Spindle blade motion state control method and system for wrap yarn spinning

Abstract

The invention relates to the technical field of motion state control, in particular to a spindle blade motion state control method and system for wrap yarn spinning. Yarn core tension, motor operation data and environment data are collected to construct a wrap yarn spinning state vector, the importance weight of each element in the wrap yarn spinning state vector to tension change is calculated, and the wrap yarn spinning state vector is weighted according to the importance weight to obtain a weighted state vector; on the basis of the weighted state vector, the predicted yarn core tension at the next sampling moment is obtained through a tension prediction model; calculating a tension error, predicting the tension error and a tension error change rate, inputting the tension error and the tension error change rate into a self-adaptive controller to obtain a motor current control quantity, calculating a motor voltage control quantity according to the motor current control quantity, controlling a spindle blade according to the motor voltage control quantity, and adjusting parameters of the self-adaptive controller in real time through an online parameter optimization method. According to the invention, the spindle blade control precision of wrap yarn spinning can be improved.

Description

Technical Field

[0001] The invention relates to the technical field of motion state control, and in particular to a method and system for controlling the motion state of a spindle used for covered yarn weaving. Background Art

[0002] As an important part of the economy, the textile industry plays a key role in many fields such as clothing, home textiles and industrial fabrics. Among them, the yarn coating process is the core link in textile production, which is mainly used to enhance the physical properties of the yarn, such as improving strength, improving hand feel and increasing wear resistance. The quality of yarn coating directly affects the performance and market competitiveness of the final product. Therefore, improving the uniformity and consistency of yarn coating and optimizing production efficiency have become important goals of textile machinery research and development.

[0003] In the process of yarn coating, the spindle bar is a key mechanical component. The control of its motion state not only affects the appearance and feel of the coated yarn, but also may affect the overall strength and durability of the yarn, and thus affect the quality and service life of the final product. The traditional spindle bar motion control method mainly relies on the design of mechanical structure and simple control algorithm, and the yarn coating is achieved by adjusting the speed and motion trajectory of the spindle bar. However, the traditional control method lacks adaptive and optimization capabilities, which limits the control accuracy of the control method in complex and changeable production environments.

[0004] To this end, a method and system for controlling the motion state of a spindle for covered yarn spinning are proposed. Summary of the invention

[0005] The object of the present invention is to provide a spindle motion state control method and system for covered yarn weaving. The covered yarn weaving state vector is constructed by collecting yarn core tension, motor operation data and environmental data, and the importance weight of each element in the covered yarn weaving state vector to the tension change is calculated, and the covered yarn weaving state vector is weighted according to the importance weight to obtain a weighted state vector; based on the weighted state vector, the predicted yarn core tension at the next sampling moment is obtained through a tension prediction model; the tension error, the predicted tension error and the tension error change rate are calculated, and input into an adaptive controller to obtain a motor current control quantity, the motor voltage control quantity is calculated according to the motor current control quantity, and the spindle is controlled according to the motor voltage control quantity, and the adaptive controller parameters are adjusted in real time through an online parameter optimization method.

[0006] To achieve the above object, the present invention provides the following technical solutions: A method for controlling the motion state of a spindle bar for covered yarn spinning, comprising: In the process of weaving the covered yarn, the yarn core tension, the motor operation data and the environmental data are collected according to the sampling period; the motor operation data includes the motor torque, the motor speed, the motor angular velocity and the motor current; the environmental data includes the environmental temperature and the environmental humidity; Constructing a covered yarn spinning state vector based on the yarn core tension, the motor operation data and the environmental data; Calculating the importance weight of each element in the covered yarn spinning state vector to the tension change to obtain an importance weight vector, and performing weighted processing on the covered yarn spinning state vector according to the importance weight vector to obtain a weighted state vector; Based on the weighted state vector, a predicted yarn core tension at the next sampling moment is obtained through a tension prediction model; Setting a target yarn core tension, calculating a tension error at a current sampling moment according to the target yarn core tension and the yarn core tension at a current sampling moment, calculating a predicted tension error according to the predicted yarn core tension and the target yarn core tension, and calculating a tension error change rate according to the tension error at a current sampling moment and a previous sampling moment; Based on the tension error, the predicted tension error and the tension error change rate, the motor current control quantity is obtained through an adaptive controller, the motor voltage control quantity is calculated according to the motor current control quantity, the spindle rod is controlled according to the motor voltage control quantity, and the adaptive controller parameters are adjusted in real time through an online parameter optimization method.

[0007] Preferably, the process of weighting the covered yarn spinning state vector comprises: The importance weight of each element in the covered yarn spinning state vector to the tension change is calculated using the following formula: ; in, For the The importance weight of the elements, is the natural exponential function, For the The basic importance coefficient of each element, is the number of elements in the covered yarn spinning state vector, For the Correlation coefficient between each element and tension; Will The importance weights constitute the importance weight vector, and the weighted state vector is obtained by weighting the covered yarn spinning state vector according to the importance weight vector, and the formula is: ; in, The current sampling time The weighted state vector, is the importance weight vector, The current sampling time The spinning state vector of the covered yarn, is the Hadamard product symbol.

[0008] Preferably, the process of the tension prediction model includes: Extract recent The weighted state vectors at the sampling moments and constructing a weighted state vector sequence; The weighted state vector sequence is input into a gated recurrent unit neural network to predict the predicted yarn core tension at the next sampling moment, including: calculating a forgetting coefficient and a retention coefficient based on the weighted state vector at the current sampling moment and the hidden state at the previous sampling moment; using the forgetting coefficient to selectively forget the hidden state at the previous sampling moment, and fusing it with the weighted state vector at the current sampling moment to obtain a candidate hidden state at the current sampling moment; using the retention coefficient to fuse the hidden state at the previous sampling moment with the candidate hidden state at the current sampling moment to obtain the hidden state at the current sampling moment; and obtaining the predicted yarn core tension at the next sampling moment through a fully connected layer based on the hidden state at the current sampling moment.

[0009] Preferably, the adaptive controller comprises: The tension error, the predicted tension error and the tension error change rate are set respectively fuzzy sets; Constructing an adaptive Gaussian membership function based on each fuzzy set, and initializing membership function parameters of each adaptive Gaussian membership function; the membership function parameters include a function center value and a function standard deviation; The membership of the tension error, the predicted tension error and the tension error change rate to the corresponding fuzzy set is calculated respectively by the adaptive Gaussian membership function; An initial fuzzy rule base is constructed based on fuzzy sets to initialize the rule parameters of each fuzzy rule; the initial fuzzy rule base includes fuzzy rules; the rule parameters include a first rule parameter, a second rule parameter and a third rule parameter; the first rule parameter is related to the tension error, the second rule parameter is related to the predicted tension error, and the third rule parameter is related to the tension error change rate; Calculate the motor current adjustment amount under each fuzzy rule according to the rule parameters combined with the tension error, the predicted tension error and the tension error change rate; The trigger strength of each fuzzy rule is calculated based on the membership degree, and the weighted motor current adjustment amount is calculated by weighted average method according to the trigger strength and the motor current adjustment amount, and the motor current control amount is obtained by the weighted motor current adjustment amount.

[0010] Preferably, adjusting the adaptive controller parameters in real time through an online parameter optimization method includes: constructing a comprehensive performance index, wherein the comprehensive performance index is composed of a control accuracy index, a control smoothness index, and a tension stability index; based on the comprehensive performance index, optimizing the rule parameters and the membership function parameters using a gradient descent method.

[0011] A spindle motion state control system for covered yarn spinning, comprising: A data acquisition unit, during the weaving process of the covered yarn, collects yarn core tension, motor operation data and environmental data according to a sampling period; the motor operation data includes motor torque, motor speed and motor current; the environmental data includes ambient temperature and ambient humidity; A tension prediction unit, which constructs a covered yarn spinning state vector based on the yarn core tension, the motor operation data and the environmental data, calculates the importance weight of each element in the covered yarn spinning state vector to the tension change to obtain an importance weight vector, performs weighted processing on the covered yarn spinning state vector according to the importance weight vector to obtain a weighted state vector, and obtains a predicted yarn core tension at the next sampling moment through a tension prediction model based on the weighted state vector; an error analysis unit, setting a target yarn core tension, calculating a tension error at a current sampling moment according to the target yarn core tension and the yarn core tension at a current sampling moment, calculating a predicted tension error according to the predicted yarn core tension and the target yarn core tension, and calculating a tension error change rate according to the tension error at a current sampling moment and a previous sampling moment; An adaptive control unit obtains a motor current control quantity through an adaptive controller based on the tension error, the predicted tension error and the tension error change rate, calculates a motor voltage control quantity according to the motor current control quantity, controls the spindle rod according to the motor voltage control quantity, and adjusts the adaptive controller parameters in real time through an online parameter optimization method.

[0012] Preferably, the process of weighting the covered yarn spinning state vector comprises: The importance weight of each element in the covered yarn spinning state vector to the tension change is calculated using the following formula: ; in, For the The importance weight of the elements, is the natural exponential function, For the The basic importance coefficient of each element, is the number of elements in the covered yarn spinning state vector, For the Correlation coefficient between each element and tension; Will The importance weights constitute the importance weight vector, and the weighted state vector is obtained by weighting the covered yarn spinning state vector according to the importance weight vector, and the formula is: ; in, The current sampling time The weighted state vector, is the importance weight vector, The current sampling time The spinning state vector of the covered yarn, is the Hadamard product symbol.

[0013] Preferably, the process of the tension prediction model includes: Extract recent The weighted state vectors at the sampling moments and constructing a weighted state vector sequence; The weighted state vector sequence is input into a gated recurrent unit neural network to predict the predicted yarn core tension at the next sampling moment, including: calculating a forgetting coefficient and a retention coefficient based on the weighted state vector at the current sampling moment and the hidden state at the previous sampling moment; using the forgetting coefficient to selectively forget the hidden state at the previous sampling moment, and fusing it with the weighted state vector at the current sampling moment to obtain a candidate hidden state at the current sampling moment; using the retention coefficient to fuse the hidden state at the previous sampling moment with the candidate hidden state at the current sampling moment to obtain the hidden state at the current sampling moment; and obtaining the predicted yarn core tension at the next sampling moment through a fully connected layer based on the hidden state at the current sampling moment.

[0014] Preferably, the adaptive controller comprises: The tension error, the predicted tension error and the tension error change rate are set respectively fuzzy sets; Constructing an adaptive Gaussian membership function based on each fuzzy set, and initializing membership function parameters of each adaptive Gaussian membership function; the membership function parameters include a function center value and a function standard deviation; The membership of the tension error, the predicted tension error and the tension error change rate to the corresponding fuzzy set is calculated respectively by the adaptive Gaussian membership function; An initial fuzzy rule base is constructed based on fuzzy sets to initialize the rule parameters of each fuzzy rule; the initial fuzzy rule base includes fuzzy rules; the rule parameters include a first rule parameter, a second rule parameter and a third rule parameter; the first rule parameter is related to the tension error, the second rule parameter is related to the predicted tension error, and the third rule parameter is related to the tension error change rate; Calculate the motor current adjustment amount under each fuzzy rule according to the rule parameters combined with the tension error, the predicted tension error and the tension error change rate; The trigger strength of each fuzzy rule is calculated based on the membership degree, and the weighted motor current adjustment amount is calculated by weighted average method according to the trigger strength and the motor current adjustment amount, and the motor current control amount is obtained by the weighted motor current adjustment amount.

[0015] Preferably, adjusting the adaptive controller parameters in real time through an online parameter optimization method includes: constructing a comprehensive performance index, wherein the comprehensive performance index is composed of a control accuracy index, a control smoothness index, and a tension stability index; based on the comprehensive performance index, optimizing the rule parameters and the membership function parameters using a gradient descent method.

[0016] Compared with the prior art, the present invention has the following beneficial effects: 1. By constructing a covered yarn textile state vector containing yarn core tension, motor operation data and environmental data, and combining the basic importance coefficient set by historical experience and the tension correlation calculated by the Pearson correlation coefficient, the importance weight of each data is calculated, and weighted processing is performed. It can adaptively adjust the influence of different parameters on tension control, and pay more attention to the key data that has a significant impact on tension changes. By combining the importance weight vector with the covered yarn textile state vector, the obtained weighted state vector more accurately reflects the actual contribution of each data to the system control. It effectively improves the response sensitivity of the control method to changes in key parameters, thereby improving the accuracy of predicting yarn core tension.

[0017] 2. A gated recurrent unit neural network is used to construct a tension prediction model. By extracting a weighted state vector sequence as input, the forget gate and update gate mechanism are used to achieve selective memory and update of historical state information. The model selectively forgets the hidden state of the previous moment by calculating the forgetting coefficient and the retention coefficient, and fuses it with the current state vector to obtain a candidate hidden state; then the historical information is fused with the new information through the retention coefficient, and finally the yarn core tension at the next moment is predicted through the fully connected layer. The prediction mechanism makes full use of the continuity characteristics of production data, can accurately capture the trend of tension changes, improve the pre-control ability of the control method, and improve the control accuracy in a changing production environment.

[0018] 3. An adaptive controller is designed. By setting the fuzzy set of tension error, predicted tension error and tension error change rate, an adaptive Gaussian membership function is constructed, and the motor current control quantity is obtained by combining the fuzzy rule base to realize the underlying control decision and avoid the interference of the production environment on the control accuracy. The controller uses comprehensive performance indicators for online parameter optimization, including control accuracy index, control stability index and tension stability index. The rule parameters and membership function parameters are optimized in real time by the gradient descent method, which reasonably balances the multi-dimensional requirements of control accuracy, stability and stability, overcomes the problem of insufficient adaptive ability of traditional control methods, and significantly improves the control accuracy of control methods in complex and changing environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 A flow chart of a spindle motion state control method for covered yarn spinning according to the present invention; Figure 2 is a flow chart of the adaptive controller of the present invention; Figure 3 The present invention is a structural diagram of a spindle motion state control system for covered yarn spinning. DETAILED DESCRIPTION

[0020] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] Embodiment 1 In order to control the yarn core tension by accurately controlling the spindle bar during the covered yarn spinning process, thereby improving product quality and work efficiency, Factory A used a spindle bar motion state control method for covered yarn spinning. Figure 1 ,include: In the process of weaving the covered yarn, the yarn core tension, the motor operation data and the environmental data are collected according to the sampling period; the motor operation data includes the motor torque, the motor speed and the motor current; the environmental data includes the environmental temperature and the environmental humidity; Constructing a covered yarn spinning state vector based on the yarn core tension, the motor operation data and the environmental data; Calculating the importance weight of each element in the covered yarn spinning state vector to the tension change to obtain an importance weight vector, and performing weighted processing on the covered yarn spinning state vector according to the importance weight vector to obtain a weighted state vector; Based on the weighted state vector, a predicted yarn core tension at the next sampling moment is obtained through a tension prediction model; Setting a target yarn core tension, calculating a tension error at a current sampling moment according to the target yarn core tension and the yarn core tension at a current sampling moment, calculating a predicted tension error according to the predicted yarn core tension and the target yarn core tension, and calculating a tension error change rate according to the tension error at a current sampling moment and a previous sampling moment; Based on the tension error, the predicted tension error and the tension error change rate, the motor current control quantity is obtained through an adaptive controller, the motor voltage control quantity is calculated according to the motor current control quantity, the spindle rod is controlled according to the motor voltage control quantity, and the adaptive controller parameters are adjusted in real time through an online parameter optimization method.

[0022] Furthermore, the target yarn core tension is an ideal yarn core tension value set according to the yarn specifications and coating process requirements, which can ensure that the coated yarn is evenly wound on the yarn core to meet the physical properties and appearance quality requirements of the coated yarn; in actual applications, the target yarn core tension can be preset according to the process requirements and appropriately adjusted according to changes in ambient temperature and humidity to ensure the quality stability of the coated yarn.

[0023] Furthermore, the process of weighting the covered yarn spinning state vector comprises: The importance weight of each element in the covered yarn spinning state vector to the tension change is calculated using the following formula: ; in, For the The importance weight of the elements, is the natural exponential function, For the The basic importance coefficient of each element is set according to historical experience. is the number of elements in the covered yarn spinning state vector, For the The correlation coefficient between each element and tension is calculated using the Pearson correlation coefficient; Will The importance weights constitute the importance weight vector, and the weighted state vector is obtained by weighting the covered yarn spinning state vector according to the importance weight vector, and the formula is: ; in, The current sampling time The weighted state vector, is the importance weight vector, The current sampling time The spinning state vector of the covered yarn, is the Hadamard product symbol.

[0024] By introducing a weight calculation method based on the natural exponential function and combining it with the basic importance coefficient and the Pearson correlation coefficient, the importance of each element of the state vector is accurately quantified, enabling the system to more accurately grasp the influence of each element on tension control, thereby improving the pertinence and effectiveness of the control system.

[0025] Furthermore, the process of the tension prediction model includes: Extract recent The weighted state vectors at the sampling moments and constructing a weighted state vector sequence; The weighted state vector sequence is input into a gated recurrent unit neural network to predict the predicted yarn core tension at the next sampling moment, including: calculating a forgetting coefficient and a retention coefficient based on the weighted state vector at the current sampling moment and the hidden state at the previous sampling moment; using the forgetting coefficient to selectively forget the hidden state at the previous sampling moment, and fusing it with the weighted state vector at the current sampling moment to obtain a candidate hidden state at the current sampling moment; using the retention coefficient to fuse the hidden state at the previous sampling moment with the candidate hidden state at the current sampling moment to obtain the hidden state at the current sampling moment; and obtaining the predicted yarn core tension at the next sampling moment through a fully connected layer based on the hidden state at the current sampling moment.

[0026] The tension prediction model based on the gated recurrent unit neural network realizes the selective memory and update of historical information through the forget gate and update gate mechanism, improves the accuracy of tension prediction, enhances the predictive control capability of the system, and makes the control process more forward-looking and efficient.

[0027] Further, refer to Figure 2 , the adaptive controller comprises: The tension error, the predicted tension error and the tension error change rate are set respectively Fuzzy sets; the specific fuzzy sets are shown in Table 1, and 5 fuzzy sets are set: {negative large, negative small, zero, positive small, positive large}; Constructing an adaptive Gaussian membership function based on each fuzzy set, and initializing membership function parameters of each adaptive Gaussian membership function; the membership function parameters include a function center value and a function standard deviation; The membership of the tension error, the predicted tension error and the tension error change rate to the corresponding fuzzy set is calculated respectively by the adaptive Gaussian membership function; An initial fuzzy rule base is constructed based on fuzzy sets to initialize the rule parameters of each fuzzy rule; the initial fuzzy rule base includes fuzzy rules; the rule parameters include a first rule parameter, a second rule parameter and a third rule parameter; the first rule parameter is related to the tension error, the second rule parameter is related to the predicted tension error, and the third rule parameter is related to the tension error change rate; According to the rule parameters combined with the tension error , the predicted tension error and the tension error change rate The motor current adjustment amount under each of the fuzzy rules is calculated using the following formula: ; in, For the The motor current adjustment amount under the fuzzy rules, For the The first rule parameter corresponding to the fuzzy rule, For the The second rule parameter corresponding to the fuzzy rule, For the The third rule parameter corresponding to the fuzzy rule; The trigger strength of each fuzzy rule is calculated based on the membership degree. The weighted motor current adjustment amount is calculated by the weighted average method according to the trigger strength and the motor current adjustment amount. The motor current control amount is obtained by the weighted motor current adjustment amount. The calculation formula is: ; in, For the The trigger strength of the fuzzy rules, Tension error The corresponding The membership degree of a fuzzy set is The predicted tension error The corresponding The membership degree of a fuzzy set is is the tension error change rate The corresponding The membership degree of a fuzzy set is The motor current control amount is obtained by adding the weighted motor current adjustment amount to the motor current at the last sampling moment.

[0028] Table 1 Fuzzy sets and corresponding membership function parameters

[0029] The formula for calculating the motor voltage control quantity according to the motor current control quantity is: ; in, is the motor voltage control quantity, is the armature resistance, is the back electromotive force coefficient, is the motor angular velocity, is the motor current control quantity. The armature resistance and back electromotive force coefficient are the equipment characteristic parameters of the motor.

[0030] Taking tension error as an example, the corresponding The Gaussian membership function of a fuzzy set is: ; in, Tension error The corresponding The Gaussian membership function of a fuzzy set is is the natural exponential function, is the function center value, is the function standard deviation. The initialization of the function center value and the function standard deviation is shown in Table 1.

[0031] Furthermore, a trigger strength threshold is set. When the trigger strength is less than the trigger strength threshold, it indicates that the corresponding fuzzy rule is not activated. A statistical period is set to statistically analyze the number of inactivation times of each fuzzy rule within the period, and the fuzzy rules whose number of inactivation times is greater than the preset threshold are deleted from the initial fuzzy rule library.

[0032] Some fuzzy rules and their corresponding rule parameters are shown in Table 2. Taking the first fuzzy rule as an example, the rule content is: map the tension error, predicted tension error and tension error change rate to the fuzzy set {negative large}, calculate the membership degree through the function center value and function standard deviation corresponding to the fuzzy set {negative large}, calculate the trigger strength of the first fuzzy rule according to the membership degree, and calculate the motor current adjustment amount under the first fuzzy rule through the corresponding first rule parameters, second rule parameters and third rule parameters.

[0033] Table 2 Fuzzy rules and corresponding parameters

[0034] The adaptive controller adopts Gaussian membership function and dynamically optimized fuzzy rule base, combined with trigger intensity threshold mechanism, to realize adaptive adjustment of control rules, improve the degree of intelligence and control accuracy, and enhance the ability of control method to cope with complex working conditions.

[0035] Furthermore, adjusting the adaptive controller parameters in real time through an online parameter optimization method includes: constructing a comprehensive performance index, wherein the comprehensive performance index is composed of a control accuracy index, a control smoothness index, and a tension stability index; based on the comprehensive performance index, optimizing the rule parameters and the membership function parameters using a gradient descent method.

[0036] The calculation formula of comprehensive performance index is: ; in, The current sampling time The first item of the formula is the control accuracy index, the second item is the control stability index, and the third item is the tension stability index. The current sampling time The tension error, The current sampling time The motor current control quantity, The current sampling time The yarn core tension, The last sampling moment The yarn core tension, , and is the weight coefficient.

[0037] By constructing a comprehensive performance index including control accuracy, smoothness and stability, and using the gradient descent method for parameter optimization, real-time optimization and adjustment of control parameters are achieved, ensuring balanced optimization on multiple performance indicators and improving the overall performance of the control method.

[0038] By constructing a covered yarn spinning state vector containing multi-dimensional data and performing weighted processing based on importance weights, it is possible to adaptively focus on key parameters that have a greater impact on tension changes. The tension prediction model constructed using a gated recurrent unit neural network makes full use of the continuity characteristics of time series data and improves the predictive control capability of the system. At the same time, the designed adaptive controller combines the online parameter optimization mechanism to output the motor current control quantity to control the spindle rod, and at the same time realizes real-time optimization and adjustment of the controller parameters. It not only ensures the precise control of the spindle rod movement, but also enhances the system's adaptability to environmental changes. It overcomes the problems of inaccurate quantification of parameter influence, lack of predictive control and insufficient adaptive ability of traditional control methods, significantly improves the control accuracy in the covered yarn spinning process, and thus improves the quality and production efficiency of the covered yarn.

[0039] Embodiment 2 In order to reduce the defect rate of covered yarn products, Factory B uses a spindle motion state control system for covered yarn spinning, referring to Figure 3 ,include: A data acquisition unit, during the weaving process of the covered yarn, collects yarn core tension, motor operation data and environmental data according to a sampling period; the motor operation data includes motor torque, motor speed and motor current; the environmental data includes ambient temperature and ambient humidity; A tension prediction unit, which constructs a covered yarn spinning state vector based on the yarn core tension, the motor operation data and the environmental data, calculates the importance weight of each element in the covered yarn spinning state vector to the tension change to obtain an importance weight vector, performs weighted processing on the covered yarn spinning state vector according to the importance weight vector to obtain a weighted state vector, and obtains a predicted yarn core tension at the next sampling moment through a tension prediction model based on the weighted state vector; an error analysis unit, setting a target yarn core tension, calculating a tension error at a current sampling moment according to the target yarn core tension and the yarn core tension at a current sampling moment, calculating a predicted tension error according to the predicted yarn core tension and the target yarn core tension, and calculating a tension error change rate according to the tension error at a current sampling moment and a previous sampling moment; An adaptive control unit obtains a motor current control quantity through an adaptive controller based on the tension error, the predicted tension error and the tension error change rate, calculates a motor voltage control quantity according to the motor current control quantity, controls the spindle rod according to the motor voltage control quantity, and adjusts the adaptive controller parameters in real time through an online parameter optimization method.

[0040] Furthermore, the process of weighting the covered yarn spinning state vector comprises: The importance weight of each element in the covered yarn spinning state vector to the tension change is calculated using the following formula: ; in, For the The importance weight of the elements, is the natural exponential function, For the The basic importance coefficient of each element, is the number of elements in the covered yarn spinning state vector, For the Correlation coefficient between each element and tension; Will The importance weights constitute the importance weight vector, and the weighted state vector is obtained by weighting the covered yarn spinning state vector according to the importance weight vector, and the formula is: ; in, The current sampling time The weighted state vector, is the importance weight vector, The current sampling time The spinning state vector of the covered yarn, is the Hadamard product symbol.

[0041] Furthermore, the process of the tension prediction model includes: Extract recent The weighted state vectors at the sampling moments and constructing a weighted state vector sequence; The weighted state vector sequence is input into a gated recurrent unit neural network to predict the predicted yarn core tension at the next sampling moment, including: calculating a forgetting coefficient and a retention coefficient based on the weighted state vector at the current sampling moment and the hidden state at the previous sampling moment; using the forgetting coefficient to selectively forget the hidden state at the previous sampling moment, and fusing it with the weighted state vector at the current sampling moment to obtain a candidate hidden state at the current sampling moment; using the retention coefficient to fuse the hidden state at the previous sampling moment with the candidate hidden state at the current sampling moment to obtain the hidden state at the current sampling moment; and obtaining the predicted yarn core tension at the next sampling moment through a fully connected layer based on the hidden state at the current sampling moment.

[0042] Furthermore, the adaptive controller comprises: The tension error, the predicted tension error and the tension error change rate are set respectively fuzzy sets; Constructing an adaptive Gaussian membership function based on each fuzzy set, and initializing membership function parameters of each adaptive Gaussian membership function; the membership function parameters include a function center value and a function standard deviation; The membership of the tension error, the predicted tension error and the tension error change rate to the corresponding fuzzy set is calculated respectively by the adaptive Gaussian membership function; An initial fuzzy rule base is constructed based on fuzzy sets to initialize the rule parameters of each fuzzy rule; the initial fuzzy rule base includes fuzzy rules; the rule parameters include a first rule parameter, a second rule parameter and a third rule parameter; the first rule parameter is related to the tension error, the second rule parameter is related to the predicted tension error, and the third rule parameter is related to the tension error change rate; Calculate the motor current adjustment amount under each fuzzy rule according to the rule parameters combined with the tension error, the predicted tension error and the tension error change rate; The trigger strength of each fuzzy rule is calculated based on the membership degree, and the weighted motor current adjustment amount is calculated by weighted average method according to the trigger strength and the motor current adjustment amount, and the motor current control amount is obtained by the weighted motor current adjustment amount.

[0043] Furthermore, adjusting the adaptive controller parameters in real time through an online parameter optimization method includes: constructing a comprehensive performance index, wherein the comprehensive performance index is composed of a control accuracy index, a control smoothness index, and a tension stability index; based on the comprehensive performance index, optimizing the rule parameters and the membership function parameters using a gradient descent method.

[0044] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for controlling the motion state of a spindle bar for covered yarn spinning, characterized in that: include: During the covered yarn spinning process, yarn core tension, motor operation data and environmental data are collected according to the sampling period; The motor operation data includes motor torque, motor speed, motor angular velocity and motor current; the environmental data includes ambient temperature and ambient humidity; Constructing a covered yarn spinning state vector based on the yarn core tension, the motor operation data and the environmental data; Calculating the importance weight of each element in the covered yarn spinning state vector to the tension change to obtain an importance weight vector, and performing weighted processing on the covered yarn spinning state vector according to the importance weight vector to obtain a weighted state vector; Based on the weighted state vector, a predicted yarn core tension at the next sampling moment is obtained through a tension prediction model; Setting a target yarn core tension, calculating a tension error at a current sampling moment according to the target yarn core tension and the yarn core tension at a current sampling moment, calculating a predicted tension error according to the predicted yarn core tension and the target yarn core tension, and calculating a tension error change rate according to the tension error at a current sampling moment and a previous sampling moment; Based on the tension error, the predicted tension error and the tension error change rate, the motor current control quantity is obtained through an adaptive controller, the motor voltage control quantity is calculated according to the motor current control quantity, the spindle rod is controlled according to the motor voltage control quantity, and the adaptive controller parameters are adjusted in real time through an online parameter optimization method.

2. A spindle motion state control method for covered yarn spinning according to claim 1, characterized in that: The process of weighting the covered yarn spinning state vector comprises: The importance weight of each element in the covered yarn spinning state vector to the tension change is calculated using the following formula: ； in, For the The importance weight of the elements, is the natural exponential function, For the The basic importance coefficient of each element, is the number of elements in the covered yarn spinning state vector, For the Correlation coefficient between each element and tension; Will The importance weights constitute the importance weight vector, and the weighted state vector is obtained by weighting the covered yarn spinning state vector according to the importance weight vector, and the formula is: ； in, The current sampling time The weighted state vector, is the importance weight vector, The current sampling time The spinning state vector of the covered yarn, is the Hadamard product symbol.

3. A spindle motion state control method for covered yarn spinning according to claim 1, characterized in that: The process of the tension prediction model includes: Extract recent The weighted state vectors at the sampling moments and constructing a weighted state vector sequence; The weighted state vector sequence is input into a gated recurrent unit neural network to predict the predicted yarn core tension at the next sampling moment, including: calculating a forgetting coefficient and a retention coefficient based on the weighted state vector at the current sampling moment and the hidden state at the previous sampling moment; using the forgetting coefficient to selectively forget the hidden state at the previous sampling moment, and fusing it with the weighted state vector at the current sampling moment to obtain a candidate hidden state at the current sampling moment; using the retention coefficient to fuse the hidden state at the previous sampling moment with the candidate hidden state at the current sampling moment to obtain the hidden state at the current sampling moment; and obtaining the predicted yarn core tension at the next sampling moment through a fully connected layer based on the hidden state at the current sampling moment.

4. A spindle motion state control method for covered yarn spinning according to claim 1, characterized in that: The adaptive controller comprises: The tension error, the predicted tension error and the tension error change rate are set respectively fuzzy sets; Constructing an adaptive Gaussian membership function based on each fuzzy set, and initializing membership function parameters of each adaptive Gaussian membership function; the membership function parameters include a function center value and a function standard deviation; The membership of the tension error, the predicted tension error and the tension error change rate to the corresponding fuzzy set is calculated respectively by the adaptive Gaussian membership function; An initial fuzzy rule base is constructed based on fuzzy sets to initialize the rule parameters of each fuzzy rule; the initial fuzzy rule base includes fuzzy rules; the rule parameters include a first rule parameter, a second rule parameter and a third rule parameter; the first rule parameter is related to the tension error, the second rule parameter is related to the predicted tension error, and the third rule parameter is related to the tension error change rate; Calculate the motor current adjustment amount under each fuzzy rule according to the rule parameters combined with the tension error, the predicted tension error and the tension error change rate; The trigger strength of each fuzzy rule is calculated based on the membership degree, and the weighted motor current adjustment amount is calculated by weighted average method according to the trigger strength and the motor current adjustment amount, and the motor current control amount is obtained by the weighted motor current adjustment amount.

5. A spindle motion state control method for covered yarn spinning according to claim 4, characterized in that: Real-time adjustment of adaptive controller parameters by online parameter optimization method includes: constructing a comprehensive performance index, wherein the comprehensive performance index is composed of a control accuracy index, a control smoothness index and a tension stability index; based on the comprehensive performance index, optimizing the rule parameters and the membership function parameters by using the gradient descent method.

6. A spindle motion state control system for covered yarn spinning, characterized in that: include: A data acquisition unit collects yarn core tension, motor operation data and environmental data according to a sampling period during the covered yarn spinning process; The motor operation data includes motor torque, motor speed and motor current; the environmental data includes ambient temperature and ambient humidity; A tension prediction unit, which constructs a covered yarn spinning state vector based on the yarn core tension, the motor operation data and the environmental data, calculates the importance weight of each element in the covered yarn spinning state vector to the tension change to obtain an importance weight vector, performs weighted processing on the covered yarn spinning state vector according to the importance weight vector to obtain a weighted state vector, and obtains a predicted yarn core tension at the next sampling moment through a tension prediction model based on the weighted state vector; an error analysis unit, setting a target yarn core tension, calculating a tension error at a current sampling moment according to the target yarn core tension and the yarn core tension at a current sampling moment, calculating a predicted tension error according to the predicted yarn core tension and the target yarn core tension, and calculating a tension error change rate according to the tension error at a current sampling moment and a previous sampling moment; An adaptive control unit obtains a motor current control quantity through an adaptive controller based on the tension error, the predicted tension error and the tension error change rate, calculates a motor voltage control quantity according to the motor current control quantity, controls the spindle rod according to the motor voltage control quantity, and adjusts the adaptive controller parameters in real time through an online parameter optimization method.

7. A spindle motion state control system for covered yarn spinning according to claim 6, characterized in that: The process of weighting the covered yarn spinning state vector comprises: The importance weight of each element in the covered yarn spinning state vector to the tension change is calculated using the following formula: ； in, For the The importance weight of the elements, is the natural exponential function, For the The basic importance coefficient of each element, is the number of elements in the covered yarn spinning state vector, For the Correlation coefficient between each element and tension; Will The importance weights constitute the importance weight vector, and the weighted state vector is obtained by weighting the covered yarn spinning state vector according to the importance weight vector, and the formula is: ； in, The current sampling time The weighted state vector, is the importance weight vector, The current sampling time The spinning state vector of the covered yarn, is the Hadamard product symbol.

8. A spindle motion state control system for covered yarn spinning according to claim 6, characterized in that: The process of the tension prediction model includes: Extract recent The weighted state vectors at the sampling moments and constructing a weighted state vector sequence; The weighted state vector sequence is input into a gated recurrent unit neural network to predict the predicted yarn core tension at the next sampling moment, including: calculating a forgetting coefficient and a retention coefficient based on the weighted state vector at the current sampling moment and the hidden state at the previous sampling moment; using the forgetting coefficient to selectively forget the hidden state at the previous sampling moment, and fusing it with the weighted state vector at the current sampling moment to obtain a candidate hidden state at the current sampling moment; using the retention coefficient to fuse the hidden state at the previous sampling moment with the candidate hidden state at the current sampling moment to obtain the hidden state at the current sampling moment; and obtaining the predicted yarn core tension at the next sampling moment through a fully connected layer based on the hidden state at the current sampling moment.

9. A spindle motion state control system for covered yarn spinning according to claim 6, characterized in that: The adaptive controller comprises: The tension error, the predicted tension error and the tension error change rate are set respectively fuzzy sets; Constructing an adaptive Gaussian membership function based on each fuzzy set, and initializing membership function parameters of each adaptive Gaussian membership function; the membership function parameters include a function center value and a function standard deviation; The membership of the tension error, the predicted tension error and the tension error change rate to the corresponding fuzzy set is calculated respectively by the adaptive Gaussian membership function; An initial fuzzy rule base is constructed based on fuzzy sets to initialize the rule parameters of each fuzzy rule; the initial fuzzy rule base includes fuzzy rules; the rule parameters include a first rule parameter, a second rule parameter and a third rule parameter; the first rule parameter is related to the tension error, the second rule parameter is related to the predicted tension error, and the third rule parameter is related to the tension error change rate; Calculate the motor current adjustment amount under each fuzzy rule according to the rule parameters combined with the tension error, the predicted tension error and the tension error change rate; The trigger strength of each fuzzy rule is calculated based on the membership degree, and the weighted motor current adjustment amount is calculated by weighted average method according to the trigger strength and the motor current adjustment amount, and the motor current control amount is obtained by the weighted motor current adjustment amount.

10. A spindle motion state control system for covered yarn spinning according to claim 9, characterized in that: Real-time adjustment of adaptive controller parameters by online parameter optimization method includes: constructing a comprehensive performance index, wherein the comprehensive performance index is composed of a control accuracy index, a control smoothness index and a tension stability index; based on the comprehensive performance index, optimizing the rule parameters and the membership function parameters by using the gradient descent method.

Images