Flow rate control method and control system for liquid silica gel injection molding

Abstract

The invention discloses a material flow rate control method and system for liquid silica gel injection molding, and the method comprises the following steps: S1, collecting temperature gradient distribution in an injection mold, silica gel initial viscosity and pressure intensity data in a mold cavity in real time through a multi-mode sensor network, and inputting the data into a pre-molding parameter model; s2, based on the pre-forming parameter model; s3, outputting a result according to the dynamic response equation; and S4, dynamically calibrating the regulation and control parameter set through a fuzzy neural network in combination with a closed-loop feedback mechanism, and driving an execution mechanism to adjust the silica gel injection rate based on the calibrated parameters. The industrial bottleneck of liquid silica gel injection molding flow rate control is systematically solved; the cooperative construction of the multi-modal sensing network and the multi-field coupling model breaks through the physical limitation of a traditional empirical formula, and realizes the improvement of the dynamic response precision under a complex working condition.

Description

Technical Field

[0001] The present invention belongs to the field of injection molding flow control, and more specifically, particularly relates to a method for controlling the flow rate of liquid silicone rubber injection molding. At the same time, the present invention also relates to a control system for the flow rate of liquid silicone rubber injection molding. Background Art

[0002] In the field of liquid silicone rubber injection molding, precise control of the flow rate is the core link to ensure the molding quality of products. The existing technical solutions have the following limitations:

[0003] Insufficient dynamic response: Traditional methods rely on fixed-parameter models and are difficult to adapt to the strong nonlinear and time-varying characteristics of the silicone curing process. The dynamic coupling effect of temperature gradient and viscosity is not fully modeled, resulting in the drift phenomenon between flow rate prediction and actual process parameters;

[0004] Lack of multi-field collaboration: The existing control system lacks a systematic characterization of the dynamic interaction of multiple physical fields such as heat-flow-force during the injection molding process, and the correlation modeling between the die geometry topology and the material rheological properties remains at the level of static empirical formulas;

[0005] Limitations of closed-loop optimization: Traditional proportional-integral-derivative controllers lack intelligent decision-making capabilities in the connection of dynamic weight adjustment and curing stage conversion, and the global optimization of control parameters is limited by fixed gain coefficients;

[0006] Weak anomaly fault tolerance: The corrective ability for sudden changes in material properties or interference from the internal microstructure of the die is insufficient, and the self-diagnosis mechanism is mostly based on threshold judgment, without realizing the deep collaboration of multi-dimensional feedback links;

[0007] Based on the above, we propose a method and control system for controlling the flow rate of liquid silicone rubber injection molding to solve the problems existing in the prior art. Summary of the Invention

[0008] The purpose of the present invention is to solve the disadvantages existing in the prior art, and to propose a method and control system for controlling the flow rate of liquid silicone rubber injection molding, which systematically solves the industry bottleneck of liquid silicone rubber injection flow rate control; the collaborative construction of a multi-modal perception network and a multi-field coupling model breaks through the physical limitations of traditional empirical formulas and realizes the improvement of dynamic response accuracy under complex working conditions.

[0009] To achieve the above purpose, the present invention provides the following technical solutions:

[0010] A method for controlling the flow rate of liquid silicone rubber injection molding, comprising the following steps:

[0011] S1. Real-time collect the temperature gradient distribution, initial silicone viscosity, and cavity pressure data inside the injection mold through a multi-modal sensor network, and input the data into the preform parameter model;

[0012] S2. Based on the preform parameter model, combined with the mold geometric topology and silicone curing kinetics characteristics, construct a non-linear dynamic flow rate response equation;

[0013] S3. According to the output result of the dynamic response equation, generate a set of control parameters through a variable-weight proportional-integral-derivative controller, and the set of control parameters includes the injection pressure gradient value, screw speed correction amount, and valve opening and closing duty cycle;

[0014] S4. Dynamically calibrate the set of control parameters through a fuzzy neural network combined with a closed-loop feedback mechanism, and drive the actuator to adjust the silicone injection rate based on the calibrated parameters.

[0015] Preferably, the deployment strategy of the multi-modal sensor network in step S1 includes a cross-scale spatial perception architecture:

[0016] Arrange an ultra-high-precision micro-sensor array in the mold gate area to capture turbulent pulsation characteristics, use a broadband resonant sensor group at the end of the cavity to detect capillary flow micro-vibration signals, and embed an anti-shear stress sensor chain at the mold parting surface;

[0017] All sensor nodes achieve time-domain synchronization through a heterogeneous communication protocol, and the de-jitter processing of viscosity data uses a sliding window dynamic baseline correction method to eliminate the interference of pseudo-signals caused by pressure mutations.

[0018] Preferably, the construction process of the non-linear dynamic flow rate response equation in step S2 includes a multi-physics field coupling optimization engine:

[0019] Discretely divide the mold through a three-dimensional unstructured grid division module, and establish a mass-momentum double conservation constraint equation set for each cell in combination with the non-isothermal rheological constitutive theory;

[0020] Introduce a viscoelastic relaxation correction technique in the iterative solution process, use an implicit Euler-Lagrange hybrid algorithm to dynamically balance and decouple the shear rate field and the temperature gradient field, and achieve numerical stability control under multi-field coupling conditions through the singular value decomposition of the Jacobian matrix.

[0021] Preferably, the parameter optimization of the variable-weight proportional-integral-derivative controller in step S3 adopts a hierarchical logic architecture:

[0022] Keep the initial setting of traditional PID parameters in the basic control layer, adjust the dynamic range of the gain coefficient according to the solid-phase transformation rate of silicone in the adaptation adjustment layer, and introduce a non-linear saturation function in the over-limit compensation layer to suppress the integral term accumulation error;

[0023] The information interaction among the three levels is realized through a bidirectional fuzzy inference channel, and in the orthogonal decomposition module of the error signal, a Haar wavelet basis is adopted to establish a multi-resolution feature representation.

[0024] Preferably, the training mechanism of the fuzzy neural network in step S4 includes a hybrid knowledge injection strategy:

[0025] In the initial model construction stage, the process expert experience rule base is imported to form a prior topological structure, and in the online learning stage, the real-time feedback data stream is fused through an incremental weight adjustment algorithm;

[0026] The activation function of the network hidden layer adopts a differentiable Gaussian membership function generator, and its central parameter is dynamically calibrated through a hybrid optimization engine of backpropagation and genetic algorithm;

[0027] During the training process, an adversarial correction module is synchronously run to enhance the generalization robustness of the control strategy.

[0028] Preferably, the closed-loop feedback mechanism includes a multi-dimensional cross-validation link:

[0029] The dynamic following of the screw thrust is realized on the millisecond time scale through the first-level fast response loop, the valve action timing strategy is corrected based on the pressure field balance index by the second-level fine adjustment loop, and the hyperparameters of the preform model are reversely adjusted according to the finished product optical interference detection data by the third-level global optimization loop;

[0030] Multi-rate data fusion is achieved among the loops through a digital twin interface, and the consistency of the control timing is ensured through an event-driven message queue.

[0031] Preferably, the actuator adjustment process includes a kinematic redundancy compensation module:

[0032] For the joint clearance error of the multi-degree-of-freedom hybrid manipulator, a dynamic compensation model based on Lyapunov stability theory is established; a predictive trajectory planning algorithm is started in the high-speed commutation stage, and a smooth speed transition curve is generated through Bezier curve interpolation;

[0033] At the same time, for the non-linear hysteresis effect of the hydraulic drive subsystem, a composite correction method combining inverse model feedforward compensation and sliding mode variable structure control is adopted.

[0034] 8. A flow rate control system for liquid silicone injection molding, which is used to implement the above method, and includes:

[0035] A multi-modal sensor network module, deployed in the injection mold, configured to collect temperature gradient distribution, initial silicone viscosity, and cavity pressure data in real time;

[0036] The data fusion center module is connected to the multi-modal sensor network module and is configured to perform multi-source heterogeneous signal fusion on the collected data and generate a dynamic characteristic data set based on a preformed parameter model;

[0037] The dynamic flow rate generation module is connected to the data fusion center module and is configured to construct a non-linear dynamic flow rate response equation according to the mold geometric topology, silicone curing kinetics characteristics, and the dynamic characteristic data set; The variable weight ratio-integral-differential controller module is connected to the dynamic flow rate generation module and is configured to analyze the dynamic flow rate response equation and generate a set of control parameters including the injection pressure gradient value, screw speed correction amount, and valve opening and closing duty cycle;

[0038] The fuzzy neural network calibration module is connected to the variable weight ratio-integral-differential controller module and is configured to dynamically calibrate the set of control parameters through a closed-loop feedback mechanism and output a calibration signal;

[0039] The actuator drive module is connected to the fuzzy neural network calibration module and is configured to convert the calibration signal into a mechanical control instruction to adjust the silicone injection rate.

[0040] Preferably, the multi-modal sensor network module includes a sensor cluster deployed distributively:

[0041] The turbulent flow characteristic sensing sub-module is arranged in the mold gate area and includes a composite sensor group of a high-frequency infrared temperature array and an ultrasonic viscosity probe;

[0042] The end vibration detection sub-module is installed at the end of the mold cavity and consists of a broadband piezoelectric resonance sensor and an anti-shear stress sensor chain;

[0043] The synchronous communication sub-module uses a time-division multiplexing protocol to achieve time-domain synchronization of all sensor nodes and eliminates electromagnetic coupling effects through an anti-interference shielding layer;

[0044] Each sub-module is connected to the input interface of the data fusion center module through a heterogeneous bus.

[0045] The technical effects and advantages of the present invention: A method and a control system for controlling the flow rate of liquid silicone injection molding provided by the present invention have the following effects compared with the prior art:

[0046] By constructing a non-linear flow rate response equation that integrates mold geometric topology and curing kinetics characteristics, dynamic analysis of the thermal-fluid-mechanical field coupling effect is realized. The three-dimensional grid meshing technology and the implicit Euler-Lagrange hybrid algorithm are used in combination to accurately characterize the phase change characteristics of the viscoelastic behavior of silicone at different temperature stages, and effectively suppress the flow rate fluctuations caused by the material shear thinning effect. The multi-physical field collaborative solver can autonomously compensate for the flow path deviation caused by mold deformation and improve the uniformity of complex cavity filling;

[0047] Adopt a dynamic calibration architecture combining fuzzy neural network and closed-loop feedback to break through the parameter rigidity bottleneck of traditional control systems. The knowledge injection interface combines incremental learning algorithms to achieve the deep integration of expert experience and real-time data, and adaptively adjusts the matching relationship between the valve opening and closing timing and the screw thrust. The bidirectional fuzzy inference channel supports the multi-resolution decomposition of error signals, significantly reducing the risk of flow rate oscillation caused by sudden changes in pressure gradients;

[0048] The joint optimization of the dynamic flow rate generation module and the multi-objective evolutionary algorithm realizes the Pareto optimal balance between energy consumption cost and molding quality. By establishing an associated model between the activation energy of silica gel and the temperature gradient, the thermodynamic distribution state of the curing reaction process is precisely controlled, reducing ineffective heat dissipation. The predictive trajectory planning module of the actuator reduces mechanical transmission energy consumption, and the adaptive backpressure regulation mechanism suppresses turbulent losses. Brief Description of the Drawings

[0049] Figure 1 It is a flow chart of the method for controlling the flow rate of liquid silicone injection molding of the present invention. Detailed Embodiments

[0050] In order to make the objectives, technical solutions and advantages of the present invention more clear and understandable, the following further details the present invention in conjunction with specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the scope of protection of the present invention.

[0051] The present invention provides a method and a control system for controlling the flow rate of liquid silicone injection molding as shown in Figure 1 Through deep mechanism innovation, the industry bottleneck of liquid silicone injection molding flow rate control is systematically solved; the collaborative construction of the multi-modal perception network and the multi-field coupling model breaks through the physical limitations of traditional empirical formulas and improves the dynamic response accuracy under complex working conditions; the combined design of the intelligent calibration mechanism and the robust control architecture ensures the stable output of the system under material parameter fluctuations and process anomalies; the dual-drive mode of the global optimization algorithm and the energy efficiency management module significantly reduces energy consumption while ensuring molding quality; finally, an intelligent control system with self-perception, real-time optimization, and fast fault tolerance capabilities is formed, providing a reliable technical guarantee for the large-scale production of high-precision silicone products.

[0052] The method for controlling the flow rate of liquid silicone injection molding includes the following steps:

[0053] S1. Real-time collect the temperature gradient distribution, initial silicone viscosity, and cavity pressure data inside the injection mold through a multi-modal sensor network, and input the data into the preform parameter model. In step S1, the multi-modal sensor network includes a high-frequency infrared temperature array, an ultrasonic viscosity probe, and a distributed piezoresistive sensor. Among them, the sampling frequency of the temperature data is 5 - 20 kHz, and the sensitivity threshold of the viscosity probe is ±0.02 Pa·s;

[0054] The deployment strategy of the multi-modal sensor network in step S1 includes a cross-scale spatial perception architecture: arrange an ultra-high-precision micro-sensor array in the mold gate area to capture turbulent pulsation characteristics, use a broadband resonant sensor group at the end of the mold cavity to detect capillary flow micro-vibration signals, and embed an anti-shear stress sensor chain at the mold parting surface; all sensor nodes achieve time-domain synchronization through heterogeneous communication protocols. Among them, the de-jitter processing of the viscosity data uses the sliding window dynamic baseline correction method to eliminate the interference of pseudo-signals caused by pressure mutations.

[0055] S2. Based on the preform parameter model, combined with the mold geometric topology structure and silicone curing kinetics characteristics, construct a non-linear dynamic flow rate response equation. The construction process of the non-linear dynamic flow rate response equation in step S2 includes a multi-physical field coupling optimization engine: discretely divide the mold through a three-dimensional unstructured grid division module, and establish a mass-momentum double conservation constraint equation set for each cell in combination with the non-isothermal rheological constitutive theory; introduce a viscoelastic relaxation correction technology in the iterative solution process, use an implicit Euler-Lagrange hybrid algorithm to dynamically balance and decouple the shear rate field and the temperature gradient field, and achieve numerical stability control under multi-field coupling conditions through the singular value decomposition of the Jacobian matrix;

[0056] The non-linear dynamic flow rate response equation in step S2 further includes a process anomaly self-diagnosis module: when the deviation between the model output flow rate prediction value and the real-time monitoring value exceeds the set threshold, start a defect tracing program based on the causal inference tree, and identify mold blockage or material phase change anomalies by decomposing the pressure drop fluctuation map; for the pressure oscillation phenomenon caused by poor exhaust, automatically generate a spiral injection path re-planning strategy, and optimize the injection speed configuration matrix in combination with the gradient descent method.

[0057] The construction of the non-linear dynamic flow rate response equation in step S2 satisfies the following formula:

[0058] Among them, Q(t) is the time-dependent silicone volume flow rate, ΔP represents the pressure difference between the mold inlet and the end, μ(t) is the temperature-dependent dynamic viscosity coefficient, α is the basic flow resistance compensation factor, β is the shear thinning index, γ is the time window adjustment coefficient, τ is the curing characteristic time constant, ω represents the temperature gradient weight parameter, and δ is the non-linear amplification coefficient;

[0059] The calculation of the dynamic viscosity coefficient μ(T) follows:

[0060] where μ0 is the reference viscosity, E α is the activation energy of silica gel, R is the gas constant, T is the real-time temperature, T g is the glass transition temperature, ξ is the temperature-sensitive adjustment factor, and σ is the steepness parameter of the phase change interval.

[0061] S3. According to the output result of the dynamic response equation, a set of control parameters is generated by a variable-weight proportional-integral-derivative controller. The set of control parameters includes the injection pressure gradient value, the correction amount of the screw speed, and the duty ratio of the valve opening and closing. In step S3, the parameter optimization of the variable-weight proportional-integral-derivative controller adopts a hierarchical logic architecture: the initial settings of the traditional PID parameters are maintained at the basic control layer, the dynamic range of the gain coefficient is adjusted according to the solid-phase transformation rate of silica gel at the adaptation adjustment layer, and a non-linear saturation function is introduced at the over-limit compensation layer to suppress the integral term accumulation error. The information interaction between the three layers is realized through a two-way fuzzy inference channel, and the orthogonal decomposition module of the error signal uses the Haar wavelet basis to establish a multi-resolution feature representation.

[0062] S4. The set of control parameters is dynamically calibrated through a fuzzy neural network combined with a closed-loop feedback mechanism, and the injection rate of silica gel is adjusted by driving the actuator based on the calibrated parameters. The training mechanism of the fuzzy neural network in step S4 includes a hybrid knowledge injection strategy: the process expert experience rule base is imported at the initial model construction stage to form a prior topological structure, and the real-time feedback data stream is fused through an incremental weight adjustment algorithm at the online learning stage. The activation function of the network hidden layer uses a differentiable Gaussian membership function generator, and its central parameter is dynamically calibrated through a hybrid optimization engine of backpropagation and genetic algorithm. The adversarial correction module is synchronously run during the training process to enhance the generalization robustness of the control strategy.

[0063] The closed-loop feedback mechanism includes a multi-dimensional cross-validation link: the dynamic following of the screw thrust is realized at the millisecond time scale through the first-level fast response loop, the valve action timing strategy is corrected based on the pressure field balance index at the second-level fine adjustment loop, and the hyperparameters of the preform model are adjusted backward according to the optical interference detection data of the finished product at the third-level global optimization loop. The multi-rate data fusion is realized between the loops through a digital twin interface, and the consistency of the control timing is ensured through an event-driven message queue.

[0064] The dynamic calibration process in step S4 adopts a multi-objective co-evolution algorithm: a Pareto front evaluation model including an energy consumption cost item, a molding cycle item, and a quality standard deviation item is established, and the optimal configuration of the regulation parameter set is searched through a non-dominated sorting genetic algorithm; an elite retention strategy and a population diversity maintenance mechanism are introduced during the evolution process, and the global search and local development capabilities are balanced through an adaptive crossover operator; the elite solution sets generated in each iteration are stored in the knowledge graph database for enhancing the knowledge reserve of feedback control.

[0065] The actuator adjustment process includes a kinematic redundancy compensation module: for the joint clearance error of a multi-degree-of-freedom hybrid manipulator, a dynamic compensation model based on Lyapunov stability theory is established; a predictive trajectory planning algorithm is started during the high-speed commutation stage, and a smooth speed transition curve is generated through Bezier curve interpolation; at the same time, for the nonlinear hysteresis effect of the hydraulic drive subsystem, a composite correction technology combining inverse model feedforward compensation and sliding mode variable structure control is adopted.

[0066] This embodiment also proposes a flow rate control system for liquid silicone injection molding for implementing the above method, including:

[0067] A multi-modal sensor network module, deployed inside the injection mold, configured to collect temperature gradient distribution, initial silicone viscosity, and cavity pressure data in real time; the multi-modal sensor network module includes a sensor cluster deployed distributively:

[0068] A turbulent flow characteristic sensing sub-module, arranged in the mold gate area, including a composite sensor group of a high-frequency infrared temperature array and an ultrasonic viscosity probe;

[0069] A terminal vibration detection sub-module, installed at the end of the cavity, composed of a broadband piezoelectric resonance sensor and a shear stress-resistant sensor chain;

[0070] A synchronous communication sub-module, which realizes the time-domain synchronization of all sensor nodes using a time-division multiplexing protocol and eliminates the electromagnetic coupling effect through an anti-interference shielding layer;

[0071] Each sub-module is connected to the input interface of the data fusion center module through a heterogeneous bus.

[0072] A data fusion center module, connected to the multi-modal sensor network module, configured to perform multi-source heterogeneous signal fusion on the collected data and generate a dynamic characteristic data set based on a preform parameter model; the data fusion center module includes:

[0073] A signal preprocessing unit, configured to perform wavelet noise reduction processing on the temperature signal, implement sliding window dynamic baseline correction on the viscosity data, and perform mutation pulse filtering on the pressure signal;

[0074] The spatio-temporal alignment unit uses a Kalman filter bank to achieve the unification of spatio-temporal benchmarks for multi-modal data;

[0075] The feature extraction unit extracts viscoelastic feature maps and thermodynamic coupling coefficients from the fused data through a convolutional neural network;

[0076] The processed data is input into the dynamic flow rate generation module through a high-speed data channel.

[0077] The dynamic flow rate generation module is connected to the data fusion center module and is configured to construct a non-linear dynamic flow rate response equation based on the die geometric topology, silicone curing kinetics characteristics, and dynamic characteristic datasets; The variable weight proportional-integral-differential controller module is connected to the dynamic flow rate generation module and is configured to analyze the dynamic flow rate response equation and generate a set of control parameters including injection pressure gradient values, screw speed correction amounts, and valve opening and closing duty cycles; The dynamic flow rate generation module includes:

[0078] The three-dimensional grid meshing unit generates an unstructured tetrahedral computational grid based on the die CAD model;

[0079] The multi-field coupling solver uses an implicit Euler-Lagrange hybrid algorithm to simultaneously solve the mass conservation equation, momentum transport equation, and energy balance equation;

[0080] The perturbation response unit is configured to inject a random field perturbation term into the solver and analyze the flow rate fluctuation tolerance;

[0081] The output end of this module is connected to the variable weight proportional-integral-differential controller module through a bidirectional data link.

[0082] The fuzzy neural network calibration module is connected to the variable weight proportional-integral-differential controller module and is configured to dynamically calibrate the set of control parameters through a closed-loop feedback mechanism and output a calibration signal; The variable weight proportional-integral-differential controller module includes a hierarchical control architecture:

[0083] The basic PID layer is a traditional proportional-integral-differential operation unit that maintains a fixed gain coefficient;

[0084] The dynamic tuning layer adjusts the weight coefficients of each control term in real time according to the silicone curing degree;

[0085] The over-limit protection layer uses a hyperbolic tangent function to suppress the saturation phenomenon of output parameters;

[0086] The seamless switching of control strategies is achieved between each layer through a fuzzy decision interface;

[0087] The fuzzy neural network calibration module includes:

[0088] The knowledge injection interface imports the process expert experience rule base to construct the initial network topology;

[0089] An incremental learning unit that dynamically updates network weights through an online error backpropagation algorithm;

[0090] A robustness enhancement unit that runs an adversarial sample generator to verify the stability of the control strategy;

[0091] The calibration signal is transmitted to the actuator drive module through an optical fiber communication ring network.

[0092] An actuator drive module, connected to the fuzzy neural network calibration module, configured to convert the calibration signal into a mechanical control instruction to adjust the silicone injection rate; the actuator drive module consists of multiple control links:

[0093] An instruction parsing unit that decodes the calibration signal into a hydraulic valve opening instruction, a screw motor torque instruction, and a back pressure adjustment amount;

[0094] A motion compensation unit that uses a predictive trajectory planning algorithm to generate a smooth velocity curve for mechanical transmission clearance;

[0095] A redundant control unit that automatically switches to a backup drive link when an actuator failure is detected; the drive instruction is transmitted to the hydraulic servo system and the electric stepping device through the CAN bus;

[0096] In addition, the system also includes:

[0097] A closed-loop feedback verification module, connected to the actuator drive module and the fuzzy neural network calibration module, including:

[0098] A laser Doppler velocimeter that obtains actual flow rate data in real time;

[0099] A residual analysis unit that calculates the deviation between the set flow rate and the measured value;

[0100] A recombination strategy generator that triggers a parameter reconstruction instruction when the deviation exceeds the limit;

[0101] The verification result is transmitted back to the dynamic flow rate generation module through the feedback channel for online model update.

[0102] In summary, this embodiment has the following effects:

[0103] Multi-field dynamic coupling modeling optimization, by constructing a non-linear flow rate response equation that integrates mold geometry topology and curing kinetics characteristics, realizes the dynamic analysis of the thermal-fluid-stress field coupling effect. The three-dimensional mesh generation technology is combined with the implicit Euler-Lagrange hybrid algorithm to accurately characterize the phase change characteristics of the viscoelastic behavior of silicone at different temperature stages, effectively suppressing the flow rate fluctuations caused by the material shear thinning effect. The multi-physics field collaborative solver can autonomously compensate for the flow path deviation caused by mold deformation, improving the uniformity of complex cavity filling.

[0104] Intelligent real-time calibration mechanism, adopting a dynamic calibration architecture of fuzzy neural network and closed-loop feedback, breaks through the parameter rigidity bottleneck of traditional control systems. The knowledge injection interface combines with the incremental learning algorithm to achieve the deep integration of expert experience and real-time data, and adaptively adjusts the matching relationship between the valve opening / closing timing and the screw thrust. The bidirectional fuzzy inference channel supports the multi-resolution decomposition of error signals, significantly reducing the risk of flow rate oscillation caused by sudden changes in pressure gradient.

[0105] Robust abnormal response ability. The multi-dimensional cross-validation mechanism quickly identifies abnormal conditions such as bubble accumulation and local curing through a combined monitoring of laser velocimetry and residual analysis. The diagnostic module based on the causal inference tree can trace the source of abnormal pressure drop and trigger the spiral injection molding path optimization program. The synergistic effect of the tabu search strategy and the redundant control unit ensures that the system stability can still be maintained through historical mode reconstruction when the sensor fails.

[0106] Global energy efficiency collaborative improvement. The joint optimization of the dynamic flow rate generation module and the multi-objective evolutionary algorithm realizes the Pareto optimal balance between energy consumption cost and molding quality. Through the correlation modeling of the activation energy of silica gel and the temperature gradient, the thermodynamic distribution state of the curing reaction process is precisely controlled, reducing ineffective heat dissipation. The predictive trajectory planning module of the actuator reduces the mechanical transmission energy consumption, and the adaptive backpressure adjustment mechanism suppresses the turbulent loss.

[0107] Finally, it should be noted that the above are only the preferred embodiments of the present invention and are not used to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, for those skilled in the art, they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for controlling the flow rate of liquid silicone rubber during injection molding, characterized in that: The following steps are involved: S1, collecting data on temperature gradient distribution, initial viscosity of silicone and pressure in the mold cavity in real time in the injection mold through a multimodal sensor network, and inputting the data into a preforming parameter model; S2. Based on the preforming parameter model, combined with the mold geometric topology and the curing dynamics of silicone, a nonlinear dynamic flow rate response equation is constructed; S3. Generate a control parameter set through a variable weight proportional-integral-differential controller according to the output result of the dynamic response equation, wherein the control parameter set includes an injection pressure gradient value, a screw speed correction value, and a valve opening and closing duty cycle; S4. Dynamically calibrate the control parameter set through a fuzzy neural network combined with a closed-loop feedback mechanism, and drive the actuator to adjust the silicone injection rate based on the calibrated parameters.

2. The flow rate control method for liquid silicone injection molding according to claim 1, characterized in that The deployment strategy of the multimodal sensor network described in step S1 includes a cross-scale spatial perception architecture: An ultra-high-precision micro-sensor array is arranged in the mold gate area to capture the turbulent pulsation characteristics, a broadband resonant sensor group is used at the end of the mold cavity to detect the capillary flow micro-vibration signal, and an anti-shear stress sensor chain is embedded at the mold parting surface; All sensor nodes are synchronized in the time domain through heterogeneous communication protocols. The de-jitter processing of viscosity data adopts the sliding window dynamic baseline correction method to eliminate the false signal interference caused by sudden pressure changes.

3. A method for controlling the flow rate of liquid silicone rubber during injection molding according to claim 1, characterized in that, The construction process of the nonlinear dynamic flow rate response equation in step S2 includes a multi-physics field coupling optimization engine: The mold is discretized through the three-dimensional unstructured grid division module, and the mass-momentum dual conservation constraint equations of each unit cell are established in combination with the non-isothermal rheological constitutive theory; Viscoelastic relaxation correction technology is introduced in the iterative solution process, and the implicit Euler-Lagrangian hybrid algorithm is used to dynamically balance the shear rate field and the temperature gradient field. The numerical stability control under multi-field coupling conditions is achieved through the Jacobian matrix singular value decomposition.

4. A method for controlling the flow rate of liquid silicone rubber injection molding according to claim 1, characterized in that, The parameter optimization of the variable weight proportional-integral-derivative controller in step S3 adopts a hierarchical logic architecture: The initial setting of traditional PID parameters is maintained in the basic control layer, the dynamic range of gain coefficient is adjusted according to the silica gel solid phase transformation rate in the adaptive adjustment layer, and a nonlinear saturation function is introduced in the over-limit compensation layer to suppress the accumulation error of the integral term; The information interaction among the three levels is realized through a bidirectional fuzzy inference channel, in which the orthogonal decomposition module of the error signal uses the Haar wavelet basis to establish a multi-resolution feature expression.

5. A method for controlling the flow rate of liquid silicone rubber during injection molding, as claimed in claim 1, characterized in that, The training mechanism of the fuzzy neural network in step S4 includes a hybrid knowledge injection strategy: In the initial model building stage, the process expert experience rule base is introduced to form a priori topological structure, and in the online learning stage, the real-time feedback data stream is integrated through the incremental weight adjustment algorithm; The activation function of the network hidden layer adopts a differentiable Gaussian membership generator, whose central parameters are dynamically calibrated through a hybrid optimization engine of back propagation and genetic algorithm; The adversarial correction module is run synchronously during the training process to enhance the generalization robustness of the control strategy.

6. A method for controlling the flow rate of liquid silicone rubber injection molding according to claim 1, characterized in that, The closed-loop feedback mechanism includes a multi-dimensional cross-validation link: The dynamic following of the screw thrust is realized on a millisecond time scale through the first-level fast response loop. The second-level fine adjustment loop corrects the valve action timing strategy based on the pressure field balance index. The third-level global optimization loop reversely adjusts the hyperparameters of the preform model according to the finished product optical interference detection data; Multi-rate data fusion is achieved between the loops through the digital twin interface, and the consistency of the control timing is ensured through the event-driven message queue.

7. A method for controlling the flow rate of liquid silicone rubber injection molding according to claim 1, characterized in that, The adjustment process of the actuator includes a kinematic redundancy compensation module: For the joint clearance error of the multi-degree-of-freedom hybrid manipulator, a dynamic compensation model based on the Lyapunov stability theory is established; the predictive trajectory planning algorithm is started during the high-speed commutation stage, and a smooth speed transition curve is generated through Bezier curve interpolation; At the same time, for the nonlinear hysteresis effect of the hydraulic drive subsystem, a composite correction method combining inverse model feedforward compensation and sliding mode variable structure control is adopted.

8. A flow rate control system for liquid silicone injection molding, characterized in that, The system is used to implement the method according to any one of claims 1-7, including: A multi-modal sensor network module, deployed in the injection mold, configured to collect temperature gradient distribution, initial silicone viscosity, and cavity pressure data in real time; A data fusion center module, connected to the multi-modal sensor network module, configured to perform multi-source heterogeneous signal fusion on the collected data and generate a dynamic characteristic data set based on the preform parameter model; A dynamic flow rate generation module, connected to the data fusion center module, configured to construct a nonlinear dynamic flow rate response equation according to the mold geometric topology, silicone curing kinetics characteristics, and dynamic characteristic data set; a variable weight proportional-integral-differential controller module, connected to the dynamic flow rate generation module, configured to analyze the dynamic flow rate response equation and generate a set of control parameters including injection pressure gradient values, screw speed correction amounts, and valve opening and closing duty cycles; A fuzzy neural network calibration module, connected to the variable weight proportional-integral-differential controller module, configured to dynamically calibrate the set of control parameters through a closed-loop feedback mechanism and output a calibration signal; An actuator drive module, connected to the fuzzy neural network calibration module, configured to convert the calibration signal into a mechanical control instruction to adjust the silicone injection rate.

9. A flow rate control system for liquid silicone injection molding according to claim 8, characterized in that, The multi-modal sensor network module includes a distributed sensor cluster: A turbulent flow characteristic sensing sub-module, arranged in the mold gate area, including a composite sensor group of a high-frequency infrared temperature array and an ultrasonic viscosity probe; A terminal vibration detection sub-module, installed at the end of the cavity, composed of a broadband piezoelectric resonance sensor and an anti-shear stress sensor chain; A synchronous communication sub-module, which realizes the time-domain synchronization of all sensor nodes using the time-division multiplexing protocol and eliminates the electromagnetic coupling effect through an anti-interference shielding layer; Each sub-module is connected to the input interface of the data fusion center module through a heterogeneous bus.

Images