IMT (in-mold transfer) quality monitoring control system suitable for automotive trim

Abstract

The invention belongs to the technical field of automobile manufacturing and intelligent detection, and particularly relates to an IMT (in-mold transfer) quality monitoring control system suitable for automobile interiors, which comprises a mold body integrated with a distributed sensing array, a multi-mode online sensing module, an edge intelligent processing unit and a central coordination controller, through three-source fusion of high-speed optical observation, embedded physical sensing and thin film impedance monitoring and in combination with space-time alignment and causal modeling, real-time sensing and self-adaptive regulation and control of key state variables are realized. Through closed-loop linkage of state feedback and pre-charging calibration, the power device is prevented from bearing overshoot current in a cold start stage; a distributed timestamp alignment mechanism is utilized to maintain the multi-node event sequence consistency under the condition of no global clock source; full-link isolation processing of sensitive data in a trusted execution environment is realized, and plaintext is prevented from being leaked to a universal operating system; and it is ensured that the fault diagnosis module can still complete anomaly recognition based on the local historical mode during communication interruption.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of automobile manufacturing and intelligent detection, and in particular relates to an IMT (In-Mold Transfer) quality monitoring control system for automobile interiors. BACKGROUND

[0002] With the evolution of the automobile industry towards intelligence, individualization and high quality, consumers have higher requirements for the visual texture, tactile experience and functional integration of automobile interiors. The IMT (In-Mold Transfer) process has been widely used in the manufacturing of interior components such as instrument panels and door panel trim strips of medium and high-end vehicles, because it can realize high-precision patterns, multi-layer composite structures and three-dimensional surface integration. The process places a functional film with a pre-printed decorative pattern in an injection mold, and completes pattern transfer and substrate forming at the same time, thereby eliminating the secondary processing procedure, improving production efficiency, and enhancing product consistency and environmental protection.

[0003] The existing IMT in-mold transfer system usually includes a film supply mechanism, a positioning and clamping device, an injection unit and a basic vision detection module, and relies on preset process parameters (injection temperature, pressure, film tension, etc.) for open-loop or semi-closed-loop control, supplemented by post-inspection or fixed-point photography to judge the integrity of the transferred pattern and positional deviation. This solution can maintain a basic yield in early low-decorative scenarios, solving the problems of high cost, high pollution and poor adhesion of traditional post-decoration processes, and has historical rationality and stage value.

[0004] The existing technology has deep-seated structural contradictions: the structural disconnection between real-time perception and full-dimensional quality characterization; with the design complexity of automobile interior parts (metallic luster gradient, micro-texture touch, etc. multi-functional integration), the IMT process has strict requirements for the dynamic stability and microscopic consistency of the transfer; the traditional discrete, single-point triggered image acquisition mechanism cannot capture key process variables such as transient deformation and interlayer slip of the film under the impact of high-temperature and high-pressure melt; the addition of detection equipment lacks spatiotemporal alignment capability with injection parameters, leading to data fragmentation and difficulty in building a quality evolution model; the post-verification + experience backtracking mode cannot form an effective control loop, making it difficult to predict micron-level defects. The root cause lies in the fact that the system breaks the mapping relationship between the process and the result, and cannot match the multi-physical field coupling evolution characteristics of the transfer quality.

[0005] Therefore, the present application provides an IMT in-mold transfer quality monitoring control system for automobile interiors. SUMMARY

[0006] In order to make up for the deficiencies of the prior art and solve at least one of the technical problems proposed in the background art.

[0007] The technical scheme adopted by the present application to solve its technical problems is: the IMT in-mold transfer quality monitoring control system suitable for automotive interiors, comprising a mold body, a film supply mechanism, an injection execution unit, a multi-modal online sensing module, an edge intelligent processing unit, a process parameter adjustment actuator, and a central coordination controller; Among them, the mold body is internally integrated with a distributed fiber bragg grating strain sensing network, a micro thermocouple array and a piezoelectric interface shear force sensor, which is used to synchronously collect the local strain field, temperature gradient distribution and interlayer shear stress time series data of the film and melt contact area during the injection filling stage; the film supply mechanism is equipped with a high-precision servo tension control system, and the tension feedback signal is uploaded in real time to the central coordination controller via the RS485 bus; the injection execution unit includes a screw position encoder, an injection pressure sensor, a pressure maintaining stage flowmeter and a melt front end temperature probe, all process parameters are packaged with IEC61850 extension specification and transmitted to the edge intelligent processing unit through industrial Ethernet.

[0008] Preferably, the multi-modal online sensing module is composed of three parts: The first part is a high-speed line scan camera array arranged inside the mold parting surface, the imaging window of which is sealed with high-temperature-resistant quartz glass, and is phase-locked with the injection screw displacement signal through a synchronous trigger signal, to ensure continuous capture of the transient topography change of the film surface with microsecond-level time resolution during the melt front advance process; The second part is an infrared thermal imaging unit arranged beside the mold cooling water channel, which is used to non-contact acquisition of the complete thermal distribution map of the film back surface during the injection pressure maintaining stage; The third part is a conductive micro-grid structure embedded in the bottom layer of the functional film, which is used as a capacitive sensing layer after the transfer is completed, but is used as an impedance measurement electrode during the injection process, which detects the dielectric constant change of the film under high temperature and high pressure through high-frequency alternating excitation signal, thereby indirectly reflecting the microstructure integrity.

[0009] Preferably, the edge intelligent processing unit is configured with a dual-core heterogeneous processor architecture, wherein the main core runs a real-time operating system to process hard real-time data streams from the mold internal sensors, and the cooperative core runs a lightweight neural network inference engine to perform image feature extraction and defect pattern recognition; The intermediate calculation results are exchanged between the two cores through a double-port RAM shared cache area, and the DMA controller triggers data transfer operation according to the preset priority strategy; The edge intelligent processing unit also has a time stamp alignment engine built-in, which marks all input data streams with nanosecond-level time stamps according to the injection cycle start pulse signal, and compensates and corrects the sampling time of different sensor channels based on a physical propagation delay model, thereby generating a strictly synchronized multi-dimensional state vector sequence.

[0010] Preferably, the central coordination controller adopts a hierarchical control architecture, with an upper layer being a knowledge graph-based quality evolution reasoning machine and a lower layer being a model predictive control-based process parameter regulator; the quality evolution reasoning machine is pre-loaded with a multi-physical field coupling knowledge base of the IMT transfer process, which is obtained by training historical good product data and contains causal correlation rules between thin film material properties, melt rheological parameters, mold geometric constraints and typical defect modes; During each injection molding cycle, the reasoning machine receives a sequence of state vectors from the edge intelligent processing unit, combines the material batch identification and environmental temperature and humidity information of the current batch, dynamically deduces the possible quality risk level in the next cycle, and generates a feedforward adjustment instruction; The process parameter regulator then outputs the optimal screw speed curve, holding pressure set value and film supply tension adjustment amount according to the feedforward instruction and real-time feedback deviation signal by solving a constrained quadratic programming problem, and sends it to the servo drivers of the injection molding execution unit and the film supply mechanism through the CANopen protocol.

[0011] Preferably, the system starts the pre-charging calibration process when the mold is closed: During the mold clamping and holding stage before the start of injection molding, the central coordination controller sends a zero-point calibration instruction to the distributed fiber grating sensing network, and activates the reference impedance measurement loop of the conductive micro-grid to eliminate system offsets introduced by mold thermal expansion or mechanical assembly errors; The calibration data is stored in the non-volatile memory of the edge intelligent processing unit and serves as a reference for the calculation of state variables in subsequent injection molding cycles; After each mold opening and ejection, the system automatically executes the lens contamination detection program of the high-speed line scan camera during the idle window period before the next mold closing. This program analyzes the pixel mean square deviation of the background dark field image to judge the cleanliness of the optical window. If it exceeds the preset safety threshold, the air curtain cleaning device is triggered for local blowing.

[0012] Preferably, a bidirectional verification feedback loop is established between the multi-modal online sensing module and the injection molding execution unit: When the edge intelligent processing unit detects image features of suspected micro-cracks or overlay offset on the film surface, it immediately queries the interface shear force sensor readings and the conductive micro-grid impedance change rate at the same time. If both show abnormal mutations, it is determined to be a real defect and triggers a stop alarm; On the contrary, if only the image features are abnormal and the physical sensing data are normal, it is considered as optical interference or reflection artifact, and the system automatically ignores the false alarm and records it to the noise sample library for subsequent model optimization. This verification mechanism effectively avoids the vulnerability of single modal sensing and improves the robustness of defect discrimination.

[0013] Preferably, the cavity surface of the mold body is provided with micron-level positioning grooves, which form mechanical interlocking with the alignment marks of the functional film edge, ensuring the initial positioning accuracy of the film under the impact of injection molding; at the same time, micro Hall sensors are embedded at the bottom of the grooves for detecting whether the film is completely in the predetermined position; the position signal is used as a necessary enable condition for injection start, and only when all the positioning points are confirmed to be in place, the central coordination controller allows the injection execution unit to enter the injection stage; This design eliminates systematic overprint errors caused by initial offset of the film from the source.

[0014] Preferably, the edge intelligent processing unit is configured with a local fault diagnosis state machine, which is preconfigured with multiple levels of abnormal response strategies: First-level abnormality corresponds to single sensor failure, the system automatically switches to redundant channel and marks the fault node; Second-level abnormality corresponds to multi-modal data logic conflict, the system suspends the current period and requests manual reset; Third-level abnormality corresponds to the same defect mode appearing in continuous multiple periods, the system automatically calls the root cause analysis module in the knowledge graph, generates a process parameter drift report and pushes it to the manufacturing execution system; All state transitions need to meet the pre-set timing verification conditions, for example, the third-level response can be triggered only when the impedance suddenly changes for three times and the shear force simultaneously rises, to prevent false actions.

[0015] Preferably, the central coordination controller establishes a secure connection with the enterprise-level manufacturing cloud platform through a TLS1.3 encrypted channel, only uploads the desensitized quality trend summary and device health index, and the original sensor data and image stream are retained on the edge side processing; The system supports a remote firmware security update mechanism, and the update package needs to be verified by an asymmetric key in the hardware security element chip before it can be loaded and executed, to ensure the integrity of the control logic is not tampered with externally.

[0016] The beneficial effects of the present application are as follows: The IMT in-mold transfer quality monitoring control system for automotive interior described in the present application avoids overshoot current of power devices in the cold start stage through closed-loop linkage of state feedback and pre-charge calibration; maintains the consistency of multi-node event sequence under the condition of no global clock source by using distributed time stamp alignment mechanism; realizes full-link isolation processing of sensitive data in a trusted execution environment to prevent plaintext leakage to the general operating system; ensures that the fault diagnosis module can still complete abnormal identification based on local historical patterns during communication interruption; and effectively suppresses parasitic oscillation caused by high-frequency switching through a thermo-electric coupling impedance matching structure; This invention further integrates the sensing of the physical field inside the mold, the optical observation of the melt front, and the monitoring of the intrinsic electrical properties of the thin film to construct a dynamic quality characterization system covering the entire IMT transfer process. This enables the system to identify hidden process anomalies that may lead to final inspection defects during the injection molding filling stage. Through a feedforward-feedback composite control strategy, process adjustment actions are brought forward to before defect formation. By utilizing a multimodal cross-validation mechanism, the risk of misjudgment caused by single sensor failure or environmental interference is significantly reduced. With the help of a hierarchical control architecture, interpretable mapping from microscopic physical quantities to macroscopic quality indicators is achieved. Ultimately, this invention achieves knowability, traceability, judgment, and controllability of IMT in-mold transfer quality. Attached Figure Description

[0017] The invention will now be further described with reference to the accompanying drawings.

[0018] Fig. 1 This is a schematic diagram of the structure of an IMT (In-Mold Transfer) quality monitoring and control system applicable to automotive interiors according to the present invention; Fig. 2 This is a system block diagram showing the collaborative operation of the edge intelligent processing unit and the central coordination controller described in this invention. Detailed Implementation

[0019] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0020] like Figs. 1-2 As shown in the embodiment of the present invention, an IMT (In-Mold Transfer) quality monitoring and control system for automotive interiors comprises seven core subsystems: a mold body, a film supply mechanism, an injection molding execution unit, a multimodal online sensing module, an edge intelligent processing unit, a process parameter adjustment actuator, and a central coordination controller. Each subsystem achieves high-bandwidth, low-latency data interaction and command coordination through industrial Ethernet, CANopen bus, and RS485 communication links.

[0021] The mold body is integrally machined from H13 hot work die steel. The cavity surface is mirror polished to Ra≤0.02μm. Micron-level positioning grooves are set in the key positioning area. The grooves are 50±2μm deep and 200±5μm wide, distributed along the edge contour of the film, and form a mechanical interlocking structure with the pre-set alignment marks of the functional film. A miniature Hall sensor is embedded at the bottom of each positioning groove, with its sensing surface 0.1 mm away from the surface of the cavity, for detecting whether the film has completely fallen into the predetermined position. The Hall sensor outputs an analog voltage signal, which is converted by a 16-bit sigma-delta ADC and uploaded to the edge intelligent processing unit at a sampling rate of 1 kHz. Only when the output values of all Hall sensors corresponding to the positioning points fall within the preset threshold interval [1.8V, 2.2V], the central coordination controller generates an injection enable signal to allow the injection execution unit to enter the injection stage.

[0022] The mold body is integrated with a distributed sensing array, including a fiber Bragg grating strain sensing network, a miniature thermocouple array, and a piezoelectric interface shear force sensor. The fiber Bragg grating strain sensing network is composed of 12 FBGs (Fiber Bragg Gratings) with center wavelengths of 1548.3 nm to 1550.7 nm engraved on a single polarization-maintaining optical fiber. The adjacent grating spacing is 15 mm, and they are arranged in a serpentine shape along the main path of the melt flow, covering the entire cavity bottom and sidewall. The fiber Bragg grating strain sensing network monitors the local strain field evolution of the film under injection impact in real time through wavelength division multiplexing technology and a demodulator. The miniature thermocouple array is composed of 24 K-type thermocouples, which are welded to the inside of the cavity wall at a distance of 0.5 mm from the surface, with a spatial resolution of 10 mm x 10 mm. It is used to collect the temperature gradient distribution of the film contact area with the melt, with a measurement range of 20°C to 350°C and an accuracy of ±0.5°C. The piezoelectric interface shear force sensor uses PZT-5H piezoelectric ceramic sheets, with a total of 8 arranged at the 8 nodes where the melt front is expected to arrive. The output charge signal is converted to a voltage signal by a charge amplifier and input to the edge intelligent processing unit, which is used to capture the transient shear stress pulse caused by interlayer slip.

[0023] The film supply mechanism consists of a pay-off roller, a tension swing lever, a servo motor, and an encoder to form a closed-loop tension control system. The servo motor is equipped with a 23-bit absolute value encoder, connected to the central coordination controller through an EtherCAT bus. A high-precision tension sensor is installed on the tension swing lever, and its output signal is uploaded to the central coordination controller through an RS485 bus at a period of 10 ms. The tension control algorithm uses an adaptive PID structure, with the proportional gain Kp automatically switched according to the current film material type (such as PC / PMMA composite film or PET base film), and the integral time Ti and the differential time Td dynamically adjusted based on environmental temperature and humidity, ensuring that the tension fluctuation is controlled within ±0.3N.

[0024] The injection molding execution unit includes a full-electric injection molding machine body, the screw position of which is monitored in real time by a magnetostrictive displacement sensor, the injection pressure is collected by a piezoresistive sensor, the flow rate in the pressure maintaining stage is measured by a Coriolis mass flowmeter, and the melt front end temperature is obtained by directly inserting a K-type armored thermocouple probe into the front end of the flow channel. All process parameters are packaged into GOOSE messages according to the IEC61850-7-4 extension specification and transmitted to the edge intelligent processing unit through a gigabit industrial Ethernet.

[0025] The multi-modal online sensing module is composed of three parts: The first part is a high-speed line scan camera array composed of 4 CMOS line array cameras, the imaging window of which is sealed with fused quartz glass and installed inside the mold parting surface 2 mm away from the surface of the cavity. The camera trigger signal is generated by the incremental pulse of the injection screw displacement encoder through FPGA frequency division, ensuring that the film surface topography is continuously captured with a time resolution of 10 μs during the melt front advancing process. The image data is transmitted to the edge intelligent processing unit through the CameraLink interface, and the original image resolution is 4096 x N (N is the number of rows), and the gray level is 12 bits.

[0026] The second part is an infrared thermal imaging unit, which is arranged in a special observation cavity beside the mold cooling water channel using a non-cooled microbolometer focal plane array. The back surface of the film is measured by a non-contact temperature measurement through a high-temperature-resistant sapphire window. The infrared thermal imaging unit collects complete thermal distribution maps at a frame rate of 30 Hz during the pressure maintaining stage, with a temperature range of 50°C to 300°C and a spatial resolution of 0.5 mm / pixel. The thermal image data is transmitted through a gigabit Ethernet, and the synchronization accuracy is guaranteed by the PTP protocol, with a time jitter of <1 μs.

[0027] The third part is a conductive microgrid structure embedded in the bottom layer of the functional film, which is prepared by a roll-to-roll printing process using silver nanowires. After transfer printing, it is used as a capacitive sensing layer. During the injection molding process, the microgrid is used as an impedance measurement electrode, with a high-frequency alternating current excitation signal applied to both ends to measure its complex impedance Z(ω)=R+jX. The impedance analyzer records the changes of the real part R and the imaginary part X at a sampling rate of 1 kHz, which is used to indirectly reflect the change of the dielectric constant εr of the film under high temperature and high pressure. When microcracks or interlayer debonding occur in the film, the decrease in εr leads to a decrease in X and an increase in R. The change rate dZ / dt can be used as an early indicator of the initiation of microscopic defects.

[0028] The edge intelligent processing unit adopts a dual-core heterogeneous processor architecture, with the main core running the FreeRTOS real-time operating system and the sub-core being an NPU (neural network processing unit). The main core is responsible for processing hard real-time data streams from the mold internal sensors, including FBG demodulation data, thermocouple readings, shear force signals, etc., with a processing period of 100μs; the co-core runs a lightweight convolutional neural network for extracting features from high-speed line-scan images and identifying defect patterns (such as micro-cracks, bubbles, and overprint shifts) with an inference delay of <5ms. The two cores exchange intermediate results through a dual-port RAM shared buffer, and the DMA controller triggers data transfer operations based on a pre-set priority strategy (sensor data > image features > control instructions).

[0029] The edge intelligent processing unit is equipped with a timestamp alignment engine, which takes the injection cycle start pulse (triggered by the mold closed limit switch) as the global time zero point, and applies nanosecond-level timestamps to all input data streams. To compensate for the physical propagation delay of different sensor channels (such as FBG demodulation delay of about 200μs and thermocouple thermal inertia delay of about 5ms), the engine is based on a pre-calibrated delay model for correction; For example, for the i-th thermocouple channel, its original sampling time is Corrected by the following formula: ; Where, is the thermal response time constant of the channel, which is obtained by step heating experiment calibration, and the typical value is 3-8ms; After correction, all modal data are resampled to a unified time grid (step size 100μs) to generate a multi-dimensional state vector sequence , k=1,2,…,N.

[0030] The central coordination controller adopts a hierarchical control architecture, with the upper layer being a quality evolution reasoning machine based on a knowledge graph. The reasoning machine loads a multi-physical field coupling knowledge base of the IMT transfer process, which is stored in the form of RDF triples and contains entities (such as PC / PMMA film, melt temperature 280°C, shear stress >1.5MPa) and their causal relationships (such as if shear stress >1.5MPa and strain rate >10s⁻¹, then micro-crack probability >85%); The knowledge base is obtained by training with historical good product data (sample size >10 5 cycles), and uses Bayesian network modeling to handle uncertainty; At the end of each injection cycle, the reasoning machine receives the state vector sequence summary uploaded by the edge unit, combines the current material batch identifier and the workshop environment temperature and humidity (collected by Modbus RTU sensors), deduces the quality risk level of the next cycle (classified into Low / Medium / High three levels), and generates feedforward adjustment instructions.

[0031] The lower layer is a model predictive control (MPC) based process parameter regulator; The regulator is based on a simplified dynamic model of the injection molding process: ; ; where, is the system state (including melt front position, film maximum strain, interface average shear force, etc.), is the control input (screw speed, holding pressure, film supply tension), is the measurable output, and are the process noise and measurement noise, respectively; The matrix is obtained through system identification (using subspace method), and the MPC solves the following quadratic programming problem with constraints: ; ; where, is the prediction horizon, is the control horizon, and are the weight matrices, is the reference trajectory (set by the inference machine feedforward command); The optimization problem is solved using the interior point method, which takes <8ms to calculate, and outputs the optimal control sequence , which is sent to the injection molding machine servo drive and film supply mechanism controller through the CANopen protocol.

[0032] The system starts the pre-charging calibration process as soon as the mold is closed. During the mold clamping and holding stage (pressure 5 MPa, duration 3 s), the central coordination controller sends zero-point calibration instructions to the FBG network, and records the center wavelengths of each grating in the strain-free state ; At the same time, the reference impedance measurement loop of the conductive micro-grid is activated, an excitation signal of 1 kHz, 0.1 Vpp is applied, and the initial impedance is recorded ; This calibration data is stored in the FRAM (Ferroelectric Memory, capacity 4MB) of the edge unit as a reference for subsequent period strain calculation and impedance change rate , where is the FBG strain sensitivity coefficient.

[0033] After each mold opening and ejection, the system automatically executes the lens contamination detection program of the high-speed line scan camera in the idle window period before the next mold closing. The procedure turns off the external light source, and collects the background dark-field image , calculates the mean square deviation of the pixel gray scale: Where, is the number of effective pixels, is the average gray scale. If (corresponding signal-to-noise ratio < 30 dB), it is determined that the optical window pollution is excessive, and the gas curtain cleaning device is triggered for local blowing.

[0034] A bidirectional verification feedback loop is established between the multi-modal online sensing module and the injection molding execution unit; When the edge unit detects a suspected micro-crack in the image (CNN confidence > 0.85), it immediately queries the interface shear force sensor reading and the rate of impedance change at the same time. If it satisfies: Where, MPa, , it is determined to be a real defect, triggering an alarm stop; otherwise, it is considered as an optical artifact, recorded to the noise sample library for online fine-tuning of the CNN model.

[0035] The edge intelligent processing unit is configured with a local fault diagnosis state machine, with a pre-set three-level abnormal response strategy; First-level anomaly: single FBG channel failure (wavelength drift > 2 nm), the system automatically switches to the adjacent redundant channel, and the fault node is marked as Degraded; Second-level anomaly: image display overprint offset but Hall positioning signal normal, determined as a logic conflict, pause the current cycle, light up the HMI red warning light, and wait for manual reset; Third-level anomaly: the same defect mode (such as impedance mutation and shear force rising synchronously) appears for 3 consecutive cycles, call the knowledge graph root cause analysis module, generate a report: possible cause: melt temperature too high causing film softening too much, push to the MES system.

[0036] The central coordination controller is connected with the enterprise-level manufacturing cloud platform through a TLS1.3 encrypted channel, only uploading desensitized data: quality trend summary (such as defect rate week-on-week), device health index (based on sensor drift rate calculation); Raw images and sensor data are processed within the edge side trusted execution environment (TEE). Remote firmware update packages need to be signed by the RSA-2048 public key in the hardware security element (SE chip) before they can be loaded.

[0037] The above shows and describes the basic principles, main features and advantages of the present application. Those skilled in the art should understand that the present application is not limited to the above-mentioned embodiments, and the above-mentioned embodiments and descriptions in the specification are only to illustrate the principles of the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. An IMT in-mold transfer quality monitoring control system suitable for use in automotive interiors, characterized by, Comprise: A mold body, a film supply mechanism, an injection molding execution unit, a multi-modal online sensing module, an edge intelligent processing unit, a process parameter adjustment actuator, and a central coordination controller; The mold body is internally integrated with a distributed fiber bragg grating strain sensing network, a micro thermocouple array, and a piezoelectric interface shear force sensor, which are used to synchronously collect local strain field, temperature gradient distribution, and interlayer shear stress time series data of the contact area between the film and the melt during the injection filling stage; The film supply mechanism is equipped with a high-precision servo tension control system, and the tension feedback signal is uploaded in real time to the central coordination controller via an industrial communication bus; The injection molding execution unit includes a screw position encoder, an injection pressure sensor, a pressure holding stage flowmeter, and a melt front temperature probe, and all process parameters are packaged in a standardized industrial communication protocol and transmitted to the edge intelligent processing unit through an industrial Ethernet; The multi-modal online sensing module includes a high-speed line scan camera array arranged inside the mold parting surface, an infrared thermal imaging unit arranged in the mold cooling water channel bypass, and a conductive microgrid structure embedded in the bottom layer of the functional film, wherein the conductive microgrid acts as an impedance measurement electrode during the injection molding process, and detects the dielectric constant change of the film through a high-frequency alternating current excitation signal to reflect its microstructure integrity; The edge intelligent processing unit is configured with a dual-core heterogeneous processor architecture, the main core runs a real-time operating system to process hard real-time sensing data streams, the co-core runs a lightweight neural network inference engine to perform image feature extraction and defect pattern recognition, the main core and the co-core exchange intermediate calculation results through a shared buffer area, and a time stamp alignment engine is built-in, which time stamps all input data streams according to the injection cycle start pulse signal, and compensates and corrects the sampling time of different sensor channels based on a physical propagation delay model, to generate a synchronized multi-dimensional state vector sequence; The central coordination controller adopts a hierarchical control architecture, the upper layer is a quality evolution reasoning machine based on a knowledge graph, and the lower layer is a process parameter regulator based on model predictive control; The quality evolution reasoning machine loads a multi-physical field coupling knowledge base of the IMT transfer process, receives the state vector sequence from the edge intelligent processing unit, dynamically deduces the next cycle quality risk level combining material batch identification and environmental information, and generates a feedforward adjustment instruction; The process parameter regulator solves a constrained quadratic programming problem according to the feedforward instruction and real-time feedback deviation signal, outputs the optimal screw speed curve, pressure holding pressure set value, and film supply tension adjustment amount, and issues them to the injection molding execution unit and the film supply mechanism through an industrial field bus.

2. The IMT in-mold transfer quality monitoring control system for automotive interior according to claim 1, wherein The cavity surface of the mold body is provided with a positioning groove, and the positioning groove and the alignment mark of the edge of the functional film form a mechanical interlocking structure; A Hall sensor is embedded in the bottom of the positioning groove for detecting whether the film has completely fallen into the predetermined position; The injection molding execution unit of the central coordination controller enters the injection stage only when all positioning points are confirmed to be in place.

3. The IMT in-mold transfer quality monitoring control system for automotive interior according to claim 1, wherein The system starts a pre-charging calibration process in the mold closed state: In the mold clamping and holding phase, the central coordination controller sends a zero-point calibration instruction to the distributed fiber grating sensor network and activates the reference impedance measurement loop of the conductive micro-grid to eliminate system offsets introduced by mold thermal expansion or mechanical assembly errors; The calibration data is stored in the non-volatile memory of the edge intelligent processing unit.

4. The IMT in-mold transfer quality monitoring control system for automotive interior according to claim 1, wherein, During the idle window period between mold opening and the next mold closing, the system automatically executes the lens contamination detection program of the high-speed line scan camera: By analyzing the pixel mean square deviation of the background dark field image, the cleanliness of the optical window is judged. If it exceeds the preset safety threshold, the gas curtain cleaning device is triggered to perform local blowing.

5. The IMT in-mold transfer quality monitoring control system for automotive interior according to claim 1, wherein, A bidirectional verification feedback loop is established between the multi-modal online sensing module and the injection execution unit: When the edge intelligent processing unit detects image features of suspected micro-cracks or overlay offset on the film surface, it immediately queries the interface shear force sensor readings and the conductive micro-grid impedance change rate at the same time; If both show abnormal mutations, it is determined to be a real defect and triggers an alarm; On the contrary, if only the image feature is abnormal and the physical sensing data is normal, it is considered as optical interference or reflection artifact, and the system automatically ignores the false alarm and records it to the noise sample library for subsequent model optimization.

6. The IMT in-mold transfer quality monitoring control system for automotive interior according to claim 1, wherein, The edge intelligent processing unit is configured with a local fault diagnosis state machine, which has multiple levels of abnormal response strategies: Level one abnormality corresponds to a single sensor failure, the system automatically switches to the redundant channel and marks the fault node; Level two abnormality corresponds to multi-modal data logic conflict, the system suspends the current period and requests manual reset; Level three abnormality corresponds to the same defect pattern appearing in multiple consecutive periods, the system automatically calls the root cause analysis module in the knowledge graph, generates a process parameter drift report and pushes it to the manufacturing execution system; All state transitions must meet the preset timing verification conditions.

7. The IMT in-mold transfer quality monitoring control system for automotive interior according to claim 1, characterized in that, The central coordination controller connects with the enterprise-level manufacturing cloud platform through an encrypted secure channel, only uploads the desensitized quality trend summary and device health index, and the original sensing data and image stream are processed in the trusted execution environment on the edge side; The system supports remote firmware security update mechanism, the update package needs to be verified by the asymmetric key in the hardware security element chip before it can be loaded and executed.

8. The IMT in-mold transfer quality monitoring control method for automotive interior trim, applicable to the IMT in-mold transfer quality monitoring control system for automotive interior trim according to any one of claims 1-7, characterized in that, The method comprises the following steps: After the mold is closed, perform pre-charging calibration to obtain the fiber grating zero-point wavelength and the reference impedance of the conductive micro-grid; During the injection filling stage, capture the film surface transient topography through the high-speed line scan camera array, and simultaneously collect strain field, temperature field, shear force and time-varying impedance data; Use the timestamp alignment engine to synchronize and delay compensate the multi-source heterogeneous data at nanosecond level, generate a multi-dimensional state vector sequence under the unified time grid; Based on the state vector sequence, the quality evolution reasoning machine combines the material batch and environmental information to infer the next period quality risk level; The model predictive controller solves the constrained optimization problem to generate a composite control instruction, dynamically adjusting the screw speed curve, holding pressure and film tension; In defect discrimination, multi-modal cross-validation is performed: only when image abnormalities and physical sensing abnormalities occur simultaneously, is it determined to be a real defect and triggers a response.

9. The IMT in-mold transfer quality monitoring control method for automotive interior according to claim 8, characterized by, The construction of the multi-dimensional state vector sequence includes: The data of thermocouple channel is compensated by time delay according to its thermal response time constant, the data of fiber Bragg grating is calculated by local strain according to wavelength shift, the data of conductive micro-mesh is calculated by change rate relative to reference value, and the high-speed line-scan image and infrared thermal image are resampled to the same time step.

10. The IMT in-mold transfer quality monitoring control method for automotive interior according to claim 8, characterized by, The system dynamic model adopted by the model predictive controller is obtained by subspace identification method, whose state variables include melt front position, film maximum strain and interface average shear force, control inputs include screw speed, holding pressure and film supply tension, and the optimization goal is to minimize the weighted deviation between output variables and reference trajectories set by the inference machine, while the control increment change rate and the physical limit of the actuator are constrained.

Images