Inspection Quality Charge Control Method Based on Improved Fruit Fly Optimization Dynamic Sliding Mode

Abstract

A method for controlling the inspection mass charge based on improved fruit fly optimization dynamic sliding mode relates to the field of inspection mass charge control. The present invention is to solve the problem that when the charge management system performs charge control, the requirements for the controller are different in different application scenarios, resulting in that the fixed controller parameters cannot maintain the optimal state for a long time. In the present invention, the parameters for constructing the dynamic sliding mode control law are taken as the positions of fruit flies, the improved fruit fly optimization algorithm is used to calculate the optimal individuals of fruit flies, and the positions of the optimal individuals of fruit flies are taken as the optimal parameters at time t; the tracking error at time t is substituted into the dynamic sliding mode control law at time t to calculate the control signal of the charge control system, and the charge control system of the inspection mass charge is controlled by using the control signal of the charge control system.

Description

Technical Field

[0001] The present invention belongs to the technical field of inspection mass charge control. Background Art

[0002] High-precision inertial sensors are scientific instruments for high-precision acceleration measurement in a microgravity environment, with advantages such as high resolution, and can be applied to various types of space science missions, including satellite gravity measurement for global gravity field inversion, space gravitational wave detection, space equivalence principle test, space Newton's inverse square law test, etc. In space gravitational wave detection, the test mass (TM) in a high-precision inertial sensor serves as an inertial reference benchmark. When a gravitational wave signal passes by, the relative position between the two test masses changes, and this change can be obtained through inter-satellite laser interferometry to detect the gravitational wave signal.

[0003] Due to the influence of external environmental factors such as solar high-energy particles and cosmic rays, the surface of the test mass in the space inertial sensor inevitably accumulates charges continuously, causing the electrostatic force between the test mass and the surrounding conductors, thereby introducing acceleration noise and affecting space gravitational wave detection. Therefore, it is necessary to manage the surface charges of the test mass, and a charge management system (CMS) is designed in the space inertial sensor system for charge control. When the charge management system works, it is affected by external factors (environmental temperature, pressure, external pollution, etc.) for a long time, resulting in the aging of ultraviolet light-emitting diode (UV LED) devices, changes in the surface characteristics (such as quantum yield) of the metal coating of the inertial sensor, external disturbances, and the problem of time delay in charge measurement in actual engineering, which affects the performance indicators of the CMS and even causes system instability in severe cases. Based on this problem, dynamic sliding mode control (DSMC) with the advantages of fast response speed and strong robustness is applied to the charge control of the charge management system. However, the requirements for the controller are different in different application scenarios, resulting in the fixed controller parameters being unable to maintain the optimal state for a long time. Summary of the Invention

[0004] The present invention is to solve the problem that when the charge management system performs charge control, the requirements for the controller are different in different application scenarios, resulting in the fixed controller parameters being unable to maintain the optimal state for a long time. Now, a test mass charge control method based on improved fruit fly optimization dynamic sliding mode is provided.

[0005] The test mass charge control method based on improved fruit fly optimization dynamic sliding mode includes:

[0006] Using the expected output value V of the test mass charge control system r And the measured value V of the surface potential of the test mass at time t m_t The difference is defined as the tracking error e at time t t ;

[0007] Take the parameters used to construct the dynamic sliding mode control law as the positions of fruit flies, randomly generate the initial positions of the fruit fly population, and use the improved fruit fly optimization algorithm to calculate the optimal individuals of the fruit flies. Take the positions of the optimal individuals of the fruit flies as the optimal parameters [c1, c2, ε, k] at time t;

[0008] Construct the dynamic sliding mode control law at time t using the optimal parameters [c1, c2, ε, k]

[0009]

[0010] Among them, sgn() is the sign function, and σ t is the second-order switching function and has and are the first-order derivatives of s t and e t respectively, and the intermediate variable A p is the actual decay rate of the environmental charging rate, and B p is the decay coefficient of the quantum yield and the ultraviolet light power. is the first-order derivative of B p , T1 is the charge measurement time delay, and are the first-order, second-order, and third-order derivatives of V r respectively, is the first-order derivative of V m_t , and V p_t is the output value of the test mass charge control system at time t;

[0011] Substitute the tracking error e t at time t into the dynamic sliding mode control law at time t to calculate the control signal u t of the charge control system at time t, and use the control signal u t of the charge control system at time t to perform charge control on the test mass charge control system.

[0012] Furthermore, the above calculation of the optimal individuals of fruit flies using the improved fruit fly optimization algorithm includes:

[0013] S1: Use the Logistic map to map the initial positions X_axis of the fruit fly population to chaotic variables CX_axis;

[0014] S2: Perform chaotic operations on the chaotic variable CX_axis using the Logistic map to obtain the intermediate variable CX;

[0015] S3: Map the intermediate variable CX to the interval of the initial positions of the fruit fly population to obtain the positions X'_axis of the fruit fly population after chaotic operation;

[0016] S4: Use the adaptive strategy to adjust the adaptive search radius of the fruit fly individuals at iteration number k to obtain the positions of each fruit fly in the fruit fly population;

[0017] S5: Substitute each fruit fly position into the dynamic sliding mode control law at time t - 1 to obtain the control signal u t ′ -1 of the charge control system corresponding to the fruit fly position. Substitute the control signal u t ′ -1 corresponding to each fruit fly position into the objective function to calculate the taste concentration value at the position of each fruit fly;

[0018] The expression of the objective function J is:

[0019]

[0020] where w1, w2, and w3 are all weight coefficients and w3 >> w1, eV t is the overshoot and eV t = V p_t - V p_t-1 , V p_t and V p_t-1 are the output values of the test mass charge control system at times t and t - 1 respectively;

[0021] S6: Select the fruit fly corresponding to the minimum taste concentration as the high-quality fruit fly at iteration number k. Judge whether k is equal to the maximum iteration number. If so, take the high-quality fruit fly at iteration number k as the optimal fruit fly individual. Otherwise, make k = k + 1, and then return to S4.

[0022] Further, the above-mentioned use of the adaptive strategy to adjust the adaptive search radius of the fruit fly individuals at iteration number k includes:

[0023] Adjust the adaptive search radius through the following formula:

[0024]

[0025] where w k is the adaptive search radius of the fruit fly individuals at iteration number k, w max and w min are the maximum and minimum search radii respectively, and k max is the maximum iteration number.

[0026] Further, mapping the initial position X_axis of the fruit fly population to the chaotic variable CX_axis using the Logistic mapping includes:

[0027] When the range of the initial position X_axis of the fruit fly population is [a, b], the chaotic variable CX_axis is obtained using the following formula:

[0028]

[0029] Further, performing a chaotic operation on the chaotic variable CX_axis using the Logistic mapping to obtain the intermediate variable CX includes:

[0030] The intermediate variable CX is obtained using the following formula:

[0031] CX = α·CX_axis·(1 - CX_axis),

[0032] where α is a control parameter. When α ∈ (3.56, 4], the closer the value of α is to 4, the more intense the chaotic state.

[0033] Further, mapping the intermediate variable CX to the interval of the initial position of the fruit fly population to obtain the position X′_axis of the fruit fly population after the chaotic operation includes:

[0034] The position X′_axis of the fruit fly population after the chaotic operation is obtained using the following formula:

[0035] X′_axis = a + CX(b - a).

[0036] Further, using an adaptive strategy to adjust the adaptive search radius of the fruit fly individual at the k-th iteration to obtain the positions of each fruit fly in the fruit fly population includes:

[0037] The position X of the i-th fruit fly is obtained using the following formula i :

[0038] X i = X′_axis + w k ·Value·rand(),

[0039] where Value is the single flight distance and rand() is a random value.

[0040] The inspection quality charge control method based on the improved fruit fly optimization dynamic sliding mode of the present invention has the following beneficial effects:

[0041] (1) The dynamic sliding mode controller can track different step input commands in a short time, with the advantage of fast response speed; it can also eliminate the influence of complex factors such as parameter time-variation, cope with time delay and external disturbances, has strong robustness, and improves the stability of the system.

[0042] (2) The present invention proposes an improved fruit fly optimization algorithm, which uses chaotic mapping to initialize the fruit fly population, increases the traversability of the population, makes the initialization range more uniform; has an adaptive search radius, balances the local optimization and global optimization capabilities; improves the optimization speed and accuracy, and avoids the defect of falling into local optimum.

[0043] In summary, the present invention uses the improved fruit fly optimization algorithm to further improve the efficiency and accuracy of the parameter tuning of the dynamic sliding mode controller. Description of the Drawings

[0044] Figure 1 It is the control block diagram of the inspection quality charge control method based on the improved fruit fly optimization dynamic sliding mode;

[0045] Figure 2 It is the flow chart of optimizing the parameters of the dynamic sliding mode controller by the improved fruit fly optimization algorithm;

[0046] Figure 3 It is the DSMC dynamic response curve diagram of three parameter tuning methods;

[0047] Figure 4 It is the TM potential output tracking error curve diagram of three parameter tuning methods when the TM potential is controlled at -10 mV. Detailed Implementation Manner

[0048] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention. It should be noted that, without conflict, the embodiments in the present invention and the features in the embodiments can be combined with each other.

[0049] Detailed Implementation Manner 1. Refer to Figures 1 to 4 Specifically describe this implementation manner. The space inertial sensor inspection quality charge control method based on the improved fruit fly optimization dynamic sliding mode described in this implementation manner includes:

[0050] S1: Design a dynamic sliding mode controller. Drawing on the idea of high-order sliding mode, using the first-order sliding surface s as the system state point, design a controller using a new second-order sliding surface σ, and prove its stability using the Lyapunov function. Specifically as follows:

[0051] S11: Give a charge control model and a charge measurement time delay model for designing a dynamic sliding mode controller.

[0052] Most space applications require multi-year on-orbit missions, so the aging problem of UV LED devices must be considered. According to the experimental data of UV LED life tests. The light power attenuation law of UV LED is obtained as follows:

[0053]

[0054] where p a (t) represents the light power of the UV LED, and α1, β1, and β2 are all fitting parameters.

[0055] Due to changes in temperature, pressure, and air pollution, the surface properties (such as quantum yield) of the conductor metal coating are different on different surfaces and even at different times on the same surface. Therefore, the change in quantum yield should be considered. According to the relevant experimental data, the fitting curve of the quantum yield is obtained as follows:

[0056]

[0057] where QY C is the change law of the quantum yield, and Q0, α3, α4, β3, and β4 are all fitting parameters.

[0058] Combining the environmental charging rate with the CMS charge and discharge rate, the total discharge rate of TM in space can be obtained That is, the charge control model:

[0059]

[0060] where A p is the actual attenuation rate of the environmental charging rate, B p is the attenuation coefficient of the quantum yield and the ultraviolet light power, and B p = 0.2·p a ·QY C , u is the control signal of the charge control system, p a is the light power of the UV LED, V p is the surface potential of the TM, that is, the output value of the proof mass charge control system.

[0061] In aerospace engineering applications, many control systems have the characteristics of system communication and feedback measurement time delay (lag), resulting in an increase in system response time and, in severe cases, disrupting the stability of the system. The time delay in the charge management system mainly comes from the charge measurement link. Therefore, in this embodiment, the system communication time delay is ignored, and a first-order inertia link is introduced into the feedback link of the closed-loop system to simulate the effect of charge measurement time delay, that is, the charge measurement time delay model:

[0062]

[0063] where T1 is the time constant of the delay link, and V m is the measured value of the surface potential of TM after delay (feedback link), is the first derivative of V m .

[0064] S12: Controller Design

[0065] Taking V r as the expected output value of the charge control system for inspection, in order to minimize the surface potential of the inspection mass, the tracking error e is defined by the difference between the measured value V m of the surface potential of TM and the expected output value V r :

[0066] e = V r - V m (5).

[0067] The stability of the control system is closely related to the switching function (sliding mode surface s). To ensure the existence and reachability of the sliding mode, a first-order sliding mode surface s is designed as follows:

[0068]

[0069] where is the first derivative of e, and the parameter c1 > 0.

[0070] Adopting the idea of the high-order sliding mode surface in high-order sliding mode control, with the sliding mode surface s as the system state, a second-order switching function σ is designed:

[0071]

[0072] where is the first derivative of s, and the parameter c2 > 0.

[0073] The first derivative of the second-order switching function σ is:

[0074]

[0075] The method of using the exponential reaching law is adopted to accelerate the reaching process, enabling the system to reach the sliding surface in a relatively short time:

[0076]

[0077] where sgn() is the sign function, the parameter ε > 0, and the parameter k > 0.

[0078] Establish the connection between the exponential reaching law and the first derivative of the second-order switching function and design the control law based on this. Let Equation (8) be equal to Equation (9), then we have:

[0079]

[0080] Then design the dynamic sliding mode control law as:

[0081]

[0082] where

[0083]

[0084]

[0085] T1 is the charge measurement time delay.

[0086] S13: Stability proof

[0087] Select the following Lyapunov function and prove its stability:

[0088]

[0089] Take the derivative of the Lyapunov function, then we have:

[0090]

[0091] Substitute Equation (11) into Equation (15), then we have:

[0092]

[0093] From Equation (16), we can see that:

[0094] (1) When σ > 0, ε > 0, k > 0, sgn(σ) > 0, thus we can obtain

[0095] (2) When σ < 0, ε > 0, k > 0, sgn(σ) < 0, thus we can obtain

[0096] In summary, according to Lyapunov stability theory: when the energy function V > 0 and certain conditions are met, the derivative of the Lyapunov function The system satisfies stability.

[0097] S2: Use an improved chaotic adaptive fruit fly optimization algorithm to automatically optimize the parameters of the dynamic sliding mode controller online, so that the controller performance remains optimal for a long time, and improve the system stability and robustness.

[0098] S21: Improvement strategy

[0099] (1) The chaotic mapping optimization algorithm (COA), as a direct search algorithm, its basic idea is to use the ergodicity of chaos for optimization. The COA algorithm uses the law of chaotic variables themselves for optimization search. Compared with random search, it can evenly traverse the entire search space, enabling the search algorithm to jump out of local extreme points.

[0100] The Logistic mapping, as a non-linear chaotic equation, is a typical model for studying complex systems such as dynamic systems and chaos. Its function expression is:

[0101] x k+1 = αx k (1 - x k ) (17),

[0102] where k is the number of iterations, x k ∈(0, 1), α is the control parameter. When α ∈ (3.56, 4], the chaotic system determined by equation (17) enters the chaotic state. The closer the value of α is to 4, the more intense the chaotic state is.

[0103] (2) Adaptive step size strategy:

[0104] Although the standard fruit fly optimization algorithm (FOA) has good convergence speed and convergence accuracy, the fixed search radius is not conducive to balancing the global optimization ability and local optimization ability. By adopting the strategy of adaptive search radius, in the early stage of the search algorithm, a larger search radius can improve the convergence speed and increase the global optimization ability. In the later stage of the search algorithm, a smaller search radius can improve the convergence accuracy and increase the local optimization ability, making it easier to jump out of local optimal values and avoid premature convergence.

[0105] The search radius is adaptively adjusted using the following formula:

[0106]

[0107] where w k is the adaptive search radius of the fruit fly individual when the number of iterations is k, w max and w min are the maximum and minimum search radii respectively, kmax is the maximum number of iterations.

[0108] S22: Objective function design

[0109] During the process of using the chaotic adaptive fruit fly optimization algorithm (GIFOA) to optimize the parameters of the dynamic sliding controller, the smell concentration value Smell is used as the update evaluation value, and the smell concentration function F(S i ) is the bridge connecting the intelligent optimization algorithm GIFOA and DSMC. Therefore, it is crucial to design a suitable objective function (also known as the smell concentration function).

[0110] Generally, when evaluating the performance of a control system, the dynamic performance (overshoot, settling time, rise time, etc.) and steady-state performance of the system need to be comprehensively considered. Therefore, the integral of the product of the simulation time t and the absolute value of the tracking error e (ITAE) is selected as one item of the objective function J.

[0111]

[0112] To prevent the control output from being too large and increase energy consumption, the square term u of the control signal of the charge control system is added 2 as one item of the objective function, so that the objective function is jointly controlled by the tracking error e, the simulation time t, and the control signal u of the charge control system.

[0113] The objective function J is as follows:

[0114]

[0115] Although the ITAE performance index function can enable the system to obtain a faster response time, there will be a large overshoot in the dynamic response process. To address this problem, this embodiment adopts a penalty mechanism, adds weight coefficients to each item of the objective function, and adds the overshoot eV t as one item. The final objective function is:

[0116]

[0117] Among them, w1, w2, and w3 are the weight coefficients of each item. Generally, w3 >> w1, and w1 = 0.999, w2 = 0.001, w3 = 100.

[0118] eV t is the overshoot and has eV t = V p_t - V p_t-1 V p_t and V p_t-1 are the output values of the test mass charge control system at times t and t - 1 respectively. When eV tWhen < 0, the set of controller parameters found by GIFOA does not meet the optimization conditions. By amplifying the overshoot term through the penalty factor w3, the value of the objective function becomes too large, thus eliminating this set of controller parameters.

[0119] S23: Steps of Optimizing DSMC by GIFOA (Improved Fruit Fly Optimization Algorithm)

[0120] S231: The general implementation steps of GIFOA are as follows:

[0121] Step 1: Set the size Sizepop of the fruit fly population, the maximum number of iterations Maxgen, and the spatial dimension, and randomly initialize the positions of the fruit fly population:

[0122] Init X_axis(23),

[0123] where X_axis is a random value within the initial range interval.

[0124] Step 2: Use the Logistic mapping to map the initial positions X_axis of the fruit fly population to chaotic variables CX_axis. The range of X_axis is [a, b], and the values of a and b are determined according to the range of the controller parameters in the actual system. The value range of the chaotic variable CX_axis is (0, 1), that is, map X_axis from [a, b] to (0, 1). The mapping expression is as follows:

[0125]

[0126] Step 3: Perform chaotic operations on the chaotic variable CX_axis using the Logistic mapping reference formula (17) to obtain the variable CX. The specific chaotic operation formula is as follows:

[0127] CX = α·CX_axis·(1 - CX_axis) (25).

[0128] Step 4: Map the variable CX obtained after the chaotic operation to an ordinary variable, that is, map CX from the interval (0, 1) to the interval [a, b]. The specific operation formula after the chaotic operation is as follows:

[0129] X′_axis = a + CX(b - a) (26).

[0130] Step 5: Use the adaptive strategy (formula 18) to adjust the adaptive random flight direction and distance of the fruit fly individuals to search for food by smell. The position X of the i-th fruit fly i :

[0131] X i = X′_axis + w k·Value·rand() (27),

[0132] where Value is the single - flight distance, rand() is a random value, and w k is the adaptive search radius of the fruit - fly individual when the iteration number is k.

[0133] Step 6: Adaptively update the fruit - fly position X i Calculate the smell concentration judgment value S i :

[0134] S i = X i (28).

[0135] Step 7: Substitute the smell concentration judgment value S i in formula (28) into the objective function of smell concentration judgment, and calculate the smell concentration value Smell of the position where the i - th fruit - fly individual is located i :

[0136] Smell i = Function(S i ) (29).

[0137] Specifically, make S i = X i = [c 1i , c 2i , ε i , k i , and obtain the control signal u 1i , c 2i , ε i , k i corresponding to the charge control system, and substitute u i into the objective function of smell concentration judgment, and the obtained result is the smell concentration value Smell i . i .

[0138] Step 8: Find the minimum value of the objective function, and find the fruit - fly individual with the best smell concentration in the fruit - fly population according to the smell concentration value:

[0139] [bestSmell bestIndex] = min(Smell i ) (30),

[0140] where bestSmell is the optimal smell concentration value of the contemporary fruit - fly population, and bestIndex is the serial number of the fruit - fly with the best smell concentration in the population.

[0141] Step 9: Save the optimal flavor concentration value and its corresponding position information (fruit fly serial number), and the fruit fly population flies towards this direction relying on its sensitive vision.

[0142]

[0143] Among them, Smellbest is the optimal flavor concentration value, which is used as the final parameter [c1, c2, ε, k] for constructing the dynamic sliding mode control law

[0144] Step 10: Perform iterative optimization. Determine whether the number of iterations reaches the maximum number of iterations. If it does not meet the maximum number of iterations, repeat Steps 5 to 9 again; otherwise, the loop ends.

[0145] S232: Execute steps according to the designed GIFOA program. Use the objective function in the GIFOA program to assign values to the parameters in the Simulink model and run Simulink to optimize the parameters online. The process of improving the fruit fly optimization algorithm to optimize the parameters of the dynamic sliding mode controller is as Figure 2 shown.

[0146] Simulation experiment

[0147] Determine the parameters and ranges:

[0148] Environmental charging rate A p is set to 3.13×10 -7 s -1 , which is equivalent to a charging rate of +20e s -1 . B p = 0.2·QY C ·p a , QY C is the quantum yield change characteristic, Q0 is 0.206, α3 is 0.274, α4 is 0.525, β3 is -394.013, β4 is -4334.395. α1 is 0.239, β1 is 2.876, β2 is 3493.317. The charge measurement time delay T1 = 1. The charge measurement noise is -0.05mV to +0.05mV. The external unknown perturbation d is -0.1mV to +0.1mV. The initial potential of TM is -15mV. The spatial dimension Dim of the fruit fly population is 4. Set the fruit fly population size Sizepop to 30, the maximum number of iterations Maxgen to 50, the limit range of the optimized parameters is c1 ∈ [0, 3], c2 ∈ [0, 3], k ∈ [0, 10], ε ∈ [0, 10], the chaotic system selects the Logistic map, a is 0.001, b is 10, α is 3.9, and the maximum step size w max in the adaptive search step size is 5, and the minimum step size w min ​is 0.01, and the weight coefficients of the objective function are set as w1 = 0.999, w2 = 0.001, and w3 = 100.

[0149] To verify the performance of the standard FOA and GIFOA in optimizing the parameters of the dynamic sliding mode controller, in this embodiment, the controller parameters obtained by the two intelligent optimization algorithms are simulated and verified. The manually adjusted controller parameters are: c1 = 0.05, c2 = 1, ε = 1, k = 1; the controller parameters after FOA optimization are: c1 = 0.0514, c2 = 1.031, ε = 1.1312, k = 1.1352; the controller parameters after GIFOA optimization are: c1 = 0.0612, c2 = 1.1731, ε = 4.1372, k = 1.5341.

[0150] In this embodiment, the simulation experiments are carried out using the controller parameters of FOA, GIFOA, and manual adjustment, and the results are as Figure 3 and Figure 4 shown. Figure 3 is the closed-loop response of the dynamic sliding mode control. The whole simulation includes 3 cycles. In each cycle, the input step command gradually increases from -12 mV to -10 mV with a step of 2 mV (each step lasts for 1000 s). The initial value is -15 mV, and the simulation time is from 0 s to 9000 s. The results show that under manual adjustment, the rise time is 44 s, the adjustment time is 60 s, and there is no overshoot. Under FOA, the rise time is 43 s, the adjustment time is 58 s, and there is no overshoot. Under GIFOA, the rise time is 36 s, the adjustment time is 49 s, and there is no overshoot. The dynamic performance of the dynamic sliding mode controller parameters optimized by GIFOA is better than that of FOA and manual adjustment.

[0151] Figure 4 shows the tracking errors of the TM potential of manual adjustment, FOA, and GIFOA when the TM potential is controlled at -10 mV after 6000 s. Figure (a) is the composite diagram of the three cases. It can be seen from Figure (a) that the tracking error after GIFOA stabilizes is smaller, which is better than the tracking errors of manual adjustment and FOA. Figure (b) shows that the tracking error under manual adjustment is from -0.03 mV to +0.04 mV. Figure (c) shows that the tracking error under FOA is from -0.03 mV to +0.04 mV. Figure (d) shows the tracking error under GIFOA, and its value is from -0.02 mV to +0.02 mV.

[0152] In summary, the dynamic sliding mode controller parameters optimized by GIFOA are better than those of FOA and manual adjustment in both dynamic performance and steady-state performance, which verifies the effectiveness of the designed GIFOA in finding the optimal solution.

[0153] Specific Embodiment 2. The method for inspecting the quality charge control of a space inertial sensor based on an improved fruit fly optimization dynamic sliding mode includes:

[0154] Step 1: Use the expected output value V of the inspection mass charge control system r and the measured surface potential value V of the inspection mass at time t m_t to define the tracking error e at time t t = V r - V m_t .

[0155] Step 2: Use the parameters for constructing the dynamic sliding mode control law as the fruit fly positions, randomly generate the initial positions of the fruit fly population, and use the improved fruit fly optimization algorithm to calculate the optimal fruit fly individuals. Specifically, it includes the following steps:

[0156] S1: When the range of the initial position X_axis of the fruit fly population is [a, b], use the Logistic mapping to map the initial position X_axis of the fruit fly population to the chaotic variable CX_axis:

[0157]

[0158] S2: Perform chaotic operations on the chaotic variable CX_axis using the Logistic mapping to obtain the intermediate variable CX:

[0159] CX = α·CX_axis·(1 - CX_axis),

[0160] where α is the control parameter. When α ∈ (3.56, 4], the closer the value of α is to 4, the more intense the chaotic state is.

[0161] S3: Map the intermediate variable CX to the interval of the initial position of the fruit fly population to obtain the position X'_axis of the fruit fly population after chaotic operations:

[0162] X'_axis = a + CX(b - a).

[0163] S4: Use the adaptive strategy to adjust the adaptive search radius of the fruit fly individual at iteration k:

[0164]

[0165] where w k is the adaptive search radius of the fruit fly individual at iteration k, w max and w min are the maximum and minimum search radii respectively, and k max is the maximum number of iterations.

[0166] Furthermore, calculate the position of each fruit fly in the fruit fly population according to the adaptive search radius:

[0167] X i = X′_axis + w k ·Value·rand(),

[0168] where X i is the position of the i-th fruit fly, Value is the single flight distance, and rand() is a random value.

[0169] S5: Substitute each fruit fly position into the dynamic sliding mode control law at time t - 1 to obtain the control signal u t ′ -1 of the charge control system corresponding to the fruit fly position. Substitute the control signal u t ′ -1 of the charge control system corresponding to each fruit fly position into the objective function, and calculate the taste concentration value at the position of each fruit fly:

[0170]

[0171] where w1, w2, and w3 are all weight coefficients and w3 >> w1, eV t is the overshoot and eV t = V p_t - V p_t-1 , V p_t and V p_t-1 are the output values of the inspection mass charge control system at times t and t - 1 respectively;

[0172] S6: Select the fruit fly corresponding to the minimum taste concentration as the high-quality fruit fly at the iteration number k. Judge whether k is equal to the maximum iteration number. If so, take the high-quality fruit fly at the iteration number k as the optimal fruit fly individual. Otherwise, make k = k + 1, and then return to S4.

[0173] Step Three: Take the position of the optimal fruit fly individual as the optimal parameter [c1, c2, ε, k] at time t.

[0174] Step Four: Use the optimal parameter [c1, c2, ε, k] to construct the dynamic sliding mode control law at time t

[0175]

[0176] where sgn() is the sign function, σ t is the second-order switching function and and are s t and e tThe first derivative, intermediate variable A p is the actual attenuation rate of the environmental charging rate, B p is the attenuation coefficient of the quantum yield and the ultraviolet light power, is the first derivative of B p T1 is the charge measurement time delay, and are the first, second, and third derivatives of V r respectively, is the first derivative of V m_t V p_t is the output value of the test mass charge control system at time t.

[0177] Step Five: Substitute the tracking error e t at time t into the dynamic sliding mode control law at time t to calculate the control signal u t of the charge control system at time t, and use the control signal u t of the charge control system at time t to perform charge control on the test mass charge control system.

Claims

1. A test quality charge control method based on improved fruit fly optimization dynamic sliding mode, characterized in that The expected output value V of the inspection mass charge control system is adopted r The surface potential measurement value V of the inspection mass at time t m_t The difference is defined as the tracking error e at time t t ; the parameters for constructing the dynamic sliding mode control law are taken as the fruit fly positions, the initial positions of the fruit fly population are randomly generated, and the improved fruit fly optimization algorithm is used to calculate the optimal fruit fly individual, and the position of the optimal fruit fly individual is taken as the optimal parameters [c1, c2, ε, k] at time t; Construct the dynamic sliding mode control law at time t by using the optimal parameters [c1, c2, ε, k] where sgn() is the sign function, and σ t is a second-order switching function and has and are the first-order derivatives of s t and e t respectively, and the intermediate variable A p is the actual decay rate of the environmental charging rate, B p is the decay coefficient of the quantum yield and the ultraviolet light power, is the first-order derivative of B p T1 is the charge measurement time delay, and are the first-order, second-order, and third-order derivatives of V r respectively, is the first-order derivative of V m_t V p_t is the output value of the inspection mass charge control system at time t; Substitute the tracking error e at time t t into the dynamic sliding mode control law at time t to calculate the control signal u of the charge control system at time t t and use the control signal u of the charge control system at time t t to perform charge control on the test mass charge control system.

2. The inspection quality charge control method based on improved fruit fly optimization dynamic sliding mode according to claim 1, characterized in that, The calculation of the optimal fruit fly individual using the improved fruit fly optimization algorithm includes: S1: Use the Logistic map to map the initial position X_axis of the fruit fly population to the chaotic variable CX_axis; S2: Perform chaotic operation on the chaotic variable CX_axis using the Logistic map to obtain the intermediate variable CX; S3: Map the intermediate variable CX to the interval of the initial position of the fruit fly population to obtain the position X′_axis of the fruit fly population after chaotic operation; S4: Use the adaptive strategy to adjust the adaptive search radius of the fruit fly individual at iteration k to obtain the positions of each fruit fly in the fruit fly population; S5: Substitute each fruit fly position into the dynamic sliding mode control law at time t-1 to obtain the control signal u' of the charge control system corresponding to the fruit fly position t-1 , and substitute the control signal u' of the charge control system corresponding to each fruit fly position t-1 into the objective function to calculate the taste concentration value at the position where each fruit fly is located; The expression of the objective function J is: wherein, w1, w2, and w3 are all weighting coefficients and w3 >> w1, eV t is the overshoot amount and eV t = V p_t - V p_t-1 , V p_t and V p_t-1 are the output values of the test mass charge control system at times t and t - 1, respectively; S6: Select the fruit fly corresponding to the minimum taste concentration as the high-quality fruit fly at iteration k, determine whether k is equal to the maximum number of iterations, if so, take the high-quality fruit fly at iteration k as the optimal fruit fly individual, otherwise make k = k + 1, and then return to S4.

3. The inspection quality charge control method based on improved fruit fly optimization dynamic sliding mode according to claim 2, characterized in that, The use of the adaptive strategy to adjust the adaptive search radius of the fruit fly individual at iteration k includes: Adjust the adaptive search radius by the following formula: Among them, w k is the adaptive search radius of the fruit fly individual when the iteration number is k, w max and w min are the maximum and minimum search radii respectively, and k max is the maximum number of iterations.

4. The inspection quality charge control method based on improved fruit fly optimization dynamic sliding mode according to claim 3, characterized in that, The use of the Logistic map to map the initial position X_axis of the fruit fly population to the chaotic variable CX_axis includes: When the range of the initial position X_axis of the fruit fly population is [a, b], the chaotic variable CX_axis is obtained by the following formula:

5. The inspection quality charge control method based on improved fruit fly optimization dynamic sliding mode according to claim 4, wherein The use of the Logistic map to perform chaotic operation on the chaotic variable CX_axis to obtain the intermediate variable CX includes: The intermediate variable CX is obtained by the following formula: CX = α·CX_axis·(1 - CX_axis), where α is the control parameter, and when α ∈ (3.56, 4], the closer the value of α is to 4, the more intense the chaotic state.

6. The inspection quality charge control method based on improved fruit fly optimization dynamic sliding mode according to claim 5, characterized in that The mapping of the intermediate variable CX to the interval of the initial position of the fruit fly population to obtain the position X′_axis of the fruit fly population after chaotic operation includes: The position X′_axis of the fruit fly population after chaotic operation is obtained by the following formula: X′_axis = a + CX(b - a).

7. The inspection quality charge control method based on improved fruit fly optimization dynamic sliding mode according to claim 6, characterized in that The use of the adaptive strategy to adjust the adaptive search radius of the fruit fly individual at iteration k to obtain the positions of each fruit fly in the fruit fly population includes: The position X of the i-th fruit fly is obtained by the following formula i :[[]]END]] X i = X′_axis + w k ·Value·rand(), where Value is the single flight distance and rand() is a random value.

Images