0032] The deviation specifically includes flow lag, flow advance, or asymmetric advancement of the resin flow front between different preset observation points; wherein: When a lag in the resin flow front is detected, the adaptive control command is to increase the injection pressure; When the resin flow front is detected to be flowing ahead of schedule or asymmetrically advancing, the adaptive control command is to reduce the injection pressure or injection flow rate, or to pause the injection for a preset period of time, allowing the resin flow front to equalize itself.

[0033] The central processing unit is further configured to: after generating the adaptive control command, verify whether the flow state deviates based on the injection parameter adjustment data fed back by the multi-sensor array, and decide whether to generate subsequent corrective adaptive control commands accordingly.

[0034] The central processing unit is also configured to: when the image recognition algorithm determines that the resin has completely filled the cavity, and the real-time data monitored by the outlet pressure sensor determines that the outlet pressure has reached and stabilized at a preset filling completion threshold, control the resin injection machine to stop injection.

[0035] The central processing unit is also configured to bind the original image sequence, real-time data from the multi-sensor array, decision process data generated based on control logic, and executed adaptive control instructions from each injection process with the corresponding product number and store them as a digital process archive for quality traceability and process optimization analysis.

[0036] In this embodiment, the system software configuration mainly includes the following characteristics: Central Processing Unit: The IPC (In-Process Control) component is responsible for processing raw image sequences and performing complex algorithm calculations. It is configured with a high-performance multi-core processor, running a general-purpose operating system and host computer software. This software is developed using C++ and open-source computer vision libraries to implement the aforementioned image recognition algorithms. PLC Section: An industrial-grade programmable logic controller (PLC) is selected, responsible for high-speed, reliable real-time data acquisition, adaptive control command output, and hard-wired interlocking control with the resin injection molding machine. Real-time data exchange between the IPC and PLC is achieved via PROFINET industrial Ethernet, with a cycle time ≤10ms.

[0037] Pre-installed on the IPC of the central processing unit: Image recognition algorithm: A program developed based on an open-source computer vision library to implement edge detection and threshold segmentation functions.

[0038] Resin Flow Process Model: This model is integrated into the host computer software as a plug-in. Essentially, it's a numerical simulation kernel based on the finite element / finite volume method to solve Darcy's law. Before use, it's necessary to obtain the permeability data of the current batch of fiber preforms in the three main directions, as well as the resin viscosity-temperature curve, through offline experiments (or supplier data), and input them into the software. During the injection process, the resin flow process model performs rapid online simulation based on real-time inlet pressure and temperature (used for looking up and updating resin viscosity), predicting the advancement process of the resin flow front.

[0039] Control logic program: Includes qualified process window and control logic. The qualified process window is set in the host computer software and mainly includes: ① the theoretical time range for the resin flow front to reach each preset observation point (corresponding to specific pixel coordinates within the window); ② the allowable inlet pressure fluctuation range. The control logic is implemented in the form of rules, for example: IF actual time > theoretical time limit THEN triggers the "flow lag" processing flow.

[0040] Example 3: A method for intelligent monitoring and adaptive control of resin flow in an RTM process, based on any of the above examples, includes the following steps: S10. System initialization: After the RTM mold is closed, the system is started and the pre-stored resin flow process model and preset qualified process window are loaded. S20. Data Synchronous Acquisition: Start the resin injection machine, synchronously trigger the image acquisition module to start continuously acquiring the original image sequence at the set acquisition frame rate, and synchronously start the multi-sensor array to record data. S30. Real-time analysis and decision-making: The central processing unit processes the raw image sequence received in real time, identifies the position of the resin flow front through the image recognition algorithm, and integrates the real-time data of the multi-sensor array (including the real-time data of the inlet pressure sensor, inlet temperature sensor and outlet pressure sensor), and combines the resin flow process model to determine whether the current flow state deviates from the qualified process window. S40, Adaptive control execution: If step S30 determines that there is a deviation, the central processing unit generates a corresponding adaptive control command and sends it to the actuator, which drives the resin injection machine to adjust the injection parameters in real time. S50, Process completion judgment: Continue to continuously acquire and record the original image sequence of step S20, and repeat steps S30 to S40 until the process completion conditions are met, then control the resin injection machine to stop injection, and form and save the digital process file of this process. The process completion conditions are as follows: based on the original image sequence recognition results, it is determined that the resin has completely filled the cavity, and based on the real-time data monitored by the outlet pressure sensor, it is determined that the outlet pressure has stably reached the preset filling completion threshold.

[0041] This embodiment combines Figure 1-2 Taking the production of "Product A" as an example, the detailed control methods are as follows: Step S10, System Initialization: The fiber preform, after being laid up and shaped according to regulations, is placed into the RTM mold cavity. The operator selects the process formula for "Product A" on the host computer software interface. This formula file automatically loads the corresponding fiber preform permeability data and resin viscosity-temperature curve. The software interface displays a two-dimensional schematic diagram of the RTM mold cavity. The operator uses the mouse to click on the area corresponding to the actual visualization window on the two-dimensional schematic diagram to set multiple (e.g., 3) preset observation points (P1, P2, P3). Based on the loaded resin flow process model and the set standard injection pressure (e.g., 0.3 MPa), the host computer software automatically simulates and calculates the theoretical times for the resin flow front to reach P1, P2, and P3, respectively: t1=85s, t2=120s, and t3=165s. After operator confirmation, the system initialization is complete.

[0042] This step transforms specific product and material characteristics into quantifiable monitoring targets (theoretical timelines) for the system, providing a benchmark for subsequent intelligent judgment.

[0043] The qualified process window does not refer to the physical observation window on the RTM mold, but rather to a series of quantitative parameter ranges preset in the central processing unit (CPU) to determine whether the process status is normal. In this embodiment, the window is mainly defined with preset observation points (P1, P2, P3) as spatial references and time as the key criterion. For example, a core rule of the qualified process window can be stated as: "The actual time for the resin flow front to reach the preset observation point P1 should be within ±10% of the theoretical time t1 at that point." Furthermore, the qualified process window may also include limitations on the fluctuation range of key parameters such as injection pressure and injection flow rate. The CPU compares the real-time sensing data with these preset quantitative parameter ranges to make a decision on whether to intervene in control.

[0044] Step S20, Data Synchronization Acquisition: a. Parameter preset: Based on the maximum expected flow rate of the resin and the system control cycle (e.g., 100ms), the acquisition frame rate of the industrial camera is set to 20 fps in the host computer software, and the industrial camera is configured to external trigger mode. b. Synchronous Trigger Start: After the operator starts the resin injection machine to begin injecting resin, the programmable logic controller (PLC) synchronously performs two operations: first, it sends a start command to the actuator; second, it generates a 24V, 50μs wide TTL pulse from one of its digital output ports as a global synchronization signal. c. Raw Image Sequence Acquisition: The TTL pulse is simultaneously sent to the external trigger port of the industrial camera. The industrial camera immediately exposes and generates a raw image frame on the rising edge of the pulse, transmitting it to the IPC's image buffer via a Gigabit Ethernet interface. The PLC cyclically sends this trigger pulse at 100ms intervals, thereby driving the industrial camera to acquire and transmit a continuous raw image sequence in perfect synchronization with the system control rhythm. The host computer software maintains a First-In-First-Out (FIFO) image queue to manage this raw image sequence; d. Synchronous Sensor Data Acquisition: The same TTL pulse signal is also used to activate the PLC's high-speed analog input module. This module, at a sampling rate of 1 kHz, begins continuously recording current signals from the inlet pressure sensor, inlet temperature sensor, and outlet pressure sensor, and converts them into engineering values. Both the original image sequence and the data from the multi-sensor array are timestamped based on the time of this synchronous TTL pulse signal, ensuring strict time-domain alignment of the multi-source data.

[0045] Hardware-synchronized startup ensures microsecond-level alignment accuracy between the original image sequence and the data from the multi-sensor array on the time axis. This is the foundation for subsequent precise data fusion and state analysis, avoiding errors caused by software delays.

[0046] Step S30, Real-time Analysis and Decision Making (Cycle: 100ms): This step is executed as a 100ms timed loop task in the host computer software.

[0047] S31, Image Processing and Forward Recognition: a. Acquisition and Preprocessing: The host computer software retrieves the latest frame of the original image (resolution 2048×1536) from the cache. First, it is converted to grayscale, and then a 5×5 pixel Gaussian filter is applied for smoothing and noise reduction. b. Edge detection: The Canny operator is used on the denoised grayscale image, with the low threshold set to 50 and the high threshold set to 150, to extract all obvious contours, including resin fronts, fiber textures, etc. c. Threshold Segmentation and Front Localization: The host computer software analyzes the grayscale histograms of the lower (filled) and upper (unfilled) parts of the latest frame of the original image, and automatically calculates an optimal segmentation threshold T (e.g., T=120) using Otsu's Method. Regions with pixel values ​​less than T are marked as resin (black), and regions with values ​​greater than T are marked as air (white), resulting in a binary image. Finally, by scanning each column of the binary image, the first pixel to transition from black to white is found, and these points are connected to form the resin flow front curve for the current frame. The host computer software calculates whether the pixel coordinates corresponding to preset observation points P1, P2, and P3 on this curve are covered by black pixels. S32. Data Fusion and Model Prediction: Within the same cycle, the system reads data from the multi-sensor array: inlet pressure P in =0.302MPa, inlet temperature T in =25.1℃, outlet pressure P out =0.101MPa (normal pressure). The host computer software, based on T... in Consult the resin viscosity-temperature curve to obtain the current resin viscosity μ. Then, using the current P... in Using μ as input, a single-step rapid simulation of the resin flow process model is performed to predict the theoretical position of the resin flow front under the current conditions. Simultaneously, based on the current injection time t... now The system records the actual occurrence time of key events, such as t, at preset observation points already covered by the resin flow front. P1-a (The actual time when the resin flow front reaches point P1). S33. Deviation Detection: The central processing unit calls the pre-stored "Qualified Process Window" parameters and compares the real-time data with the pre-stored "Qualified Process Window" parameters. The specific determination logic is as follows: a. Flow lag / early determination: tP1-a The theoretical arrival time t of the resin flow front at point P1 in the "qualified process window" P1-t and the allowable time deviation threshold δ t (e.g. t) P1-t Compare with ±10% of: If t P1-a >(t) P1-t +δ t If the condition is met, then "money lagging" is determined to have occurred. If t P1-a <(t) P1-t -δ t If the above conditions are met, then it is determined that "early movement" has occurred; b. Flow asymmetry determination: Compare the preset observation points at symmetrical positions, such as point P on the left. left and point P on the right right Actual arrival time t left-a With t right-a Calculate the time deviation Δt = |t left-a -t right-a |; If Δt>Δt max (Δt) max If the maximum allowable asymmetric time difference is pre-stored (e.g., 15 seconds), then "flow asymmetry" is determined to have occurred. c. Pressure anomaly detection: Inlet pressure P in The target pressure range set in the "Qualified Process Window" [P] min ,P max (where P) min P is the preset minimum allowable pressure. max Compare with the preset maximum allowable pressure. If P in Persistently higher than P max or below P min If so, it is determined to be "pressure control abnormality", which may be caused by actuator failure or external interference; S34. Cross-validation: This sub-step is activated when the host computer software detects that the actual resin flow front has covered all preset observation points. Assuming the system identifies that all preset observation points have been covered by the resin flow front (determined as "image full"), the system monitors the outlet pressure P. out The changes occur over the next 2-3 seconds. Under normal resin filling conditions, the vent is sealed by the resin, P out The pressure should rise rapidly from normal atmospheric pressure. If the image is full, P... outIf no significant increase (e.g., not exceeding 0.05 MPa) is observed within a preset time (e.g., 3 seconds), the system comprehensively determines that there is a "suspected local dry spot or outlet pressure sensor abnormality" and triggers an alarm. This verification improves the system's ability to detect hidden defects and hardware failures. Simultaneously, the event is recorded in detail in the digital process archive. S35. Decision Output: Based on the deviation judgment and cross-validation results of S33 and S34, the system generates clear decision labels (such as "normal status", "flow lags behind P1", "left and right asymmetry", "verification alarm", etc.) and prepares to generate specific adaptive control instructions in the next step (S40) according to the preset control logic.

[0048] Step S40: Adaptive control execution: The goal of this step is to convert the deviation judgment and decision labels generated in step S30 into actual control actions on the resin injection machine without delay and with high accuracy, forming a closed-loop feedback. The specific execution process is as follows: 1. Command Generation and Mapping: The central processing unit (CPU) queries its internally preset control strategy mapping table based on the decision label output by S35. This table is a defined set of "if-then" rules that maps specific types of deviations to specific adaptive control command parameters. For example: IF Decision Label = "Flow Lags Behind P1" THEN Adaptive Control Command = {Type: "Pressure Adjustment", Target Value: "Current Pressure + 0.03 MPa", Mode: "Stepped"}; IF Decision Label = "Flow Asymmetry (Left Faster, Right Slower)" THEN Adaptive Control Instruction = {Type: "Flow Control", Action: "Pause Injection", Duration: "3.0 seconds"}; IF Decision Label = "Pressure Control Abnormal (Too High)" THEN Adaptive Control Instruction = {Type: "Pressure Adjustment", Target Value: "P";} max -0.02MPa", mode: "immediately"}; 2. Adaptive control command transmission: The generated adaptive control command is sent from the IPC to the programmable logic controller (PLC) in real time via Industrial Ethernet (PROFINET). The PLC parses this adaptive control command into low-level control signals. 3. Low-level control signal output and equipment action: The PLC drives the corresponding actuator according to the adaptive control instruction type. For pressure / flow adjustment commands: The PLC's analog output module converts the target pressure value (e.g., 0.33 MPa) into a corresponding 4-20mA analog current signal, which is then output to the proportional valve controller or servo driver. This proportional valve controller or servo driver linearly adjusts the valve opening or motor speed within 100-500ms, ensuring a smooth transition of the injection pressure to the new set value. For the pause injection command: The PLC's digital output module first outputs a "close injection valve" signal. Simultaneously, the PLC's internal high-precision timer starts. After the preset pause duration (e.g., 3.0 seconds) is reached, the PLC automatically outputs a "open injection valve" signal and resumes injection at a pressure slightly lower than before the pause (e.g., a reduction of 0.05 MPa) to avoid secondary impact upon resumption. For alarm commands: The PLC triggers the audible and visual alarm and displays specific alarm information (such as "suspected local dry spots or outlet pressure sensor abnormality") through the human-machine interface (HMI), but usually does not automatically stop the process unless a safety interlock is involved (such as overpressure). 4. Action Confirmation and Closed Loop: After the actuator moves, the inlet pressure sensor provides real-time feedback of the new pressure value. The central processing unit reads this adjusted feedback data in step S30 / S32 of the next control cycle (e.g., after 100ms) to verify whether the adaptive control command has been executed correctly, thus completing one closed-loop control cycle.

[0049] Step S50, Process Completion and Archiving: The goal of this step is to accurately and reliably determine the completion of cavity filling in the RTM mold, and to systematically create and preserve traceable digital process archives.

[0050] The specific execution process is as follows: 1. Dual judgment of process completion conditions: The system runs two independent judgment logics in parallel, using "AND" logic to confirm completion, ensuring reliability. Judgment A (Visual Main Criterion): The central processing unit continuously monitors the recognition results of the original image sequence. When it is recognized that the resin flow front has completely covered the last preset observation point (such as P3) and this condition is continuously met for several consecutive control cycles (such as 3) (300ms), and the position of the resin flow front is stable and does not advance, it is judged as "image full"; Judgment B (Pressure Verification Criterion): Simultaneously with the "image full" condition being triggered, the system begins monitoring real-time data from the outlet pressure sensor. Under normal full filling conditions, the exhaust port is sealed, and the pressure should rise rapidly. If P... out If the pressure continuously rises from atmospheric pressure (0.101 MPa) and steadily exceeds the preset filling completion threshold (e.g., 0.4 MPa) within 3 seconds, it is determined as "pressure confirmed full". Final determination: The system will only make a comprehensive determination that "the resin has completely filled the cavity" if and only if both judgment A and judgment B are satisfied. If only A is satisfied and B is not satisfied, the cross-validation alarm in S34 will be activated, and the operator will need to intervene. 2. Termination of Injection and Post-processing: Once the "resin has completely filled the cavity" determination is met, the PLC immediately executes: Send a "stop injection" command to the actuator to shut down the injection valve and feed pump of the resin injection machine; As needed, a command can be sent to initiate the heating and curing process of the RTM mold; 3. Generation and Storage of Digital Process Archives: After injection stops, the central processing unit automatically initiates the archiving process, organizing the data scattered in memory and cache into a structured, tamper-proof digital process archive. The digital process archive includes the following: a. Metadata file (metadata.json): Records product number, RTM mold number, resin batch, fiber preform batch, operator, start / end time, etc.; b. Core process data: Image Log / : This folder stores the original keyframe images (e.g., JPEGs with one frame every 2 seconds) named in chronological order and their corresponding binary images; Sensor Data.csv: A tabular file containing all sensors (P... in T in P out Complete time-series data, sampled at 1 kHz, with timestamps aligned to the original image frames; Event Decision Log.csv: A tabular file that records all system events and decisions by timestamp, such as: "100.2s: Flow lag behind P2 detected, time deviation 16.7%", "101.5s: Execution pressure increased to 0.332MPa", "165.8s: Image filled", "168.8s: Outlet pressure reached 0.45MPa, confirming that the resin has completely filled the cavity", etc. Process Summary.pdf: A report file that automatically generates a summary of key process parameters, such as total filling time, average injection pressure, whether interventions occurred and the number of interventions, and final quality judgment (OK / warning), etc. c. Archiving and Association: All the above files are packaged into a compressed archive named "Product Number Timestamp.zip" and automatically uploaded to a specified directory in the factory's Manufacturing Execution System (MES) or local server. The path of this archive is bidirectionally associated with the product number in the MES or database, enabling one-click traceability.

[0051] This step not only completes the process loop, but more importantly, it generates a complete digital process archive, providing an immutable data foundation for quality traceability, process auditing, and big data analysis.

[0052] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. An RTM process resin flow intelligent monitoring and adaptive control system, characterized in that, The system is used in conjunction with RTM molds and resin injection machines, and includes: An image acquisition module is configured to acquire raw images of the interior of the RTM mold cavity in real time to obtain a raw image sequence of the resin flow front. The multi-sensor array includes an inlet pressure sensor and an inlet temperature sensor disposed at the resin injection port of the RTM mold, and an outlet pressure sensor disposed at the vent port of the RTM mold. A central processing unit, which is communicatively connected to the image acquisition module and the multi-sensor array, is configured to: Based on the original image sequence acquired by the image acquisition module, the position of the resin flow front in the cavity is identified in real time through an image recognition algorithm; Based on the position of the resin flow front and the real-time data of the multi-sensor array, combined with the pre-stored resin flow process model and the preset qualified process window, it is determined whether the resin flow state deviates. The central processing unit determines whether the flow state deviates by comparing the time when the resin flow front, identified in real time, reaches the preset observation point with the theoretical arrival time calculated by the resin flow process model based on real-time data; wherein, a time deviation exceeding a preset threshold is determined to be a deviation. The deviation specifically includes the flow lag of the resin flow front, the flow advance, or the asymmetry of the advance between different preset observation points. When a deviation is detected, an adaptive control command is generated; After generating the adaptive control command, the flow state is verified to see if it deviates based on the injection parameter adjustment data fed back by the multi-sensor array, and a decision is made on whether to generate subsequent corrective adaptive control commands accordingly. An actuator, connected to the central processing unit, is configured to receive the adaptive control commands and adjust the injection parameters of the resin injection machine accordingly.

2. The intelligent monitoring and adaptive control system for resin flow in RTM process according to claim 1, characterized in that, The system is configured to be used in conjunction with an RTM mold having at least one local visualization window. The image acquisition module has an optical axis that is configured to be aligned with the local visualization window for observation of the interior of the cavity.

3. The intelligent monitoring and adaptive control system for resin flow in RTM process according to claim 1, characterized in that, When a lag in the resin flow front is detected, the adaptive control command is to increase the injection pressure; When the resin flow front is detected to be ahead of schedule or asymmetrical, the adaptive control command is to reduce the injection pressure or injection flow rate, or to pause the injection.

4. The intelligent monitoring and adaptive control system for resin flow in RTM process according to claim 1, characterized in that, The image acquisition module includes a high-resolution industrial camera and a high-brightness LED ring light source, which is arranged to surround the lens of the industrial camera.

5. The intelligent monitoring and adaptive control system for resin flow in an RTM process according to claim 1, characterized in that, The central processing unit is also configured to bind the original image sequence, real-time data from the multi-sensor array, decision process data generated based on control logic, and executed adaptive control instructions from each injection process with the corresponding product number and store them as a digital process archive for quality traceability and process optimization analysis.

6. The intelligent monitoring and adaptive control system for resin flow in an RTM process according to claim 1, characterized in that, The actuator is a servo motor driver or a proportional valve controller.

7. The intelligent monitoring and adaptive control system for resin flow in an RTM process according to claim 1, characterized in that, The central processing unit is further configured to: when the image recognition algorithm determines that the resin has completely filled the cavity, and the real-time data monitored by the outlet pressure sensor determines that the outlet pressure has reached and stabilized at a preset filling completion threshold, control the resin injection machine to stop injection.

8. A method for intelligent monitoring and adaptive control of resin flow in an RTM process, employing the intelligent monitoring and adaptive control system for resin flow in an RTM process as described in any one of claims 1-7, characterized in that, Includes the following steps: S10. System initialization: After the RTM mold is closed, the system is started and the pre-stored resin flow process model and preset qualified process window are loaded. S20. Data Synchronous Acquisition: Start the resin injection machine, synchronously trigger the image acquisition module to start continuously acquiring the original image sequence at the set acquisition frame rate, and synchronously start the multi-sensor array to record data. S30. Real-time analysis and decision-making: The central processing unit processes the raw image sequence received in real time, identifies the position of the resin flow front through the image recognition algorithm, and fuses the real-time data of the multi-sensor array. It then combines the resin flow process model to determine whether the current flow state deviates from the qualified process window. S40, Adaptive control execution: If step S30 determines that there is a deviation, the central processing unit generates a corresponding adaptive control command and sends it to the actuator, which drives the resin injection machine to adjust the injection parameters in real time. S50, Process completion judgment: Continue to continuously acquire and record the original image sequence of step S20, and repeat steps S30 to S40 until the process completion conditions are met, then control the resin injection machine to stop injection, and form and save the digital process file of this process. The process completion conditions are as follows: based on the original image sequence recognition results, it is determined that the resin has completely filled the cavity, and based on the real-time data monitored by the outlet pressure sensor, it is determined that the outlet pressure has stably reached the preset filling completion threshold.

Images