Search

CN-121978924-A - Fuzzy neural network-based force control system with jump head grinder Mao Jizhang

CN121978924ACN 121978924 ACN121978924 ACN 121978924ACN-121978924-A

Abstract

The invention discloses a tension control system of a head-jumping roughening machine based on a fuzzy neural network, which comprises a data acquisition module, a dual-mode judging mechanism construction module, a coarse tuning control module, a fine tuning control module, a fusion control module and a driving module, wherein the data acquisition module generates a tension control input data set, the dual-mode judging mechanism construction module sets a dual-mode judging threshold value based on the tension control input data set, constructs a dual-mode judging mechanism and outputs a dual-mode judging result, the coarse tuning control module caches a coarse tuning control instruction in a continuous fusion buffer area, the fine tuning control module caches a fine tuning control instruction in the continuous fusion buffer area, the fusion control module carries out continuous fusion processing on the coarse tuning control instruction and the fine tuning control instruction in the continuous fusion buffer area to obtain a fusion control output signal, and the driving module converts the fusion control output signal into a driving instruction and sends the driving instruction to a tension adjustment executing mechanism to enable the tension adjustment executing mechanism to carry out real-time adjustment on fabric tension. The invention ensures the continuity and the accuracy of tension regulation and control of the roughening machine with the jump head under different working conditions.

Inventors

  • ZHANG ZHENGYOU

Assignees

  • 吴江市海威机械橡胶制品有限公司

Dates

Publication Date
20260505
Application Date
20260114

Claims (8)

  1. 1. The utility model provides a take jump mill Mao Jizhang force control system based on fuzzy neural network which characterized in that includes: The data acquisition module acquires and preprocesses tension control signals of the head-jumping roughening machine to generate a tension control input data set; The double-mode judging mechanism constructing module is used for setting a double-mode judging threshold value based on the tension control input data set, constructing a double-mode judging mechanism, judging whether the system is in a disturbance stage or a steady state stage currently by utilizing the fabric tension real-time signal and the jump mechanism state real-time signal, and outputting a double-mode judging result; The coarse control module activates a coarse control mode when the dual-mode judging result shows that the system is in a disturbance stage, generates a coarse control instruction according to a tension control input data set, and caches the coarse control instruction in the continuous fusion buffer area; The fine tuning control module activates a fine tuning control mode when the dual-mode discrimination result shows that the system is in a steady state stage, calls a fuzzy neural network tension control model, calculates a fine tuning control instruction according to a tension control input data set, and caches the fine tuning control instruction in a continuous fusion buffer area; The fusion control module is used for carrying out continuous fusion processing on the coarse control instruction and the fine control instruction in the continuous fusion buffer zone, and dynamically calculating fusion weights according to the activation state of the coarse control mode and the activation state of the fine control mode to obtain a fusion control output signal; And the driving module converts the fusion control output signal into a driving instruction and sends the driving instruction to the tension adjusting executing mechanism, so that the tension adjusting executing mechanism can adjust the fabric tension in real time.
  2. 2. The fuzzy neural network-based tension control system with a jump head grinder Mao Jizhang for preprocessing is characterized by comprising the steps of constructing a multi-source signal acquisition system, acquiring a fabric tension real-time signal, a main driving motor rotating speed real-time signal, a grinding roller rotating speed real-time signal and a jump head mechanism state real-time signal in real time, synchronizing and timing the real-time signals, and carrying out noise filtering and amplitude normalization processing on the synchronized and timed fabric tension real-time signal, the main driving motor rotating speed real-time signal, the grinding roller rotating speed real-time signal and the jump head mechanism state real-time signal to generate a tension control input data set.
  3. 3. The fuzzy neural network-based force control system with a jump head grinder Mao Jizhang according to claim 1, wherein the dual-mode discrimination mechanism building module comprises: Determining a fabric tension real-time signal, a target tension set value and a jump head mechanism state real-time signal based on the tension control input data set; calculating a tension error signal by using the difference between the real-time fabric tension signal and the target tension set value; subtracting the tension error signal of the last sampling time from the tension error signal of the current time, dividing the tension error signal by the sampling period, and calculating a tension error change rate signal; Subtracting the tension error change rate signal at the last sampling time from the tension error change rate signal at the current time, dividing the tension error change rate signal by the sampling period, and calculating a tension error second-order differential signal; subtracting the real-time signal of the state of the head-jumping mechanism at the last sampling moment from the real-time signal of the state of the head-jumping mechanism at the current moment, dividing the real-time signal by the sampling period, and calculating a head-jumping action edge signal; Constructing a jump disturbance window based on the tension control input data set, and calculating a jump disturbance index; performing self-adaptive baseline estimation on the jump disturbance index based on the tension control input data set, and constructing a disturbance self-adaptive baseline and a disturbance self-adaptive scale; Based on the disturbance self-adaptive base line and the threshold scale coefficient multiplied by the sum of the disturbance self-adaptive scale, respectively calculating a disturbance entering hysteresis threshold and a disturbance exiting hysteresis threshold, and introducing a jump feedforward trigger item; based on the disturbance entering hysteresis threshold and the disturbance exiting hysteresis threshold, a dual-mode discrimination logic with hysteresis and minimum residence time constraint is constructed, and a dual-mode discrimination result is output.
  4. 4. The fuzzy neural network based force control system of the jump head grinder Mao Jizhang, wherein the dual mode discrimination logic comprises: When the current time jump feedforward triggering item is 1 or the jump disturbance index is larger than a disturbance entering hysteresis threshold value in the dual-mode discrimination logic, judging as a disturbance stage discrimination result; when the current time jump feedforward triggering item is 0, the jump disturbance index is smaller than or equal to a disturbance exit hysteresis threshold value in the dual-mode discrimination logic, and the time from the last discrimination switching moment is larger than or equal to the minimum residence time, the discrimination result is judged as a steady-state stage discrimination result.
  5. 5. The fuzzy neural network based force control system of the jump head mill Mao Jizhang, wherein the coarse control module comprises: when the dual-mode judging result is a disturbance stage judging result, determining the current control moment, constructing a jump head phase variable, and updating the jump head phase variable in a coarse control mode to obtain an updated jump head phase variable; Based on the mechanical characteristics of a head-jumping mechanism of the roughening machine with the head jumping, indexing a corresponding preset compensation moment value by utilizing a current head-jumping phase variable, and multiplying the preset compensation moment value by an indication of whether a head-jumping action edge signal is greater than zero or whether a head-jumping mechanism state real-time signal is 1 to obtain a feedforward moment compensation quantity; in the coarse control mode, obtaining damping compensation moment based on weighted summation of the current tension error signal and the tension error change rate signal; Adding the feedforward torque compensation quantity and the damping compensation torque to obtain a coarse non-limiting torque command; and (3) performing improved anti-saturation limiting processing on the coarse non-limiting moment instruction to obtain a coarse control instruction, writing the coarse control instruction into a continuous fusion buffer zone, and marking the valid bit of the coarse control mode.
  6. 6. The fuzzy neural network based force control system of the jump head grinder Mao Jizhang, wherein the improved anti-saturation clipping process comprises: Comparing the coarse non-limiting moment instruction with a maximum moment instruction and a minimum moment instruction which are allowed to be output by the tension adjusting executing mechanism in intervals, and outputting the maximum moment instruction as a coarse control instruction when the coarse non-limiting moment instruction is larger than the maximum moment instruction; when the coarse non-limiting moment command is smaller than the minimum moment command, outputting the minimum moment command as a coarse control command; and when the coarse non-limiting moment command is positioned between the maximum moment command and the minimum moment command, the coarse non-limiting moment command is directly output as a coarse control command.
  7. 7. The fuzzy neural network based force control system of the jump head grinder Mao Jizhang, wherein the fine tuning control module comprises: when the dual-mode judging result is a steady-state stage judging result, activating a fine tuning control mode, calling a fuzzy neural network tension control model, and combining a tension error signal and a tension error change rate signal at the current moment into a network input vector of the fine tuning control mode; in a fuzzy neural network tension control model, respectively carrying out fuzzification mapping on network input vectors to respectively obtain membership components of tension error signals on each error fuzzy set and membership components of tension error change rate signals on each change rate fuzzy set; Based on the product of the membership component of the tension error signal on each error fuzzy set and the membership component of the tension error change rate signal on each change rate fuzzy set, constructing a corresponding rule activation intensity matrix, and carrying out normalization processing to obtain a normalization rule activation intensity matrix; configuring a unique corresponding linear output for each fuzzy rule, and carrying out weighted summation on the linear outputs according to an activation intensity matrix of a normalization rule to obtain a fine-tuning non-limiting moment instruction; comparing the fine-tuning non-limiting moment instruction with the maximum moment instruction and the minimum moment instruction in intervals, and adopting the maximum moment instruction as output when the fine-tuning non-limiting moment instruction exceeds the maximum moment instruction; When the fine-tuning non-limiting moment command is lower than the minimum moment command, the minimum moment command is adopted as output; When the fine-tuning non-limiting moment instruction is positioned between the maximum moment instruction and the minimum moment instruction, directly adopting the fine-tuning non-limiting moment instruction as a fine-tuning control instruction output to obtain a fine-tuning control instruction; And caching the fine tuning control instruction in the continuous fusion buffer area, and marking the valid bit of the fine tuning control mode.
  8. 8. The fuzzy neural network based force control system of the jump head grinder Mao Jizhang, wherein the fusion control module comprises: reading a coarse control command, a fine control command, a coarse control mode valid bit and a fine control mode valid bit from the continuous fusion buffer zone at the current control moment; based on the coarse adjustment control mode valid bit and the fine adjustment control mode valid bit, constructing and maintaining a fusion weight state quantity in the continuous fusion buffer zone; Continuously updating the fusion weight state quantity based on the sampling period and a preset smoothing time constant to obtain an updated fusion weight state quantity; based on the updated fusion weight state quantity, carrying out continuous fusion processing on the coarse control instruction and the fine control instruction to obtain a fusion control output signal, writing back to a continuous fusion buffer zone, and marking a fusion output valid bit.

Description

Fuzzy neural network-based force control system with jump head grinder Mao Jizhang Technical Field The invention relates to the technical field of a head-jumping sanding machine, in particular to a force control system of a head-jumping sanding Mao Jizhang based on a fuzzy neural network. Background Along with the continuous improvement of the intelligent level of the textile industry, the sanding machine with the jump head is used as high-end fabric after-finishing equipment, and plays an important role in the aspects of fine processing and improving the surface quality of fabrics. The sanding machine with the jump head realizes the periodic cutting-in and cutting-out of the sanding roller to the surface of the fabric through a special jump head mechanism so as to obtain unique appearance and hand feeling of the fabric. In the prior art, a tension control system of a roughening machine usually adopts a feedback regulation loop based on a tension sensor, a proportional-integral-derivative control method with fixed parameters is generally adopted or a control parameter is set based on static process experience, the conventional control strategy generally regards the roughening machine with a jump head as a linear and continuous power system, periodic abrupt disturbance caused by a mechanical structure in the jump head action process is ignored, and when the jump head mechanism performs cutting-in or cutting-out operation, the fabric tension is subjected to obvious step load impact, so that response lag, tension overshoot and even oscillation of the system are easily caused. Aiming at the abrasion of different fabrics or machine parts and the change working condition of friction force, the prior art relies on manual repeated debugging and experience correction parameters, and is difficult to adapt to the requirements of high flexibility and high stability of the production line. The traditional dual-mode distinguishing and switching control method takes a tension error threshold value as a mode switching basis, so that the advanced distinguishing and effective isolation of a strong disturbance process caused by the head jumping action are difficult to realize, the problems of distinguishing hysteresis, frequent mode shaking or unstable switching are easy to occur, and the fabric tension is difficult to keep uniform and stable under the periodic disturbance of the head jumping. The conventional PID control or simple threshold switching mechanism can not actively identify and compensate tension energy transition caused by mechanical impact under the nonlinear and abrupt load working conditions of cutting-in and cutting-out of the jump head mechanism, so that the processing quality and the production efficiency of the high-end fabric with the jump head sanding machine are limited. Disclosure of Invention The invention aims to provide a fuzzy neural network-based force control system for a head-jumping grinder Mao Jizhang, which ensures the continuity and accuracy of tension regulation and control of the head-jumping grinder under different working conditions. According to the embodiment of the invention, the tension control system of the fuzzy neural network-based head-jumping roughening machine comprises the following components: The data acquisition module acquires and preprocesses tension control signals of the head-jumping roughening machine to generate a tension control input data set; The double-mode judging mechanism constructing module is used for setting a double-mode judging threshold value based on the tension control input data set, constructing a double-mode judging mechanism, judging whether the system is in a disturbance stage or a steady state stage currently by utilizing the fabric tension real-time signal and the jump mechanism state real-time signal, and outputting a double-mode judging result; The coarse control module activates a coarse control mode when the dual-mode judging result shows that the system is in a disturbance stage, generates a coarse control instruction according to a tension control input data set, and caches the coarse control instruction in the continuous fusion buffer area; The fine tuning control module activates a fine tuning control mode when the dual-mode discrimination result shows that the system is in a steady state stage, calls a fuzzy neural network tension control model, calculates a fine tuning control instruction according to a tension control input data set, and caches the fine tuning control instruction in a continuous fusion buffer area; The fusion control module is used for carrying out continuous fusion processing on the coarse control instruction and the fine control instruction in the continuous fusion buffer zone, and dynamically calculating fusion weights according to the activation state of the coarse control mode and the activation state of the fine control mode to obtain a fusion control output signal; And the driving module converts the fusion con