CN-121973469-A - Thermoplastic laying bandwidth stabilization method based on nerve guidance sign regression and process three-parameter selector

CN121973469ACN 121973469 ACN121973469 ACN 121973469ACN-121973469-A

Abstract

The invention provides a thermoplastic laying bandwidth stabilization method based on nerve guidance sign regression and a process three-parameter selector, which synchronously collects power, speed, pressure, strip temperature, effective bandwidth and material identification and calculates compaction conditions; the method comprises the steps of constructing a process influence item set by utilizing an input alignment, sectional gating and symbolic regression structure, explicitly calculating an edge outflow risk value, identifying a safety adjustment interval and a maximum adjustable amplitude of power, speed and pressure, constructing a single parameter adjustment path in a process three parameter selector, comprehensively estimating the risk descent quantity, the compaction influence quantity and response time delay, generating a first adjustment parameter instruction, acquiring response data through single step adjustment and adjustment, judging bandwidth stability back according to risk change, effective bandwidth difference, response time delay and compaction conditions, and outputting a sequence control instruction. The invention can reduce the outflow risk of the edges of the strip, ensure the compaction quality and improve the stability and the forming quality of the thermoplastic laying process.

Inventors

CHEN YIWEI
FAN CONGZE
ZHU SHU
SONG WENZHE
ZHENG JINGHUA
ZHOU JIANFENG

Assignees

南京航空航天大学
东华大学

Dates

Publication Date: 20260505
Application Date: 20260114

Claims (9)

1. A thermoplastic deposition bandwidth stabilization method based on neural guided symbology regression and a process three parameter selector, comprising: s1, acquiring power, speed, pressure, strip temperature, effective bandwidth and material identification, generating working condition state data and calculating compaction conditions; S2, according to working condition state data and material identification, a process influence item set is constructed, an edge outflow risk value is calculated in nerve guidance sign regression, influence item weight and monotonic direction are given out by a neural network, under the sectional constraint of vitrification to a melting temperature range, a safe adjustment interval and a maximum adjustable amplitude of power, speed and pressure are identified, and an edge outflow risk value, the safe adjustment interval and the maximum adjustable amplitude are obtained; s3, under the compaction condition, forming a power adjustment path, a speed adjustment path and a pressure adjustment path according to the edge outflow risk value, the safety adjustment interval and the maximum adjustable amplitude, and measuring the risk reduction of each path in the allowable amplitude, the impact quantity on compaction and the response time delay, and determining an initial adjustment parameter instruction comprising parameters to be adjusted, an adjustment direction and the adjustment amplitude; S4, performing single-step adjustment according to the first-adjustment parameter instruction and following the safety adjustment interval and the maximum adjustable amplitude to generate adjustment response data; S5, calculating a bandwidth stability judging result according to the adjustment response data, the edge outflow risk value and the compaction condition, and forming the bandwidth stability judging result when the set threshold and the compaction condition are met; s6, forming a sequence control instruction according to the first adjustment parameter instruction and the bandwidth stability judging result.
2. The method for stabilizing the thermoplastic laying bandwidth based on the neural guided symbol regression and the process three-parameter selector according to claim 1, wherein S1 is specifically: collecting power, speed, pressure, strip temperature, effective bandwidth and material identification at the same time point to form a numerical sequence; arranging the original signal sets according to the sequence of power, speed, pressure, strip temperature and effective bandwidth to generate working condition state data containing five values, converting material identification into a numerical vector with ten dimensions, and aligning the numerical vector with the working condition state data in an input layer; judging the effective bandwidth by adopting a set threshold according to the power, the speed, the pressure and the strip temperature in the working condition state data, and calculating compaction conditions, namely the upper and lower boundaries of the range of the power, the speed and the pressure and the upper and lower boundaries of the strip temperature in the range from vitrification to melting; And (3) carrying out process influence item set calculation on the working condition state data and the material identification for nerve guidance sign regression, extracting the numerical value and the change trend of the strip temperature, aligning the numerical value and the change trend with the effective bandwidth, and keeping the compaction condition consistent with the physical constraint of edge outflow risk value identification.
3. The method for stabilizing the thermoplastic laying bandwidth based on the neural guided symbol regression and the process three-parameter selector according to claim 1, wherein the step S2 is specifically: aligning the working condition state data and the material mark in an input alignment layer to form fifteen-dimensional input vectors, and mapping the fifteen-dimensional input vectors to vector positions according to the sequence of power, speed, pressure, strip temperature and effective bandwidth to serve as input of an influence item generation layer; Calculating a process influence item set at an influence item generation layer for fifteen-dimensional input vectors, generating item by item according to a combination item of a power related item, a speed related item, a pressure related item, a temperature related item, a bandwidth related item and a material identification related item, and keeping alignment with working condition state data and a material identification; calculating a sectional gating signal vitrified to a melting temperature range in a sectional gating layer according to the material identification and the strip temperature to obtain gating coefficients for a low-temperature section and a high-temperature section, and performing sectional starting and restraining on a process influence item set; inputting the output of the influence item generation layer and the gating coefficient of the segmented gating layer into a symbol coefficient layer, generating influence item weight and monotonic direction, wherein the monotonic direction corresponds to power, speed and pressure respectively and is used for limiting the parameter change direction; Assembling explicit expression in a symbolic regression head by using a process influence item set and influence item weight, calculating an edge outflow risk value under the constraint of a gating coefficient of a segmented gating layer, and keeping alignment with working condition state data; the power, the speed and the pressure are scanned unidirectionally in a safe adjustment interval computing head according to the monotonic direction and the gating coefficient of the segmented gating layer, and the upper and lower boundaries of three sections of intervals are output to form a safe adjustment interval; And calculating the maximum adjustable amplitude of the head generation power, speed and pressure according to the variation trend of the strip temperature and the effective bandwidth and combining the length of the safety adjustment interval at the maximum adjustable amplitude, and obtaining an edge outflow risk value, the safety adjustment interval and the maximum adjustable amplitude.
4. A thermoplastic deposition bandwidth stabilization method based on neural guided symbolic regression and process three parameter selector according to claim 3, characterized in that the step of assembling explicit expression and calculating edge outflow risk value at symbolic regression head with process influence term set and influence term weight, wherein the edge outflow risk value is calculated using a risk formula: ; Wherein, the The edge outflow risk value, the single value aligned with the operating condition, To affect the item index, take one to twenty-four, To influence item weight Numerical value and The alignment of the two parts is performed, Set of process influence items Numerical value of The generation of the product is carried out, Control coefficient at the first for low temperature Duan Men The number on the item is affected by the number, Is of high Wen Duanmen control coefficient at The number on the item is affected by the number, For the segment coefficients, the values are between zero and one, and the temperature of the strip material is used for the segment coefficients Relative to the glass transition temperature threshold With a melting temperature threshold Generated by linear interpolation, and truncated to maintain the zero to one range when the threshold interval is exceeded, For the glass transition temperature threshold, for the material identification of the corresponding value, For the melting temperature threshold, corresponding values are identified for the material, the variables are obtained directly in the form of values in the model and are associated with And (3) with Maintaining alignment.
5. The method for stabilizing the thermoplastic laying bandwidth based on the neural guided symbol regression and the process three-parameter selector according to claim 1, wherein the step S3 is specifically: Inputting compaction conditions, edge outflow risk values, a safety adjustment interval, maximum adjustable amplitude and working condition state data into a process three-parameter selector, and determining that a path starting point is power, speed and pressure in the working condition state data; Sequentially constructing a power adjustment path, a speed adjustment path and a pressure adjustment path in a monotonic direction in a safety adjustment interval by taking the maximum adjustable amplitude as a step length, wherein each path takes ten candidate points to form path data; Calculating risk reduction for each path according to the working condition state data and the edge outflow risk value, and taking the difference value of the edge outflow risk values of the path starting point and the path ending point as the risk reduction; calculating the influence quantity of each path on compaction according to the compaction condition and the working condition state data of the path end point, judging the offset by adopting a set threshold value, and generating an influence quantity value; the time quantity of the primary parameter adjustment conducted to the effective bandwidth is used as response time delay, and the response time delay is calculated for each path according to the change relation between the working condition state data and the effective bandwidth; aligning the risk descent quantity, the impact quantity on compaction and the response time delay into a path calculation vector, wherein the dimension is three, and fusing the path calculation vector with the upper and lower boundaries of the safety adjustment interval and the maximum adjustable amplitude to form a path constraint vector, wherein the dimension is nine, and calculating a path scoring value; And screening the path with the maximum grading value according to the path grading value, outputting parameters to be regulated, a regulating direction and a regulating amplitude to form a first regulating parameter instruction, performing constraint checking according to a safe regulating interval and the maximum adjustable amplitude, and performing amplitude cutting and direction correction on the instruction which does not meet the constraint to obtain the first regulating parameter instruction with compliance.
6. The method for stabilizing thermoplastic deposition bandwidth based on neural guided symbol regression and process three parameter selector according to claim 5, wherein the step of aligning the risk reduction, the impact on compaction and the response time delay into a path calculation vector, wherein the dimension is three, and the path calculation vector is formed by fusing the upper and lower boundaries of the safety adjustment interval with the maximum adjustable amplitude, wherein the dimension is nine, and wherein in the step of calculating the path scoring value, the path scoring value is calculated by adopting a path scoring formula: ; Wherein, the Scoring a value for a path, indexing a value for a path The individual values of the alignment are chosen to be, Taking a value from zero to one as a constraint coefficient, The risk reduction weight is dimensionless number, The weight of the impact on compaction is a dimensionless value, In response to the delay weight, the unit is the inverse of time, The risk reduction is the difference between the edge outflow risk values of the start point and the end point of the path, To influence compaction, three offsets and The maximum ratio of (2) is a dimensionless number, For response time delay, for first exceeding effective bandwidth For the difference between the detection time and the parameter adjustment time The calculation of (2) is performed according to the following flow, and the parameters of the current path are recorded as Will be when the end point of the path crosses the boundary Setting zero, when the boundary is not crossed, calculating the residual distance from the end point to the nearest boundary divided by the corresponding interval length to obtain the interval residual ratio, and recording as Calculating the complement value of the ratio of the accumulated step length of the end point to the starting point and the maximum adjustable amplitude of the parameter to obtain the residual step length ratio, and recording the residual step length ratio as Will be And (3) with Cut off to a range of zero to one, and take the smaller value of the two as 。
7. The method for stabilizing the thermoplastic laying bandwidth based on the neural guided symbol regression and the process three-parameter selector according to claim 1, wherein the step S4 is specifically: Aligning the first adjustment parameter instruction with the safety adjustment interval and the maximum adjustable amplitude, judging compliance according to the upper and lower boundaries and the step upper limit, and cutting off according to the upper and lower boundaries and correcting according to the monotonic direction when the compliance is not met; adopting single-step adjustment under compaction conditions, and executing one-time change of parameters to be adjusted according to the corrected adjustment amplitude, so as to keep two parameters which are not selected unchanged; Immediately acquiring power, speed, pressure, strip temperature and effective bandwidth after adjustment to form adjusted working condition state data, and establishing a front-back corresponding relation with the working condition state data in alignment; And calculating the difference value of the effective bandwidth and the response time delay according to the front-back correspondence, combining the corrected adjustment amplitude, the difference value of the effective bandwidth and the response time delay, generating adjustment response data, and aligning the adjustment response data with the edge outflow risk value.
8. The method for stabilizing the thermoplastic laying bandwidth based on the neural guided symbol regression and the process three-parameter selector according to claim 1, wherein the step S5 is specifically: Aligning the adjustment response data, the edge outflow risk value and the compaction condition, organizing the adjustment response data, the edge outflow risk value and the compaction condition into input for judgment according to the time sequence of the adjusted working condition state data, and keeping alignment with the working condition state data; checking the corrected adjustment amplitude in the adjustment response data according to the safety adjustment interval and the maximum adjustable amplitude, and entering calculation only when the upper and lower boundaries are not exceeded and the step upper limit is not exceeded; Calculating risk reduction according to the difference value of the effective bandwidth in the adjustment response data and the response time delay and combining the change of the edge outflow risk value, comparing the difference value of the effective bandwidth with a set threshold value, and simultaneously comparing the response time delay with the set threshold value; checking whether the power, the speed and the pressure in the adjusted working condition state data are within the upper and lower boundaries of the range of the compaction condition according to the compaction condition, and judging that the working condition is not stable when the working condition is not within the range; and forming a bandwidth stability judging result when the risk reduction satisfies a set threshold value, the difference value of the effective bandwidth satisfies the set threshold value and the compaction condition is satisfied, and not forming the bandwidth stability judging result when any condition is not satisfied.
9. The method for stabilizing the thermoplastic laying bandwidth based on the neural guided symbol regression and the process three-parameter selector according to claim 1, wherein the step S6 is specifically: Aligning the first-tuning parameter instruction with a bandwidth stability judging result, and carrying out compliance verification according to the upper and lower boundaries of the safety adjustment interval and the step upper limit of the maximum adjustable amplitude to obtain an executable instruction set; When the bandwidth stabilizing judgment result meets the set threshold and compaction conditions, setting the sequence control instruction to stop adjusting, and keeping the power, speed and pressure unchanged; When the bandwidth stable-returning judging result is not formed, setting the sequence control instruction to be along the monotone direction of the first adjustment parameter instruction, and advancing to the next candidate point in the safety adjustment interval by a step length not exceeding the maximum adjustable amplitude; and (3) sectionally limiting the starting temperature range of the sequence control instruction according to the gating coefficient of the sectionally gating layer, so that the sequence control is aligned with the sectionally constraint of vitrification to the melting temperature range.

Description

Thermoplastic laying bandwidth stabilization method based on nerve guidance sign regression and process three-parameter selector Technical Field The invention relates to the technical field of automatic thermoplastic composite material laying process control, in particular to a thermoplastic laying bandwidth stabilization method based on nerve guidance sign regression and a process three-parameter selector. Background The thermoplastic composite material is widely focused in the field of automatic laying and compression molding due to the characteristics of meltability, short molding cycle and the like. In order to ensure the dimensional accuracy and the interface quality of the components, a great deal of researches are conducted around the prediction regulation and control of the laying deformation of the plastic composite material and the prediction of the molding performance of the thermoplastic composite material, and the laying deformation, the residual stress and the molding performance are analyzed and optimized by establishing a material constitutive model, a laying path geometric model and a molding process warm-pressing field model. However, such research has focused on offline simulation and process window design, with relatively inadequate focus on online bandwidth stability and edge outflow risk regulation in automated placement processes. In the existing thermoplastic laying process control, the common practice is to determine process windows such as power, speed, pressure and the like according to experience or experiment, manually or simply carry out closed loop adjustment on the basis of monitoring the temperature and forming quality of the strip, and also to collect data of the power, speed, pressure, temperature and laying quality by utilizing a sensor, construct regression or neural network models to predict laying bandwidth, temperature field or forming defects for guiding parameter setting, wherein part of schemes try to combine on-line measurement results, carry out amplitude limiting adjustment on single parameters or carry out linkage adjustment on the basis of preset rules so as to reduce the occurrence probability of the defects. The methods improve the stability of the thermoplastic laying process to a certain extent, but are mostly based on a fixed process window or a black box prediction model, and lack a systematic control strategy which is tightly coupled with the material temperature domain characteristics and takes the bandwidth and the compaction quality into account. In the prior art, the bandwidth control and process parameter adjustment usually do not explicitly introduce the sectional temperature range constraint of materials from vitrification to melting, so that the influence of temperature change on the laying bandwidth and edge outflow behavior is difficult to accurately reflect, the parameter suggestion based on experience or a black box model stays at a given recommended value or is subjected to single correction, the quantitative evaluation and closed loop verification of the bandwidth change and edge outflow risk in the process of linkage adjustment of power, speed and pressure in a safety range are lacking, and meanwhile, the interpretable risk index, the clear safety adjustment interval and the maximum adjustable amplitude description are generally lacking, so that the automatic generation of a sequence control instruction and a single step adjustment path are difficult under the premise of ensuring compaction conditions, and the online bandwidth stabilization requirement under complex working conditions is difficult to meet. Accordingly, a thermoplastic deposition bandwidth stabilization method that addresses the above-described deficiencies of the prior art is a problem that one of ordinary skill in the art would be able to address. Disclosure of Invention The invention aims to provide a thermoplastic laying bandwidth stabilization method based on nerve guidance sign regression and a process three-parameter selector, which aims to solve the core technical problems of how to construct a closed-loop control method with an interpretable edge outflow risk assessment and a power, speed and pressure three-parameter safety adjustment mechanism under the combined action of a glass transition to melt temperature region segmentation constraint and a compaction condition in the automatic laying process of thermoplastic composite materials, so as to realize on-line stabilization and sequential self-adaptive adjustment of the effective laying bandwidth. According to the embodiment of the invention, the thermoplastic laying bandwidth stabilization method based on the neural guided symbol regression and the process three-parameter selector comprises the following steps of: s1, acquiring power, speed, pressure, strip temperature, effective bandwidth and material identification, generating working condition state data and calculating compaction conditions; S2, according to