CN-121978965-A - Intelligent optimization method and system for composite process parameters of multilayer polypropylene film

CN121978965ACN 121978965 ACN121978965 ACN 121978965ACN-121978965-A

Abstract

The invention belongs to the field of intelligent optimization of equipment parameters, and relates to an intelligent optimization method and system of composite technological parameters of a multilayer polypropylene film, comprising the steps of constructing an assumption pool; constructing a reinforcement learning model and a reinforcement learning strategy network, verifying and optimizing assumptions, updating the reinforcement learning model and the reinforcement learning strategy network, determining optimal joint parameters through the reinforcement learning model and the reinforcement learning strategy network, judging whether an assumption pool and a model need to be updated based on multilayer polypropylene film production data, supplementing the assumptions when the assumption needs to be updated, repeating the updating until the multilayer polypropylene film production data reaches production requirements, maintaining the optimal joint parameters for production when the update is not needed, so as to improve the performance stability of the film, realize cooperative optimization and self-adaption of equipment, and reduce the production cost.

Inventors

ZHANG YI
WEN LIANG
DU WENJIE
JIANG LIFEI

Assignees

四川新康意众申新材料有限公司

Dates

Publication Date: 20260505
Application Date: 20260403

Claims (10)

1. The intelligent optimization method for the composite technological parameters of the multilayer polypropylene film is characterized by comprising the following steps of: Step 1, preprocessing multi-source data of a multi-layer polypropylene film compounding process and constructing an assumption pool, wherein the multi-source data comprises compounding equipment data, recovery raw material characteristic data, process-performance detection data and rule data; Step 2, constructing a reinforcement learning model and a reinforcement learning strategy network based on Monte Carlo sampling, wherein the reinforcement learning model comprises a state space, an action space and a reward function; step 3, verifying and optimizing the hypothesis in the hypothesis pool through the reinforcement learning model and the reinforcement learning strategy network, and updating the reinforcement learning model and the reinforcement learning strategy network according to the verification result; Step 4, obtaining optimal joint parameters through the updated reinforcement learning model and reinforcement learning strategy network, wherein the optimal joint parameters comprise optimal parameters of recovery equipment and optimal parameters of a composite process; step 5, adjusting the multi-layer polypropylene film compounding equipment based on the optimal joint parameters to obtain production data of the multi-layer polypropylene film, and judging whether an optimized assumption pool needs to be updated or not based on the production data; And 6, supplementing assumptions into the assumption pool when the assumption pool needs to be updated, and repeating the steps 3-5 until the production data reach the production requirement, and maintaining the optimal joint parameters for production when the assumption pool does not need to be updated.
2. The intelligent optimization method for the composite process parameters of the multilayer polypropylene film according to claim 1, wherein the step 1 comprises the following steps: step 1.1, collecting multidimensional data related to a multilayer polypropylene film compounding process to obtain multisource data; Step 1.2, carrying out standardization processing on multi-source data to obtain standard data; And 1.3, constructing a hypothesis pool based on the rule data.
3. The intelligent optimization method for the composite process parameters of the multilayer polypropylene film according to claim 1, wherein the step 2 comprises the following steps: step 2.1, constructing a reinforcement learning model based on multi-source data; Step 2.2, optimizing the near-end strategy as a reinforcement learning strategy network; and 2.3, processing the reinforcement learning model and the reinforcement learning strategy network through Monte Carlo sampling to obtain probability distribution of combined actions, wherein the combined actions are related to the combined changes of the characteristics of the recovery equipment, the composite equipment and the raw materials.
4. The intelligent optimization method for the composite process parameters of the multilayer polypropylene film according to claim 1, wherein the step 3 comprises the following steps: step 3.1, constructing a joint state transition model based on the transmission mechanism characteristics of the recovery equipment; Step 3.2, predicting the joint state through Monte Carlo sampling to obtain a predicted joint state, wherein the joint state is related to the associated states of the recovery equipment, the compound equipment and the raw material characteristics; and 3.3, verifying and optimizing the hypothesis based on the prediction joint state to obtain an optimized hypothesis pool, and updating the reinforcement learning model and the reinforcement learning strategy network based on the optimized hypothesis pool.
5. The intelligent optimization method for the composite process parameters of the multilayer polypropylene film according to claim 4, wherein the combined state transfer model is as follows: ; Wherein, the Is a joint state transition model; is a transfer function; is a random disturbance term; is a state space; Is an action space.
6. The intelligent optimization method for the composite process parameters of the multilayer polypropylene film according to claim 4, wherein the step 3.3 comprises the following steps: step 3.3.1, verifying the hypothesis based on the prediction joint state to obtain the validity of the hypothesis; Step 3.3.2, when the hypothesis is valid, determining that the hypothesis is valid, and when the hypothesis is invalid, performing self-adaptive optimization on the hypothesis to obtain an optimized hypothesis; step 3.3.3, merging the verification error of the optimization hypothesis into a reward function to obtain a new reward function, and updating parameters of the reinforcement learning strategy network based on the new reward function until the hypothesis is valid; and 3.3.4, when all the hypotheses are valid, obtaining an optimized hypothesis pool, an updated reinforcement learning model and a reinforcement learning strategy network.
7. The intelligent optimization method for the composite process parameters of the multilayer polypropylene film according to claim 1, wherein the step 4 comprises the following steps: step 4.1, determining a prediction joint state through Monte Carlo sampling; Step 4.2, calculating a cost function of each joint action under the prediction joint state to obtain the value of the joint action; and 4.3, obtaining the optimal parameters of the recovery equipment and the optimal parameters of the composite process based on the combined action with the maximum value and the historical optimal reference parameters.
8. The intelligent optimization method for the composite process parameters of the multilayer polypropylene film according to claim 1, wherein the step 6 comprises the following steps: Step 6.1, collecting production data of the multilayer polypropylene film, wherein the production data comprises film performance data and recovery raw material characteristic data; And 6.2, determining a deviation value of the production data and the predicted production data, generating a new hypothesis when the deviation value is greater than or equal to a preset deviation threshold value, and putting the new hypothesis into a hypothesis pool.
9. The intelligent optimization method for the composite process parameters of the multilayer polypropylene film according to claim 1, wherein the state space is as follows: ; Wherein, the Is a state space; Is normalized compression roller abrasion; The heating efficiency of the extruder after standardization; is the winding tension fluctuation after standardization; the temperature stability of the die head after standardization; the rotation speed of the active crushing roller is standardized; the wear amount of the crushing roller is standardized; the particle size of the PP particles after being crushed is standardized; is the particle size uniformity after standardization; The action space is as follows: ; Wherein, the Is an action space; The die temperature adjustment amount; Is the composite pressure adjustment quantity; is the linear velocity adjustment quantity; The coating amount is adjusted; The rotation speed of the crushing roller is adjusted; The rotating speed of the cutting blade is adjusted; the reward function is: ; Wherein, the Is a reward function; is a state space; Is an action space; 、、、、 And First, second, third, fourth, fifth and sixth weight coefficients, respectively; 、 And Respectively an interlayer peeling strength, a minimum value of the interlayer peeling strength and a maximum value of the interlayer peeling strength; And Thickness uniformity and maximum film thickness uniformity, respectively; And The apparent flaw density and the maximum apparent flaw density are respectively; 、 And The particle size uniformity of the recovered raw material, the particle size uniformity of the minimum recovered raw material and the particle size uniformity of the maximum recovered raw material are respectively; Adjusting a cost function for the process; To assume a verification error; The parameters of the reinforcement learning model are updated as follows: ; ; Wherein, the Gradient of the policy objective function with respect to the network parameter θ; The method comprises the steps of selecting a combined track number, wherein the combined track number is a combined track index number, the time step index in a single track, and the prediction period of the single combined track; is a parameter gradient operator; is natural logarithm operation; the joint action probability distribution is output for the strategy network; in the j-th track, the joint action at the moment t+k is performed; the joint state at the moment t+k in the j-th track; for the j-th joint track Is a cumulative discount prize of (2); Is the learning rate.
10. An intelligent optimization system for the composite process parameters of the multilayer polypropylene film is characterized by being used for realizing the intelligent optimization method for the composite process parameters of the multilayer polypropylene film according to any one of claims 1-9, and comprising a hypothesis pool construction module, a reinforcement learning model construction module, a Monte Carlo sampling module, a parameter optimization module, a collaborative verification module and a circulation module; The presumption pool construction module is used for preprocessing multi-source data of the multi-layer polypropylene film compounding process and constructing a presumption pool, wherein the multi-source data comprises compounding equipment data, recovery raw material characteristic data, process-performance detection data and rule data; The reinforcement learning model building module is used for building a reinforcement learning model and a reinforcement learning strategy network based on Monte Carlo sampling, wherein the reinforcement learning model comprises a state space, an action space and a reward function; The Monte Carlo sampling module is used for verifying and optimizing the hypothesis in the hypothesis pool through the reinforcement learning model and the reinforcement learning strategy network, and updating the reinforcement learning model and the reinforcement learning strategy network according to the verification result to obtain a new hypothesis pool, a new reinforcement learning model and a new reinforcement learning strategy network; the parameter optimization module is used for obtaining optimal joint parameters through a new reinforcement learning model and a new reinforcement learning strategy network, wherein the optimal joint parameters comprise optimal parameters of recovery equipment and optimal parameters of a composite process; the collaborative verification module is used for adjusting the multi-layer polypropylene film compounding equipment based on the optimal joint parameters to obtain production data of the multi-layer polypropylene film, and judging whether an assumption pool needs to be updated or not based on the production data; The circulation module is used for supplementing hypothesis into the hypothesis pool when the hypothesis pool needs to be updated, repeating the steps 3-5 until the production data reach the production requirement, and maintaining the optimal joint parameters for production when the hypothesis pool does not need to be updated.

Description

Intelligent optimization method and system for composite process parameters of multilayer polypropylene film Technical Field The invention relates to the technical field of intelligent optimization of equipment parameters, and particularly discloses an intelligent optimization method and system for composite process parameters of a multilayer polypropylene film. Background The multilayer polypropylene film is widely applied to the fields of packaging, electronics, building materials and the like because of the characteristics of light weight, good barrier property, excellent mechanical property and the like, and the rationality of the composite technological parameters directly determines the core properties of the film such as interlayer binding force, thickness uniformity and the like. The existing multi-layer polypropylene film composite process parameter determination method has three main core defects that firstly static experience dependence is not combined with random evolution rules of the full life cycle state of composite equipment, dynamic changes of the running state of the recovery equipment are not considered, process demand changes caused by random characteristics such as equipment abrasion, sensor drift and the like cannot be adapted, secondly a hypothesis-verification system is lacking, a hypothesis generation-verification mechanism is not introduced in the traditional reinforcement learning method, deep association among process parameters, equipment states, film performances and recovery raw material characteristics is difficult to excavate, thirdly the coupling relation between the composite equipment and the recovery equipment is neglected, the cooperative influence of the characteristics such as the particle size, uniformity and the like of the recovery raw materials on the composite process is not considered in independent optimization of the composite equipment, so that film performance fluctuation is large, and the interlayer binding force deviation rate is generally higher. In the prior art, although an optimization method aiming at composite process parameters or an operation regulation method of recovery equipment independently exist, an intelligent process parameter optimization scheme for deeply fusing the synergistic characteristics of a hypothesis generation and verification framework, monte Carlo reinforcement learning and recovery equipment and the composite equipment does not exist, and the requirements of high-precision and dynamic optimization of a multilayer PP film composite process cannot be met. Therefore, there is a need for an intelligent optimization method for technological parameters, which can realize the cooperation of recovery equipment and composite equipment, adapt to the random state of the equipment, and have a hypothetical verification closed loop. In view of the above, the invention provides an intelligent optimization method and system for the composite process parameters of a multilayer polypropylene film, which solve the technical problems existing in the conventional determination method for the composite process parameters of the multilayer polypropylene film, improve the performance stability of the film, realize the cooperative optimization and self-adaption of equipment and reduce the production cost. Disclosure of Invention The invention aims to provide an intelligent optimization method and system for composite process parameters of a multilayer polypropylene film, which solve the problems of automatically realizing dynamic and accurate regulation and control of the composite process parameters of the multilayer polypropylene film, and specifically adopts the following scheme: The intelligent optimization method for the multi-layer polypropylene film compounding process parameters comprises the steps of 1, preprocessing multi-source data of the multi-layer polypropylene film compounding process and constructing an assumption pool, wherein the multi-source data comprise compounding equipment data, recycling raw material characteristic data, process-performance detection data and rule data, 2, constructing a reinforcement learning model and a reinforcement learning strategy network based on Monte Carlo sampling, wherein the reinforcement learning model comprises a state space, an action space and a reward function, 3, verifying and optimizing assumptions in the assumption pool through the reinforcement learning model and the reinforcement learning strategy network, updating the reinforcement learning model and the reinforcement learning strategy network according to verification results, 4, obtaining optimal joint parameters through the updated reinforcement learning model and the reinforcement learning strategy network, and adjusting the multi-layer polypropylene film compounding equipment based on the optimal joint parameters to obtain production data of the multi-layer polypropylene film, judging whether the assumption pool after optimization