A glass fiber cloth warp tension control system

Abstract

The application provides a glass fiber cloth warp tension control system, comprising a double-sensor measurement unit, a signal processing unit, a control unit and an execution unit connected electrically; the double-sensor measurement unit comprises a front-end strain type tension sensor and a rear-end strain type tension sensor; the front-end strain type tension sensor is installed between a warp feed roller and a warp stopper of the loom and is used for collecting tension data at an upstream position of the warp; the rear-end strain type tension sensor is installed between a reed and a cloth roller of the loom and is used for collecting tension data of the whole fabric; the signal processing unit comprises a first signal conditioning module, a second signal conditioning module and a data fusion module; the control unit comprises a basic PID module, a tension prediction module for running a tension prediction algorithm based on an LSTM neural network, a compensation calculation module and a control fusion module; and the execution unit comprises a warp feed motor. The application improves the control precision of the warp tension control of the loom.

Description

Technical Field

[0001] This invention relates to a warp tension control system, specifically a glass fiber fabric warp tension control system, belonging to the field of textile machinery control technology. Background Technology

[0002] Fiberglass cloth is an important industrial fabric with excellent high-temperature resistance, corrosion resistance, and insulation properties, and is widely used in composite material reinforcement, electronic circuit boards, and insulation materials. Due to the characteristics of fiberglass filaments, problems such as filament breakage and fuzzing are prone to occur during the weaving process, and these problems are closely related to the control of warp tension. Therefore, the control of warp tension has a significant impact on the quality of fiberglass cloth fabrics.

[0003] Currently, commonly used warp tension control systems primarily employ a single-sensor measurement combined with PID control. The tension sensor is installed between the feed roller and the stop roller, measuring only the tension changes in the warp area. This single-point measurement method cannot comprehensively reflect the tension variation patterns during the weaving process, neglecting the cumulative effect of warp tension and failing to accurately reflect the impact of the weaving process on warp tension. This limits the control accuracy of warp tension during fiberglass fabric weaving. Furthermore, the control parameters of PID control are relatively fixed, making it difficult to adapt to dynamic changes in weaving conditions. The control response exhibits lag and cannot effectively cope with sudden tension fluctuations.

[0004] Therefore, there is an urgent need in this field for a novel glass fiber fabric warp tension control system. Summary of the Invention

[0005] Based on the above background, the purpose of this invention is to provide a glass fiber fabric warp tension control system, which improves the control accuracy of warp tension control on a loom by using dual-sensor collaborative measurement, combined with tension prediction and dual-level control.

[0006] To achieve the above-mentioned objectives, the present invention provides the following technical solution:

[0007] A glass fiber fabric warp tension control system is used to control the warp tension of a loom. The glass fiber fabric warp tension control system includes an electrically connected dual-sensor measurement unit, a signal processing unit, a control unit, and an execution unit.

[0008] The dual-sensor measurement unit includes a front-end strain gauge tension sensor and a rear-end strain gauge tension sensor. The front-end strain gauge tension sensor is installed between the warp feed roller and the stop warp plate of the loom and is used to collect tension data at the upstream position of the warp yarn. The rear-end strain gauge tension sensor is installed between the reed and the take-up roller of the loom and is used to collect tension data of the entire fabric.

[0009] The signal processing unit includes a first signal conditioning module for processing sampling data of the front-end strain gauge tension sensor, a second signal conditioning module for processing sampling data of the back-end strain gauge tension sensor, and a data fusion module for running an adaptive weighting algorithm;

[0010] The control unit includes a basic PID module, a tension prediction module for running a tension prediction algorithm based on LSTM neural network, a compensation calculation module, and a control fusion module;

[0011] The execution unit includes a let-off motor.

[0012] Preferably, the adaptive weighting algorithm uses the following formula to calculate the fused measured tension value F,

[0013] F=w1·Fs+w2·Fp;

[0014] In the formula, Fs is the tension data output by the first signal conditioning module, Fp is the tension data output by the second signal conditioning module, and w1 and w2 are weight coefficients;

[0015] Wherein, the values of w1 and w2 follow the following formulas:

[0016] When v≤100m / min, w1=0.7, w2=0.3;

[0017] When 100m / min<v≤200m / min, w1=0.5, w2=0.5;

[0018] When v>200m / min, w1=0.3, w2=0.7;

[0019] In the formula, v is the running speed of the loom.

[0020] Preferably, the basic PID module is configured to calculate the basic torque adjustment amount ΔM of the let-off motor according to the deviation between the set tension value and the fused measured tension value and the deviation change rate through the following formula b (t),

[0021] ΔM b (t)=K p ×e(t)+K i ×∫e(t)dt+K d ×de(t) / dt;

[0022] In the formula, K p is the proportional coefficient, K i is the integral coefficient, K d is the differential coefficient, e(t) is the deviation between the set tension value and the fused measured tension value, ∫ is the integral symbol, and d is the differential symbol.

[0023] Preferably, the input to the tension prediction algorithm is an input vector composed of a fused sequence of measured tension values, a loom speed sequence, a temperature parameter sequence, and a humidity parameter sequence from a set number of historical sampling periods. The output of the tension prediction algorithm is a sequence of predicted tension values ​​for a set number of future sampling periods. The tension prediction algorithm is constructed based on an LSTM neural network, comprising an input layer, three LSTM layers, two fully connected layers, and an output layer arranged sequentially. The first LSTM layer contains 128 neurons, the second LSTM layer contains 64 neurons, and the third LSTM layer contains 32 neurons. The first fully connected layer contains 64 neurons, and the second fully connected layer contains 32 neurons. Both fully connected layers use the ReLU activation function, and the output layer uses a linear activation function.

[0024] Preferably, the number of historical sampling periods is greater than the number of future sampling periods.

[0025] Preferably, the compensation calculation module is used to calculate a predicted tension deviation sequence for a set number of future sampling periods based on the predicted tension value sequence for a set number of future sampling periods output by the tension prediction module. The comprehensive prediction deviation is calculated using a weighted time series method based on the predicted tension deviation sequence, and the compensation torque adjustment ΔM of the motor is calculated using the following formula. f (t),

[0026] ΔM f (t)=K pf ×e p (t)+K if ×∫e p (t)dt+K df ×de p (t) / dt;

[0027] In the formula, K pf K is the compensation ratio coefficient. if To compensate for the integral coefficient, K df To compensate for the differential coefficients, e p (t) represents the overall prediction bias, ∫ represents the integral sign, and d represents the differential sign.

[0028] Preferably, the comprehensive prediction deviation is calculated using the following formula.

[0029] e p (t)=Σ(ui×e'(t+i));

[0030] In the formula, ui is the weight coefficient of the i-th deviation value in the predicted tension deviation sequence, e'(t+i) is the i-th deviation value in the predicted tension deviation sequence, i = 1 to M, and M is the set number of future sampling periods.

[0031] Preferably, the control fusion module is used to calculate the final torque adjustment ΔM(t) of the feed motor based on the basic torque adjustment of the feed motor output by the basic PID module and the compensated torque adjustment of the feed motor output by the compensation calculation module, using the following formula.

[0032] ΔM(t)=αΔM b (t)+βΔM f (t);

[0033] In the formula, ΔM b (t) is the basic torque adjustment amount, ΔM f (t) represents the compensation torque adjustment amount, and α and β are the variation weight coefficients. When the prediction accuracy index PA≥0.95, α=0.7, β=0.3; when 0.9≤PA<0.95, α=0.5, β=0.5; when PA<0.9, α=0.3, β=0.7.

[0034] Preferably, the prediction accuracy index is calculated using the following formula.

[0035] PA = 1 - |T'(t) - T(t)| / T(t);

[0036] In the formula, T'(t) is the predicted tension value in the t sampling period of the predicted tension value sequence output by the tension prediction module in the t-1 sampling period, and T(t) is the fused measured tension value output by the data fusion module in the t sampling period.

[0037] Compared with the prior art, the present invention has the following advantages:

[0038] The system employs a dual-sensor collaborative measurement approach. A front-end strain gauge tension sensor is installed in the warp area to detect tension fluctuations generated by the warp feeding system in a timely manner, while a rear-end strain gauge tension sensor is installed in the fabric forming area to detect tension changes caused by the weaving process. After adaptive weighted fusion, the measurement data from the two sensors can simultaneously reflect the instantaneous changes and cumulative effects of warp tension, effectively improving control accuracy.

[0039] Using a basic PID module as the control foundation, an LSTM neural network is used to achieve tension prediction. Combined with compensation calculation and control fusion, a two-level control is achieved, ultimately realizing precise control with feedforward compensation for the torque regulation of the feed motor. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0041] Figure 1 This is a schematic diagram of a glass fiber cloth warp tension control system according to the present invention;

[0042] In the diagram: 1. Front-end strain gauge tension sensor; 2. Rear-end strain gauge tension sensor; 3. Warp feed roller; 4. Warp stop plate; 5. Reed; 6. Fabric winding roller; 7. Heald frame; 8. Warp yarn; 9. Fabric; 10. First signal conditioning module; 11. Second signal conditioning module; 12. Data fusion module; 13. Basic PID module; 14. Tension prediction module; 15. Compensation calculation module; 16. Control fusion module; 17. Warp feed motor. Detailed Implementation

[0043] The technical solution of the present invention will be further described in detail below through specific embodiments and in conjunction with the accompanying drawings. It should be understood that the implementation of the present invention is not limited to the following embodiments, and any modifications and / or alterations made to the present invention will fall within the protection scope of the present invention.

[0044] In this invention, unless otherwise specified, all parts and percentages are by weight, and the equipment and raw materials used are commercially available or commonly used in the art. Unless otherwise specified, the methods in the following embodiments are conventional methods in the art. Unless otherwise specified, the components or equipment in the following embodiments are general standard parts or components known to those skilled in the art, and their structures and principles can be learned by those skilled in the art through technical manuals or conventional experimental methods.

[0045] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings. In this detailed description, numerous specific details are set forth to facilitate explanation and provide a thorough understanding of the embodiments of the present invention. However, one or more embodiments may be practiced by those skilled in the art without these specific details.

[0046] An embodiment of the present invention discloses a glass fiber fabric warp tension control system for controlling the warp tension of a loom. The glass fiber fabric warp tension control system includes an electrically connected dual-sensor measurement unit, a signal processing unit, a control unit, and an execution unit.

[0047] The dual-sensor measurement unit includes a front-end strain gauge tension sensor 1 and a rear-end strain gauge tension sensor 2. The front-end strain gauge tension sensor 1 is installed between the warp feed roller 3 and the stop warp plate 4 of the loom to collect tension data at the upstream position of the warp yarn 8. The rear-end strain gauge tension sensor 2 is installed between the reed 5 and the take-up roller 6 of the loom to collect tension data of the entire fabric 9.

[0048] The signal processing unit includes a first signal conditioning module 10 for processing the sampling data of the front-end strain gauge tension sensor 1, a second signal conditioning module 11 for processing the sampling data of the back-end strain gauge tension sensor 2, and a data fusion module 12 for running an adaptive weighting algorithm.

[0049] The control unit includes a basic PID module 13, a tension prediction module 14 for running a tension prediction algorithm based on an LSTM neural network, a compensation calculation module 15, and a control fusion module 16.

[0050] The execution unit includes a feed motor 17.

[0051] The front-end strain gauge tension sensor 1 is installed between the warp feed roller 3 and the stop warp strip 4 on the loom, at a distance of 450-80mm from the stop warp strip. The rear-end strain gauge tension sensor 2 is installed between the reed 5 and the take-up roller 6 on the loom, at a distance of 100-120mm from the reed 5. Both sensors are mounted on adjustable mounting bases via independent vibration damping brackets. The height and horizontal position of the mounting bases can be adjusted within ±20mm and ±15mm, respectively. The strain gauge tension sensor is existing technology, and its specific structure will not be described in detail. The sampling frequency of the front-end strain gauge tension sensor 1 is 1000–1200Hz, and the sampling frequency of the rear-end strain gauge tension sensor 2 is 500–600Hz. This is because the tension fluctuation frequency in the warp yarn area 8 is relatively high, requiring a higher sampling frequency for accurate capture, while the tension change in the fabric area 9 is relatively gentle, and a lower sampling frequency is sufficient for measurement.

[0052] The first signal conditioning module 10 and the second signal conditioning module 11 adopt the same structure, both including a first-stage amplifier circuit, a second-order Butterworth low-pass filter circuit, and a second-stage amplifier circuit. The difference lies in that the cutoff frequency of the low-pass filter circuit in the first signal conditioning module 10 is 200Hz, while the cutoff frequency of the low-pass filter circuit in the second signal conditioning module 11 is 100Hz, thus employing different filtering parameters to match the tension characteristics at different positions.

[0053] The adaptive weighting algorithm running in data fusion module 12 calculates the fused measured tension value F using the following formula:

[0054] F = w1·Fs + w2·Fp;

[0055] where Fs is the tension data output by the first signal conditioning module 10, Fp is the tension data output by the second signal conditioning module 11, and w1 and w2 are weight coefficients;

[0056] wherein the values of w1 and w2 follow the following formulas:

[0057] When v≤100m / min, w1=0.7, w2=0.3;

[0058] When 100m / min<v≤200m / min, w1=0.5, w2=0.5;

[0059] When v>200m / min, w1=0.3, w2=0.7;

[0060] where v is the running speed of the loom.

[0061] In the warp yarn 8 area, the front-end strain-type tension sensor 1 can timely perceive the tension fluctuation generated by the warp let-off system, and in the formed fabric area, the rear-end strain-type tension sensor 2 can detect the tension change caused by the weaving process. After the measurement data of the two sensors are subjected to adaptive weighted fusion, the instantaneous change and cumulative effect of the tension of the warp yarn 8 can be reflected at the same time, and the control accuracy is effectively improved.

[0062] The basic PID module 13 is configured to calculate the basic torque adjustment amount ΔM of the warp let-off motor 17 according to the deviation between the set tension value and the fused measured tension value and the deviation change rate through the following formula b (t),

[0063] ΔM b (t)=K p ×e(t)+K i ×∫e(t)dt+K d ×de(t) / dt;

[0064] where K p is the proportional coefficient, K i is the integral coefficient, K d is the differential coefficient, e(t) is the deviation between the set tension value and the fused measured tension value, ∫ is the integral symbol, and d is the differential symbol.

[0065] In the tension prediction algorithm run by the tension prediction module 14, the input is an input vector composed of the fused measured tension value sequence of a set number of historical sampling periods, the loom speed sequence, the temperature parameter sequence and the humidity parameter sequence, and the output of the tension prediction algorithm is a predicted tension value sequence of a set number of future sampling periods.

[0066] The tension prediction algorithm is based on an LSTM neural network, which is suitable for processing time series data. The algorithm consists of an input layer, three LSTM layers, two fully connected layers, and an output layer, arranged sequentially. The first LSTM layer has 128 neurons, the second has 64 neurons, and the third has 32 neurons. The first fully connected layer has 64 neurons, and the second has 32 neurons. Both fully connected layers use the ReLU activation function, and the output layer uses a linear activation function. The three-layer LSTM structure extracts short-term, medium-term, and long-term features, while the fully connected layers achieve deep feature fusion.

[0067] The number of historical sampling periods is set greater than the number of future sampling periods.

[0068] The compensation calculation module 15 is used to calculate the predicted tension deviation sequence for a set number of future sampling periods based on the predicted tension value sequence for a set number of future sampling periods output by the tension prediction module 14. The comprehensive prediction deviation is calculated using a weighted time series method based on the predicted tension deviation sequence. The compensation torque adjustment ΔM of the feed motor 17 is then calculated using the following formula. f (t),

[0069] ΔM f (t)=K pf ×e p (t)+K if ×∫e p (t)dt+K df ×de p (t) / dt;

[0070] In the formula, K pf K is the compensation ratio coefficient. if To compensate for the integral coefficient, K df To compensate for the differential coefficients, e p (t) represents the overall prediction bias, ∫ represents the integral sign, and d represents the differential sign.

[0071] Basic PID control alone is insufficient to cope with complex dynamic changes. Predictive compensation is used to anticipate tension changes, and the compensation control uses the same PID structure as the basic PID control, thereby maintaining coordination with the basic PID control.

[0072] The overall prediction deviation is calculated using the following formula.

[0073] e p (t)=Σ(ui×e'(t+i));

[0074] In the formula, ui is the weight coefficient of the i-th deviation value in the predicted tension deviation sequence, e'(t+i) is the i-th deviation value in the predicted tension deviation sequence, i = 1 to M, and M is the set number of future sampling periods.

[0075] The control fusion module 16 is used to calculate the final torque adjustment ΔM(t) of the feed motor 17 based on the basic torque adjustment amount of the feed motor 17 output by the basic PID module 13 and the compensated torque adjustment amount of the feed motor 17 output by the compensation calculation module 15, using the following formula.

[0076] ΔM(t)=αΔM b (t)+βΔM f (t);

[0077] In the formula, ΔM b (t) is the basic torque adjustment amount, ΔM f (t) represents the compensation torque adjustment amount, and α and β are the variation weight coefficients. When the prediction accuracy index PA≥0.95, α=0.7, β=0.3; when 0.9≤PA<0.95, α=0.5, β=0.5; when PA<0.9, α=0.3, β=0.7.

[0078] Basic control and predictive compensation are dynamically balanced based on the reliability of the prediction. When the prediction is accurate, the compensation effect is increased, and when the prediction is inaccurate, basic control takes the lead, thereby achieving the optimal balance between control stability and foresight.

[0079] The prediction accuracy index is calculated using the following formula.

[0080] PA = 1 - |T'(t) - T(t)| / T(t);

[0081] In the formula, T'(t) is the predicted tension value of the tension prediction module 14 in the predicted tension value sequence output in the t-1 sampling period, and T(t) is the fused measured tension value output by the data fusion module 12 in the t sampling period.

[0082] The fiberglass fabric warp tension control system employs dual-sensor collaborative measurement. The front-end strain gauge tension sensor 1 is installed in the warp yarn 8 area to promptly detect tension fluctuations generated by the warp feeding system. The rear-end strain gauge tension sensor 2 is installed in the fabric forming area to detect tension changes caused by the weaving process. After adaptive weighted fusion, the measurement data from the two sensors can simultaneously reflect the instantaneous changes and cumulative effects of the warp yarn 8 tension, effectively improving control accuracy.

[0083] The glass fiber fabric warp tension control system uses the basic PID module 13 as the control basis, uses LSTM neural network to realize tension prediction, and performs two-level control by combining compensation calculation and control fusion, and finally realizes precise control with feedforward compensation for the torque adjustment of the warp motor 17.

[0084] This article uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. It should be noted that those skilled in the art can make several improvements and modifications to the present invention without departing from the principles of the present invention, and these improvements and modifications also fall within the protection scope of the claims of the present invention.

Claims

1. A glass fiber fabric warp tension control system for controlling the warp tension of a loom, characterized in that: This warp tension control system for glass fiber fabric comprises a dual-sensor measurement unit, a signal processing unit, an intelligent control unit, a control unit and an execution unit; Said dual-sensor measurement unit includes a front-end strain-type tension sensor (1) and a back-end strain-type tension sensor (2); the front-end strain-type tension sensor (1) is installed between the let-off roller (3) and the dropper (4) of the loom, and is used to collect tension data at the upstream position of the warp (8); the back-end strain-type tension sensor (2) is installed between the reed (5) and the cloth take-up roller (6) of the loom, and is used to collect tension data of the whole fabric (9); The signal processing unit includes a first signal conditioning module (10) for processing the sampling data of the front-end strain-type tension sensor (1), a second signal conditioning module (11) for processing the sampling data of the back-end strain-type tension sensor (2), and a data fusion module (12) for running the adaptive weighting algorithm; The control unit includes a basic PID module (13), a tension prediction module (14) for running the LSTM neural network-based tension prediction algorithm, a compensation calculation module (15), and a control fusion module (16); The execution unit includes a let-off motor (17); Wherein, the adaptive weighting algorithm uses the following formula to calculate the fused measured tension value F, F = w1·Fs + w2·Fp; In the formula, Fs is the tension data output by the first signal conditioning module (10), Fp is the tension data output by the second signal conditioning module (11), and w1 and w2 are weight coefficients; Wherein, the values of w1 and w2 follow the following formulas: When v≤100m / min, w1=0.7, w2=0.3; When 100m / min<v≤200m / min, w1=0.5, w2=0.5; When v>200m / min, w1=0.3, w2=0.7; In the formula, v is the running speed of the loom; The basic PID module (13) is used to calculate the basic torque adjustment amount ΔM of the feed motor (17) based on the deviation between the set tension value and the fused measured tension value and the rate of change of the deviation, using the following formula. b (t), ΔM b (t)= K p ×e(t)+K i ×∫e(t)dt+K d ×de(t) / dt; In the formula, K p K is the proportionality coefficient. i K is the integral coefficient. d ∫ is the differential coefficient, e(t) is the deviation between the set tension value and the measured tension value after fusion, ∫ is the integral sign, and d is the differential sign; The compensation calculation module (15) is used to calculate the predicted tension deviation sequence for a set number of future sampling periods based on the predicted tension value sequence for a set number of future sampling periods output by the tension prediction module (14). The comprehensive prediction deviation is calculated using the weighted time series method based on the predicted tension deviation sequence. The compensation torque adjustment amount ΔM of the feed motor (17) is calculated using the following formula. f (t), ΔM f (t)= K pf ×e p (t)+K if ×∫e p (t)dt+K df ×de p (t) / dt; In the formula, K pf K is the compensation ratio coefficient. if To compensate for the integral coefficient, K df To compensate for the differential coefficients, e p (t) represents the overall prediction bias, ∫ represents the integral sign, and d represents the differential sign; The comprehensive prediction deviation is calculated by the following formula, e p (t)= Σ(ui×e'(t+i)); In the formula, ui is the weight coefficient of the i-th deviation value in the predicted tension deviation sequence, e'(t+i) is the i-th deviation value in the predicted tension deviation sequence, i=1~M, and M is the set number of future sampling cycles; Said control fusion module (16) is configured to calculate the final torque adjustment amount ΔM(t) of the let-off motor (17) through the following formula according to the basic torque adjustment amount of the let-off motor (17) output by the basic PID module (13) and the compensation torque adjustment amount of the let-off motor (17) output by the compensation calculation module (15), ΔM(t)= αΔM b (t) + βΔM f (t); In the formula, ΔM b (t) is the basic torque adjustment amount, ΔM f (t) represents the compensation torque adjustment amount, and α and β are the variation weight coefficients. When the prediction accuracy index PA≥0.95, α=0.3 and β=0.7; when 0.9≤PA<0.95, α=0.5 and β=0.5; when PA<0.9, α=0.7 and β=0.

3. The prediction accuracy index is calculated by the following formula, PA=1-|T'(t)-T(t)| / T(t); In the formula, T'(t) is the predicted tension value at the t-th sampling period in the predicted tension value sequence output by the tension prediction module (14) at the (t-1)-th sampling period, and T(t) is the fused measured tension value output by the data fusion module (12) at the t-th sampling period.

2. The glass fiber fabric warp tension control system according to claim 1, characterized in that: The input to the tension prediction algorithm is an input vector composed of a fused sequence of measured tension values, a loom speed sequence, a temperature parameter sequence, and a humidity parameter sequence from a set number of historical sampling periods. The output of the tension prediction algorithm is a predicted tension value sequence for a set number of future sampling periods. The tension prediction algorithm is built on an LSTM neural network and includes an input layer, three LSTM layers, two fully connected layers, and an output layer arranged sequentially. The first LSTM layer includes 128 neurons, the second LSTM layer includes 64 neurons, and the third LSTM layer includes 32 neurons. The first fully connected layer includes 64 neurons, and the second fully connected layer includes 32 neurons. Both fully connected layers use the ReLU activation function, and the output layer uses a linear activation function.

3. The glass fiber cloth warp tension control system according to claim 2, characterized in that: The set number of historical sampling periods is greater than the set number of future sampling periods.

Images