Multi-stage roots pump control method and system based on artificial intelligence

Abstract

The invention relates to the technical field of multi-stage roots pumps, and discloses a multi-stage roots pump control method and system based on artificial intelligence, and the method comprises the steps that real-time operation data signals of a multi-stage roots pump are collected through a Hall sensor, a pressure sensor and an angular velocity sensor, and the data signals comprise air outlet pressure, magnetic field intensity and rotating speed signals; carrying out dynamic feature extraction on the data signal, and carrying out normalized model processing to generate standardized feature data; and constructing a data twinborn model based on the actual working condition data, and migrating the standardized feature data to the data twinborn model by using a transfer learning mechanism. By constructing the data twinning model, standardized feature data is quickly migrated to the actual working condition model, the training period is shortened, and the generalization ability of the model is improved; through lightweight edge deployment, real-time data acquisition and prediction deviation calculation are realized, and the phenomenon of local overload or efficiency loss is ensured not to occur.

Description

Technical Field

[0001] The present invention relates to the technical field of multi-stage Roots pumps, and particularly to a control method and system for multi-stage Roots pumps based on artificial intelligence. Background Art

[0002] As a core device in a vacuum system, the operating efficiency and stability of a multi-stage Roots pump directly affect the performance of industrial production. Traditional control methods for multi-stage Roots pumps mostly rely on fixed-parameter PID control or empirical rules, and it is difficult to cope with the pressure coupling problems in each individual pumping chamber under complex dynamic working conditions.

[0003] A prior patent discloses a variable phase angle multi-stage Roots vacuum pump and its working method (publication number CN116517828A). The variable phase angle multi-stage Roots vacuum pump includes: a pump casing, a right rotor 1, and a left rotor 2. The right rotor 1 and the left rotor 2 rotate in opposite directions around their respective rotation axes. In the technology disclosed in this patent, since the pressure between each level is not balanced during the operation of the multi-stage Roots pump and it is difficult to perform dynamic pressure regulation on each individual pumping chamber, this easily causes local overload or efficiency loss. Summary of the Invention

[0004] The present invention provides a control method and system for multi-stage Roots pumps based on artificial intelligence to solve the existing technical problems, and solves the problem of different pressures in the Roots pumps at each level.

[0005] To solve the above technical problems, according to one aspect of the present invention, more specifically, a control method for multi-stage Roots pumps based on artificial intelligence includes:

[0006] S1. Collect real-time operation data signals of the multi-stage Roots pump through Hall sensors, pressure sensors, and angular velocity sensors. The data signals include outlet pressure, magnetic field intensity, and rotational speed signals; perform dynamic feature extraction on the data signals, and process and generate standardized feature data through a normalization model;

[0007] S2. Construct a data twin model based on actual working condition data, and use a transfer learning mechanism to transfer the standardized feature data to the data twin model;

[0008] S3. Taking the maximization of the energy efficiency ratio and the minimization of the deviation amount as a joint optimization goal, train reinforcement learning using the deep deterministic policy gradient algorithm to generate an optimal decoupling control strategy for the inter-stage pressure distribution of the multi-stage Roots pump;

[0009] S4. According to the decoupling control strategy, adjust the rotational speed of each level motor in real time through a PID controller;

[0010] S5. Perform lightweight edge deployment based on the data twin model, collect control effect data in real time and calculate the prediction deviation; when the prediction deviation exceeds the preset threshold, trigger the incremental update mechanism of the single Roots pump control model.

[0011] Furthermore, the normalization model is used to perform min-max normalization processing on each sensor signal separately and scale it to the interval [0, 1]. There is:

[0012]

[0013] In the formula, c represents the standard data after normalization processing of the i-th type of sensor; c i represents the original data of the i-th type of sensor; c min , c max respectively represent the minimum and maximum theoretical values detected by the i-th type of sensor.

[0014] Furthermore, the data twin model calculates the prediction deviation based on the deviation of the outlet pressure under the cycle time and the deviation of the outlet pressure change cycle from the magnetic field strength change cycle. There is:

[0015]

[0016] In the formula, g represents the deviation of the multistage Roots pump calculated by the data twin model; m represents the deviation of the outlet pressure within the cycle time; a represents the deviation of the outlet pressure change cycle from the magnetic field strength change cycle.

[0017] Furthermore, when g ≥ 78%, it means that the prediction deviation exceeds the preset threshold, and then triggers the incremental update mechanism of the single Roots pump control model;

[0018] When g < 78%, it means that there is no need to trigger the model incremental update mechanism.

[0019] Furthermore, in step S3, the reward function of the deep deterministic policy gradient algorithm is defined as a weighted inverse function of the energy efficiency ratio and the deviation.

[0020] Furthermore, the specific process of the PID controller in step S4 is as follows:

[0021] 1). Compare the rotation speed signal fed back by the angular velocity sensor with the input signal for controlling the motor rotation speed in the single Roots pump control model;

[0022] 2). Compensate the deviation between the input signal after comparison and the actual output into the input signal.

[0023] Furthermore, the decoupling control strategy is used to perform customized control according to the states of different single Roots pump control models in the multistage Roots pump.

[0024] An artificial intelligence-based multi-stage Roots pump control system, comprising:

[0025] A multi-source data perception module for collecting real-time operation data signals of the multi-stage Roots pump, wherein the data signals include outlet pressure, magnetic field intensity and rotational speed signals;

[0026] A data twin modeling module for constructing a data twin model according to actual working condition data and migrating standardized feature data to the data twin model by using a transfer learning mechanism;

[0027] A hierarchical reinforcement learning module for training reinforcement learning by using a deep deterministic policy gradient algorithm to generate an optimal decoupling control strategy for the inter-stage pressure distribution of the multi-stage Roots pump;

[0028] An inter-stage decoupling control module for adjusting the rotational speed of each level of motor in real time through a PID controller;

[0029] A lightweight edge deployment module for collecting control effect data in real time and calculating prediction deviation; when the prediction deviation exceeds a preset threshold, triggering an incremental update mechanism for a single Roots pump control model.

[0030] The artificial intelligence-based multi-stage Roots pump control method and system provided by the present invention, compared with the prior art, have the following effects:

[0031] 1. By using the deep deterministic policy gradient algorithm, the present invention takes the maximization of the energy efficiency ratio and the minimization of the deviation amount as the joint optimization objectives to generate an optimal inter-stage pressure distribution strategy, significantly reducing the coupling effect and improving the overall energy efficiency by about 15%-0%; and the decoupling control strategy supports customized regulation of the multi-stage Roots pump to adapt to the operation state differences of different levels of equipment.

[0032] 2. By constructing a data twin model, the present invention quickly migrates standardized feature data to the actual working condition model, shortening the training cycle and improving the model generalization ability; through lightweight edge deployment, real-time data collection and prediction deviation calculation are realized to ensure that there will be no phenomenon of local overload or efficiency loss.

[0033] 3. By triggering incremental update based on the prediction deviation threshold, the present invention optimizes only the Roots pump model of the abnormal level, reducing the consumption of computing resources and improving the system robustness at the same time, and this mechanism can improve the model update efficiency by 40% and reduce the risk of failure shutdown.

[0034] 4. By performing dynamic feature extraction and min-max normalization processing on the Hall sensor, pressure sensor and angular velocity signal, the present invention eliminates the dimension difference and improves the data consistency, providing high-quality input for subsequent algorithms. Description of the Drawings

[0035] Figure 1 This is the flowchart of the control method in the present invention;

[0036] Figure 2 This is the air extraction principle of the Roots pump in the present invention;

[0037] Figure 3 This is the schematic diagram of the distribution of sensors in the present invention;

[0038] Figure 4 This is the schematic diagram of the periodic change of magnetic field and pressure in the present invention;

[0039] Figure 5 This is the schematic diagram of the deviation amount g and the deviation amount m in the present invention;

[0040] Figure 6 This is the schematic diagram of the deviation amount g and the deviation amount a in the present invention;

[0041] Figure 7 This is the model diagram of the deviation amount g, the deviation amount m, and the deviation amount a in the present invention.

[0042] In the figure: 1. Hall sensor; 2. Pressure sensor; 3. Magnetic rotor. Detailed implementation manners

[0043] To make the technical solution of the present invention clearer, the following further describes the present invention in detail with reference to the accompanying drawings and specific embodiments.

[0044] As Figure 2 , 3 , and as shown in Figure 4, the present invention provides a multi-stage Roots pump control method or system based on artificial intelligence, which is applied to a multi-stage Roots pump. The multi-stage Roots pump is composed of multiple (generally an even number) Roots pump control models (as Figure 3 shown, two Roots pump control models are connected in series to form a multi-stage Roots pump. And two magnetic rotors 3 form a single air extraction chamber).

[0045] Among them, each single Roots pump control model includes two magnetic rotors 3 (the magnetic rotors are permanent magnets with different magnetic poles at both ends), a Hall sensor 1, and a pressure sensor 2 located at the exhaust port. The pressure sensor 2 is used to monitor the air pressure at the exhaust port.

[0046] Among them, the air extraction principle of a single Roots pump is as follows:

[0047] a. As Figure 2 shown in (1) of it, at this time, the magnetic field intensity detected by the Hall sensor 1 is as Figure 4 shown in a(1) of it, and the pressure detected by the pressure sensor 2 is as Figure 4 shown in b(1) of it;

[0048] b. As shown in Figure 2 (2), at this time, the magnetic field intensity detected by the Hall sensor 1 is as shown in Figure 4 a(2) in it, and the pressure detected by the pressure sensor 2 is as shown in Figure 4 b(2) in it;

[0049] c. As shown in Figure 2 (3), at this time, the magnetic field intensity detected by the Hall sensor 1 is as shown in Figure 4 a(3) in it, and the pressure detected by the pressure sensor 2 is as shown in Figure 4 b(3) in it;

[0050] d. As shown in Figure 2 (3), at this time, the magnetic field intensity detected by the Hall sensor 1 is as shown in Figure 4 a(3) in it, and the pressure detected by the pressure sensor 2 is as shown in Figure 4 b(3) in it;

[0051] e. As shown in Figure 2 (4), at this time, the magnetic field intensity detected by the Hall sensor 1 is as shown in Figure 4 a(4) in it, and the pressure detected by the pressure sensor 2 is as shown in Figure 4 b(4) in it;

[0052] f. As shown in Figure 2 (5), at this time, the magnetic field intensity detected by the Hall sensor 1 is as shown in Figure 4 a(5) in it, and the pressure detected by the pressure sensor 2 is as shown in Figure 4 b(5) in it.

[0053] The multi-stage Roots pump control system based on artificial intelligence includes: a multi-source data perception module for collecting real-time operation data signals of the multi-stage Roots pump, where the data signals include outlet pressure, magnetic field intensity, and rotational speed signals; a data twin modeling module for constructing a data twin model according to actual working condition data and migrating standardized feature data to the data twin model using a transfer learning mechanism; a hierarchical reinforcement learning module for training reinforcement learning using the deep deterministic policy gradient algorithm to generate an optimal decoupling control strategy for the inter-stage pressure distribution of the multi-stage Roots pump; an inter-stage decoupling control module for adjusting the rotational speed of each level of motor in real time through a PID controller; a lightweight edge deployment module for collecting control effect data in real time and calculating the prediction deviation; when the prediction deviation exceeds a preset threshold, triggering an incremental update mechanism for a single Roots pump control model.

[0054] Embodiment 1

[0055] Collect the real-time operation data signals of the multi-stage Roots pump through Hall sensors, pressure sensors and angular velocity sensors. The data signals include outlet pressure, magnetic field intensity and rotational speed signals; extract the dynamic characteristics of the data signals, and process them through a normalization model to generate standardized characteristic data;

[0056] The normalization model is used to perform min-max normalization processing on each sensor signal separately and scale it to the interval [0, 1]. There is:

[0057]

[0058] In the formula, c represents the standard data after normalization processing of the i-th type of sensor; c i represents the original data of the i-th type of sensor; c min and c max represent the minimum and maximum theoretical values detected by the i-th type of sensor respectively.

[0059] Among them, the deviation of the outlet pressure within the cycle time is calculated. Calculate the standard data after normalization processing of the pressure sensor within the first cycle (such as Figure 4 the pressure at time (2) in b of min ). The pressure of the pressure sensor within the first cycle is taken as c1 = 0.8 (MPa), and the minimum and maximum theoretical values detected by the pressure sensor are c max = 0.1 (MPa) and c

[0060]

[0061] Calculate the standard data after normalization processing of the pressure sensor within the second cycle (such as Figure 4 the pressure at time (4) in b of

[0062]

[0063] From the above calculations, it can be known that the deviation of the outlet pressure between the first cycle time and the second time cycle is m = 70% - 60% = 10%.

[0064] Among them, the deviation between the outlet pressure change cycle and the magnetic field intensity change cycle is calculated. Calculate the standard data after normalization processing of the magnetic field intensity change cycle (such as Figure 4 the time of cycle (1)-(5) in a of min ). The magnetic field intensity change cycle time is taken as c3 = 8 (ms), and the minimum and maximum theoretical values of the magnetic field change cycle are c max= 15 (ms). Then there is:

[0065]

[0066] For the pressure change period (such as Figure 4 the time of the period (1)-(5) in b in the figure), calculate the standard data after normalization processing. The pressure change period time is taken as c4 = 12 (ms), and the minimum and maximum theoretical values of the magnetic field change period are c min = 5 (ms), c max = 16 (ms). Then there is:

[0067]

[0068] From the above calculations, it can be known that the deviation between the pressure change period and the magnetic field strength change period is a = 63% - 46.1% = 16.9%.

[0069] Example 2

[0070] Build a digital twin model based on the actual working condition data, and use the transfer learning mechanism to transfer the standardized feature data to the digital twin model. Based on the digital twin model, perform lightweight edge deployment, collect control effect data in real time and calculate the prediction deviation; when the prediction deviation exceeds the preset threshold, trigger the incremental update mechanism of the single Roots pump control model. By building a digital twin model and combining the transfer learning mechanism, the standardized feature data can be quickly transferred to the actual working condition model, shortening the training cycle and improving the model generalization ability; through lightweight edge deployment, real-time data collection and prediction deviation calculation are realized to ensure low-latency control response.

[0071] The digital twin model calculates the prediction deviation according to the deviation of the outlet pressure under the cycle time and the deviation between the outlet pressure change period and the magnetic field strength change period. There is:

[0072]

[0073] In the formula, g represents the deviation of the multi-stage Roots pump calculated by the digital twin model; m represents the deviation of the outlet pressure within the cycle time; a represents the deviation between the outlet pressure change period and the magnetic field strength change period.

[0074] The above formula is for the empirical analysis to solve the problem that the pressures of the Roots pumps in each layer are different during the control process of the multi-stage Roots pump, and the data and the characteristic relationships between the data obtained therefrom are the external manifestations of the empirical formula. Then the reasoning process is as follows:

[0075] 1), Fit the formula for calculating the deviation g of the multi-stage Roots pump calculated by the digital twin model.

[0076] Among them, the data twin model calculates the deviation g of the multi-stage Roots pump, which can be expressed by the proportion of the pressure borne by the valves in the sample multi-stage Roots pump.

[0077] For example, if the data of 100 multi-stage Roots pump cabinet samples are collected and the pressure of a multi-stage Roots pump valve is manually evaluated (such as Figure 3 The valve at the middle connection, when the valve is subjected to a greater pressure, the deviation between the two Roots pumps is greater) exceeds the data in the other 50 samples, then it means that the data twin model calculates the deviation of the multi-stage Roots pump g = 50%.

[0078] Other data are calculated and fitted using data collected by the device itself.

[0079] 2) Establish a mathematical model for the relationship between the deviation g and the deviation m (such as Figure 5 As shown in the figure, the red dots are the distribution of 100 samples collected), then:

[0080]

[0081] In the above formula 1, k represents an empirical constant for adjusting the sensitivity of the above model.

[0082] 3) Establish a mathematical model for the relationship between the deviation g and the deviation a (such as Figure 6 As shown in the figure, the red dots are the distribution of 100 samples collected), then:

[0083]

[0084] In the above formula 2, k represents an empirical constant for adjusting the sensitivity of the above model.

[0085] 4) Establish a mathematical model for the relationship between the deviation g, deviation m, and deviation a (such as Figure 7 As shown, the model can be used to know that the deviation m and the deviation a are positively correlated), and combined with the characteristic relationship of the above formula 1 and formula 2, we have:

[0086] g=(Formula 1)×(Formula 2);

[0087] According to the above derivation, the condition coefficient g for the energy storage cabinet to call the backup power module is:

[0088]

[0089] Among them, the deviation of the outlet pressure in the first cycle and the second cycle is m=10%, and the deviation of the pressure change period and the magnetic field intensity change period is a=16.9%. Then:

[0090]

[0091] From the above calculations, it can be known that the deviation of the multi-stage Roots pump calculated by the digital twin model is g = 83.9%, which means that the prediction deviation exceeds the preset threshold, thus triggering the incremental update mechanism of the single Roots pump control model. And there is:

[0092] Table 1 Partial Detection Parameters and Control Relationships

[0093]

[0094]

[0095] From the data in Table 1 above, it can be known that when the implementation data tends to infinity, the value of the deviation g can be used as a boundary to determine whether to perform incremental update. When g ≥ 78%, it means that the prediction deviation exceeds the preset threshold, thus triggering the incremental update mechanism of the single Roots pump control model (which means that the rotational speed of the magnetic rotor 3 in the single Roots pump to be detected needs to be increased); when g < 78%, it means that there is no need to trigger the model incremental update mechanism.

[0096] Example 3

[0097] Taking the maximization of the energy efficiency ratio and the minimization of the deviation as the joint optimization objectives, the deep deterministic policy gradient algorithm is used to train the reinforcement learning to generate the optimal decoupling control strategy for the inter-stage pressure distribution of the multi-stage Roots pump; the decoupling control strategy is used to perform customized control according to the states of different single Roots pump control models in the multi-stage Roots pump. The reward function of the deep deterministic policy gradient algorithm is defined as the weighted inverse ratio function of the energy efficiency ratio and the deviation. Based on the deep deterministic policy gradient (DDPG) algorithm, taking the maximization of the energy efficiency ratio and the minimization of the deviation as the joint optimization objectives, the optimal inter-stage pressure distribution strategy is generated, significantly reducing the coupling effect and improving the overall energy efficiency.

[0098] Example 4

[0099] According to the decoupling control strategy, the rotational speeds of the motors at each level are adjusted in real time through a PID controller; the specific process of the PID controller is as follows:

[0100] 1) Compare the rotational speed signal fed back by the angular velocity sensor with the input signal for controlling the rotational speed of the motor in the single Roots pump control model;

[0101] 2) Compensate the deviation between the input signal after comparison and the actual output into the input signal.

[0102] The above-described embodiments merely represent several implementation manners of the present invention. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the patent for the present invention. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all fall within the protection scope of the present invention. Therefore, the protection scope of the patent for the present invention shall be subject to the appended claims.

Claims

1. An artificial intelligence-based control method for a multistage Roots pump, characterized in that, Including: S1. Collect the real-time operation data signals of the multistage Roots pump through Hall sensors, pressure sensors and angular velocity sensors. The data signals include outlet pressure, magnetic field intensity and rotational speed signals; extract the dynamic characteristics of the data signals, and process them through a normalization model to generate standardized feature data; S2. Build a data twin model based on the actual working condition data, and use the transfer learning mechanism to transfer the standardized feature data to the data twin model; S3. Taking the maximum energy efficiency ratio and the minimum deviation as the joint optimization objectives, use the deep deterministic policy gradient algorithm to train reinforcement learning to generate the optimal decoupling control strategy for the inter-stage pressure distribution of the multistage Roots pump; S4. According to the decoupling control strategy, adjust the rotational speed of each level motor in real time through a PID controller; S5. Perform lightweight edge deployment based on the data twin model, collect the control effect data in real time and calculate the prediction deviation; when the prediction deviation exceeds the preset threshold, trigger the incremental update mechanism of the single Roots pump control model.

2. The multi-stage Roots pump control method based on artificial intelligence according to claim 1, characterized in that: The normalization model is used to perform min-max normalization processing on each sensor signal separately, and scale it to the interval [0, 1]. There is: where c represents the standard data after normalization processing of the i-th type of sensor; c i represents the original data of the i-th type of sensor; c min , c max respectively represent the minimum and maximum theoretical values detected by the i-th type of sensor.

3. The control method for a multi-stage Roots pump based on artificial intelligence according to claim 1, characterized in that: The data twin model calculates the prediction deviation according to the deviation of the outlet pressure under the cycle time and the deviation of the outlet pressure change cycle and the magnetic field intensity change cycle. There is: In the formula, g represents the deviation of the multistage Roots pump calculated by the data twin model; m represents the deviation of the outlet pressure within the cycle time; a represents the deviation of the outlet pressure change cycle and the magnetic field intensity change cycle.

4. The method for controlling a multistage Roots pump based on artificial intelligence according to claim 3, wherein: When g≥78%, it means that the prediction deviation exceeds the preset threshold, and then trigger the incremental update mechanism of the single Roots pump control model; When g<78%, it means that there is no need to trigger the model incremental update mechanism.

5. The multi-stage Roots pump control method based on artificial intelligence according to claim 1, characterized in that: In the step S3, the reward function of the deep deterministic policy gradient algorithm is defined as the weighted inverse function of the energy efficiency ratio and the deviation.

6. The control method of a multi-stage Roots pump based on artificial intelligence according to claim 1, characterized in that: The specific process of the PID controller in the step S4 is as follows: 1). Compare the rotational speed signal fed back by the angular velocity sensor with the input signal for controlling the motor rotational speed in the single Roots pump control model; 2). Compensate the deviation of the input signal after comparison with the actual output into the input signal.

7. The multi-stage Roots pump control method based on artificial intelligence according to claim 1, characterized in that: The decoupling control strategy is used to perform customized control according to the states of different single Roots pump control models in the multistage Roots pump.

8. The multi-stage Roots pump control system based on artificial intelligence is characterized in that Applied to the control method described in any one of claims 1-7, the control system includes: A multi-source data perception module for collecting the real-time operation data signals of the multistage Roots pump. The data signals include outlet pressure, magnetic field intensity and rotational speed signals; A data twin modeling module for building a data twin model based on the actual working condition data, and using the transfer learning mechanism to transfer the standardized feature data to the data twin model; A hierarchical reinforcement learning module that uses the deep deterministic policy gradient algorithm to train reinforcement learning to generate the optimal decoupling control strategy for the inter-stage pressure distribution of the multistage Roots pump; An inter-stage decoupling control module that adjusts the rotational speed of each level motor in real time through a PID controller; A lightweight edge deployment module that collects control effect data in real time and calculates prediction deviations; when the prediction deviation exceeds a preset threshold, it triggers an incremental update mechanism for a single Roots pump control model.

Images