Anfis system for precise pressure control of high speed on-off valve and control method thereof

Abstract

The application provides an ANFIS system for precise pressure control of a high-speed switch valve and a control method thereof. An air inlet of an air inlet valve is connected with an air source through a pneumatic three-way joint, and an air outlet of the air inlet valve is connected with a fixed volume chamber. An air inlet of an air outlet valve is connected with the fixed volume chamber, and an air outlet of the air outlet valve is connected with an atmospheric environment. The air inlet valve and the air outlet valve are respectively connected with a control system. The application has strong adaptability and learning ability, can adjust a control strategy on line to adapt to changes of system dynamics and uncertainty, effectively improves system performance, and improves pressure tracking precision. The use of the ANFIS enables the system to better adapt to nonlinear and time-varying characteristics of the volume chamber pressure system, and improves robustness and performance of the controller. Slow pressurization and depressurization can be realized in the joint control strategy, so as to realize fine adjustment of pressure and reduce overshoot.

Description

Technical Field

[0001] This invention provides an ANFIS system and its control method for precise pressure control of high-speed switching valves, belonging to the field of pneumatic control technology. Background Technology

[0002] Pneumatic servo systems utilize the energy of compressed air, which is then converted back into mechanical energy through pneumatic components. Pneumatic systems offer advantages such as high-speed response, reliability, durability, and zero pollution, and are widely used in automated production lines, industrial manufacturing, transportation systems, and many other fields. The commonly used adjustment method for pneumatic pressure servo systems is proportional valve control. However, the large size, complex structure, and high cost of proportional valves limit their wider application. High-speed switching valves, due to their low cost, strong anti-pollution capability, high repeatability, and long service life, have received widespread attention in the field of pneumatic system research. High-speed switching valves operate in a fully open or fully closed state, and the systems they control are highly discrete and nonlinear, making it difficult to establish accurate mathematical models.

[0003] The existing control methods for pneumatic pressure servo systems mainly include PID (proportional-integral-derivative) control and fuzzy control.

[0004] PID control is a linear control method. As a strongly nonlinear system, pneumatic pressure servo systems require careful parameter adjustment when controlling them, making online real-time adjustment difficult and resulting in low control accuracy. Fuzzy control design and adjustment of fuzzy rules are relatively complex, requiring professional knowledge, and its control performance is limited by expert knowledge. Summary of the Invention

[0005] This invention addresses the problem that the design of fuzzy controllers is highly dependent on human experience. It proposes an Adaptive Neural Fuzzy Inference System (ANFIS) that improves fuzzy control using neural networks, and replaces expensive proportional valves with inexpensive, high-speed switching valves. This strategy combines the learning ability of neural networks with the reasoning ability of fuzzy logic to effectively handle the complex nonlinear characteristics of pneumatic pressure servo systems, thereby achieving precise pressure control.

[0006] The specific technical solution is as follows:

[0007] The ANFIS system for precise pressure control of high-speed switching valves includes two two-position two-way high-speed switching valves, namely an inlet valve and an outlet valve.

[0008] The air inlet of the air inlet valve is connected to the air source through a pneumatic triplet, and the air outlet of the air inlet valve is connected to a fixed cavity.

[0009] The air inlet of the vent valve is connected to the fixed cavity, and the air outlet of the vent valve is connected to the atmospheric environment.

[0010] The intake valve and the exhaust valve are respectively connected to the control system;

[0011] The pneumatic triplet is equipped with a first pressure sensor, and the fixed cavity is equipped with a second pressure sensor. The first pressure sensor and the second pressure sensor are connected to the control system.

[0012] The control system includes a pressure error calculation module, a differentiator, a fuzzy controller, an ANFIS controller, and a three-mode switching controller.

[0013] The PWM signal output terminal of the ANFIS controller is connected to the signal input terminals of the intake valve and the exhaust valve.

[0014] The fuzzy controller determines the PWM duty cycle in the current state based on the error between the system target pressure and the actual pressure, and the part of the fuzzy controller with good control effect is used as the training data of ANFIS.

[0015] The ANFIS controller trains the input data e and de provided by the fuzzy controller, outputs the PWM duty cycle, and then gives a control signal to control the opening and closing of the intake valve and the exhaust valve. It then combines with the three-mode switching controller to precisely control the pressure of the fixed cavity.

[0016] The control method for an ANFIS system used for precise pressure control of high-speed switching valves includes the following steps:

[0017] Step 1: Measure the actual pressure P in the fixed cavity, and then determine the desired pressure P. r The pressure error e is calculated from the difference between the actual pressure P and the pressure error e, and the rate of change of pressure error de is calculated using a differentiator.

[0018] Step 2: The fuzzy controller transforms the pressure error e and the pressure error change rate de to the range of the universe of discourse and performs fuzzy processing to change the precise input quantity into a fuzzy value. E and EC are the pressure error and the pressure error change rate, respectively, and are represented by the corresponding fuzzy linguistic values. At the same time, the universe of discourse range and fuzzy linguistic value of the fuzzy controller output D are set, where D is the PWM duty cycle.

[0019] In step 2, the universes of discourse for E and EC are divided into 7 fuzzy levels: negative large NB, negative medium NM, negative small NS, zero ZO, positive small PS, positive medium PM, and positive large PB. For D, 5 fuzzy subsets are used for coverage: NM, NS, ZO, PS, and PM.

[0020] Step 3: Set up a fuzzy rule table, perform fuzzy inference calculations on the fuzzy controller output D, pressure error E, and pressure error change rate EC to obtain the fuzzy value of the fuzzy controller output D;

[0021] Step 4: Sharpen the output D of the fuzzy controller obtained by fuzzy inference to obtain the sharpened value within the universe of discourse;

[0022] Step 5: Use the better controlled parts of the fuzzy controller as training data for ANFIS. The ANFIS structure consists of five parts: fuzzification, rule base, fuzzy inference, defuzzification, and parameter adjustment. The fuzzy rule base is generated based on the input, i.e., the error and the rate of change of the error.

[0023] Step 6: Training the samples, (1) Forward propagation: Input data into the fuzzy rule base and calculate the activation degree of each rule. Use the activation degree as a weight and perform a weighted average to obtain the fuzzy output. Pass the fuzzy output and the input of the neural network to the neural network to obtain the output of the neural network. (2) Calculate the error: Calculate the error between the actual output and the model output. (3) Backward propagation: Use the backward propagation algorithm to update the weights and biases of the neural network and update the parameters of the fuzzy rule base to reduce the error.

[0024] Step 7: Repeat the training, defining the termination condition as 50 repetitions.

[0025] Step 8: Apply the trained model to the pressure servo system to generate duty cycle prediction or control signals.

[0026] Step 9: By combining with three-mode switching, a joint control strategy is proposed. Then, digital signals are output from a digital signal board to control the on / off state of two high-speed switching valves, achieving pressurization and depressurization of the fixed cavity, ultimately maintaining the cavity pressure at the desired pressure P. r .

[0027] High-speed switching valves are typically controlled using pulse width modulation (PWM) technology. PWM-controlled switching valves have continuous output characteristics similar to proportional valves. However, due to the hysteresis and valve inertia of high-speed switching valves, PWM technology suffers from significant control dead zones, nonlinear regions, and saturation regions. Mode switching can effectively avoid the effects of control dead zones. Furthermore, combining PWM with an adaptive neural fuzzy inference system (ANFIS) improves pressure control accuracy and provides a new approach to the control of pneumatic servo systems.

[0028] The cavity pressure control method proposed in this invention, based on high-speed switching valve mode switching and ANFIS, has significant advantages over traditional pressure control systems using fuzzy controllers and PID control. It possesses strong adaptability and learning capabilities, enabling online adjustment of the control strategy to adapt to changes in system dynamics and uncertainties, effectively improving system performance and pressure tracking accuracy. The use of ANFIS allows the system to better adapt to the nonlinear and time-varying characteristics of the cavity pressure system, improving the robustness and performance of the controller. The combined control strategy allows for slow pressurization and depressurization, thereby achieving fine-tuned pressure regulation and reducing overshoot. Attached Figure Description

[0029] Figure 1 This is a system block diagram of the present invention;

[0030] Figure 2 This is a logic block diagram of the overall control of the present invention;

[0031] Figure 3 This is the ANFIS architecture of the present invention. Detailed Implementation

[0032] The specific technical solution of the present invention will be described in conjunction with the accompanying drawings.

[0033] like Figure 1 As shown, the ANFIS system for precise pressure control of high-speed switching valves includes two two-position two-way high-speed switching valves, namely inlet valve 1 and vent valve 2.

[0034] The air inlet of the air inlet valve 1 is connected to the air source 4 through the pneumatic triplet 3, and the air outlet of the air inlet valve 1 is connected to the fixed cavity 5.

[0035] The air inlet of the vent valve 2 is connected to the fixed cavity 5, and the air outlet of the vent valve 2 is connected to the atmospheric environment.

[0036] Intake valve 1 and exhaust valve 2 are respectively connected to control system 6;

[0037] The pneumatic triplet 3 is equipped with a first pressure sensor 8, and the fixed cavity 5 is equipped with a second pressure sensor 7. The first pressure sensor 8 and the second pressure sensor 7 are connected to the control system 6.

[0038] The control system 6 includes a pressure error calculation module, a differentiator, a fuzzy controller, an ANFIS controller, and a three-mode switching controller;

[0039] The PWM signal output terminal of the ANFIS controller is connected to the signal input terminals of the intake valve 1 and the exhaust valve 2.

[0040] The fuzzy controller determines the PWM duty cycle in the current state based on the error between the system target pressure and the actual pressure, and the part of the fuzzy controller with good control effect is used as the training data of ANFIS.

[0041] The ANFIS controller trains the input data e and de provided by the fuzzy controller, outputs the PWM duty cycle, and then the ANFIS controller gives a control signal to control the opening and closing of the intake valve 1 and the exhaust valve 2. Then, it combines with the three-mode switching controller to precisely control the pressure of the fixed cavity 5.

[0042] The control method for an ANFIS system used for precise pressure control of high-speed switching valves includes the following steps:

[0043] Step 1: Measure the actual pressure P in the fixed cavity 5, and then calculate the pressure P based on the desired pressure. r The pressure error e is calculated from the difference between the actual pressure P and the pressure error e, and the rate of change of pressure error de is calculated using a differentiator.

[0044] Step 2: The fuzzy controller transforms the pressure error e and the pressure error change rate de to the range of the universe of discourse and performs fuzzy processing to change the precise input quantity into a fuzzy value. E and EC are the pressure error and the pressure error change rate, respectively, and are represented by the corresponding fuzzy linguistic values. At the same time, the universe of discourse range and fuzzy linguistic value of the fuzzy controller output D are set, where D is the PWM duty cycle.

[0045] In step 2, the universes of discourse for E and EC are divided into 7 fuzzy levels: negative large NB, negative medium NM, negative small NS, zero ZO, positive small PS, positive medium PM, and positive large PB. For D, 5 fuzzy subsets are used for coverage: NM, NS, ZO, PS, and PM.

[0046] Step 3: Set up a fuzzy rule table, perform fuzzy inference calculations on the fuzzy controller output D, pressure error E, and pressure error change rate EC to obtain the fuzzy value of the fuzzy controller output D;

[0047] In step 3, the fuzzy quantities need to be processed using the inference rules in fuzzy logic; the fuzzy rules between the fuzzy controller output D, pressure error E, and pressure error change rate EC are shown in Table 1 below:

[0048] Table 1 Fuzzy Rules

[0049]

[0050] Step 4: Sharpen the output D of the fuzzy controller obtained by fuzzy inference to obtain the sharpened value within the universe of discourse;

[0051] Step 5: The better-controlled portions of the fuzzy controller are used as training data for ANFIS. The ANFIS structure consists of five parts: fuzzification, rule base, fuzzy inference, defuzzification, and parameter adjustment. The fuzzy rule base is generated based on the input (error) and the rate of change of the error. The ANFIS architecture is as follows: Figure 3 As shown.

[0052] Step 6: Training the samples, (1) Forward propagation: Input data into the fuzzy rule base and calculate the activation degree of each rule. Use the activation degree as a weight and perform a weighted average to obtain the fuzzy output. Pass the fuzzy output and the input of the neural network to the neural network to obtain the output of the neural network. (2) Calculate the error: Calculate the error between the actual output and the model output. (3) Backward propagation: Use the backward propagation algorithm to update the weights and biases of the neural network and update the parameters of the fuzzy rule base to reduce the error.

[0053] Step 7: Repeat the training, defining the termination condition as 50 repetitions.

[0054] Step 8: Apply the trained model to the pressure servo system to generate duty cycle prediction or control signals.

[0055] Step 9: By combining with three-mode switching, a joint control strategy is proposed. Then, the digital signal output card controls the on / off state of two high-speed switching valves to achieve pressurization and depressurization of the fixed cavity 5, ultimately maintaining the cavity pressure at the desired pressure P. r .

Claims

1. An ANFIS system for precise pressure control of high-speed switching valves, characterized in that, It includes two two-position two-way high-speed switching valves, namely an intake valve (1) and an exhaust valve (2). The intake port of the intake valve (1) is connected to the air source (4) through the pneumatic triplet (3), and the exhaust port of the intake valve (1) is connected to the fixed cavity (5); The air inlet of the vent valve (2) is connected to the fixed cavity (5), and the exhaust port of the vent valve (2) is connected to the atmospheric environment; The intake valve (1) and the exhaust valve (2) are respectively connected to the control system (6); The pneumatic triplet (3) is equipped with a first pressure sensor (8), and the fixed cavity (5) is equipped with a second pressure sensor (7). The first pressure sensor (8) and the second pressure sensor (7) are connected to the control system (6). The control system (6) includes a pressure error calculation module, a differentiator, a fuzzy controller, an ANFIS controller, and a three-mode switching controller; The PWM signal output terminal of the ANFIS controller is connected to the signal input terminals of the intake valve (1) and the exhaust valve (2); The fuzzy controller is based on the system's desired pressure P. r The pressure error e is calculated from the difference between the actual pressure P and the pressure error e, which determines the PWM duty cycle under the current state. The part of the fuzzy controller with good control effect is used as the training data for ANFIS. After training the pressure error e provided by the fuzzy controller and the pressure error change rate de calculated by the differentiator, the ANFIS controller outputs the PWM duty cycle. Then, the ANFIS controller gives a control signal to control the opening and closing of the intake valve (1) and the exhaust valve (2). Then, it is combined with the three-mode switching controller to accurately control the pressure of the fixed cavity (5).

2. A control method for an ANFIS system used for precise pressure control of high-speed switching valves, characterized in that, The ANFIS system for precise pressure control of high-speed switching valves as described in claim 1, wherein the method comprises the following steps: Step 1: Measure the actual pressure P of the fixed cavity (5), and determine the desired pressure P. r The pressure error e is calculated from the difference between the actual pressure P and the pressure error e, and the rate of change of pressure error de is calculated using a differentiator. Step 2: The fuzzy controller transforms the pressure error e and the pressure error change rate de to the range of the universe of discourse and performs fuzzy processing to change the precise input quantity into a fuzzy value. E and EC are the pressure error and the pressure error change rate, respectively, and are represented by the corresponding fuzzy linguistic values. At the same time, the universe of discourse range and fuzzy linguistic value of the fuzzy controller output D are set, where D is the PWM duty cycle. Step 3: Set up a fuzzy rule table, perform fuzzy inference calculations on the fuzzy controller output D, pressure error E, and pressure error change rate EC to obtain the fuzzy value of the fuzzy controller output D; Step 4: Sharpen the output D of the fuzzy controller obtained by fuzzy inference to obtain the sharpened value within the universe of discourse; Step 5: Use the parts controlled better by the fuzzy controller as training data for ANFIS. The ANFIS structure consists of five parts: fuzzification, rule base, fuzzy inference, defuzzification, and parameter adjustment. The fuzzy rule base is generated based on the input, i.e., the error and the rate of change of the error. Step 6: Training the samples, (1) Forward propagation: Input data into the fuzzy rule base and calculate the activation of each rule; use the activation as weight and perform a weighted average to obtain the fuzzy output; pass the fuzzy output and the input of the neural network to the neural network to obtain the output of the neural network; (2) Calculate the error: Calculate the error between the actual output and the model output; (3) Backward propagation: Use the backward propagation algorithm to update the weights and biases of the neural network and update the parameters of the fuzzy rule base to reduce the error; Step 7: Repeat the training, defining the termination condition as 50 repetitions; Step 8: Apply the trained model to the pressure servo system to generate duty cycle prediction or control signals; Step 9: By combining with three-mode switching, a joint control strategy is proposed. Then, the digital signal output of the digital signal board controls the on / off state of two high-speed switching valves to realize the pressurization and depressurization of the fixed cavity (5), and finally maintain the cavity pressure at the desired pressure P. r .

3. The control method of the ANFIS system for precise pressure control of high-speed switching valves according to claim 2, characterized in that, In step 2, the universes of discourse of E and EC are divided into 7 fuzzy levels, namely negative large NB, negative medium NM, negative small NS, zero ZO, positive small PS, positive medium PM and positive large PB; D is covered by 5 fuzzy subsets, namely NM, NS, ZO, PS and PM.

Images