An online intelligent optimization method for control parameters of a digital accelerometer servo control system

Abstract

The present invention discloses an online intelligent optimization method for control parameters of a digital accelerometer servo control system, including: boundary testing under specified performance conditions of various control parameters of the controller; integrating various performance indicators of the system and setting the system fitness function; setting the range of control parameters to be optimized, and initializing the population in combination with the system processor performance; using the symbiotic optimization algorithm to perform online intelligent optimization of the control parameters based on the fitness function value; when the symbiotic optimization algorithm meets the termination condition, the control parameters after comprehensive performance indicators can be obtained, and the optimized parameters are directly assigned to the controller to complete the online intelligent optimization of the control parameters of the digital accelerometer servo control system. The present invention can be applied to digital quartz flexible accelerometers, and is suitable for typical single-loop PID control systems that meet the digital quartz flexible accelerometer model.

Description

Technical Field

[0001] The invention relates to the technical field of digital accelerometer control, and in particular to an online intelligent optimization method for control parameters of a digital accelerometer servo control system. Background Art

[0002] The digital accelerometer includes a meter and a digital servo control circuit. The servo circuit collects the meter capacitance and transmits it to the digital controller through AD conversion, and then applies feedback torque to the meter through DA conversion and the drive circuit to form a closed loop. Compared with traditional analog accelerometers, the control parameter debugging of digital accelerometers can be directly debugged through software programming. With the large-scale production and application of digital accelerometers, higher requirements are put forward for the parameter debugging and optimization of the closed-loop control system of digital accelerometers.

[0003] In previous project designs, the control parameters (proportional coefficient, integral coefficient, differential coefficient) of the digital accelerometer closed-loop system PID controller can be debugged by relying on the control rate of the PID control algorithm itself or empirical formulas for engineering application parameter tuning to achieve system control. However, without a certain theoretical foundation and debugging experience, the debugging and optimization of control parameters will not only greatly increase the workload of on-site personnel, but also increase the risk of unstable operation of the digital accelerometer closed-loop system.

[0004] In addition, this debugging method cannot adjust parameters for clear system indicators. In order to obtain better control parameters that meet various performance indicators of the system, the debugging personnel are often required to have rich debugging experience, and it is impossible to obtain the optimal control parameters that have both dynamic and static performance indicators of the system.

[0005] The algorithms disclosed in the prior art for PID parameter optimization include particle swarm algorithm (such as Chinese patent application with publication number CN115236969A), improved sparrow algorithm (such as Chinese patent application with publication number CN114397807A), PSO-SOA fusion algorithm (such as Chinese patent application with publication number CN110308649A), dragonfly algorithm (such as Chinese patent application with publication number CN107703751A), etc.

[0006] These prior art algorithms are not combined with the characteristics of the digital accelerometer servo control system itself. In order to improve the applicability of digital accelerometer products, it is very important to find the optimal control parameters corresponding to the system under different predetermined indicators. Summary of the invention

[0007] In view of the above technical problems and the shortcomings in the art, the present invention provides an online intelligent optimization method for control parameters of a digital accelerometer servo control system, which can optimize the control parameters online and obtain the optimal control parameters that weigh various performance indicators of the system while ensuring the hard indicators of the system, greatly reducing the ability and experience requirements of the debugging personnel, improving the debugging efficiency, and achieving the best performance under the specified indicator requirements of the system. The present invention has been successfully applied to digital quartz flexible accelerometers and is suitable for typical single-loop PID control systems that conform to the digital quartz flexible accelerometer model.

[0008] An online intelligent optimization method for control parameters of a digital accelerometer servo control system is used to optimize online and obtain the optimal PID control parameters that balance dynamic and static indicators while ensuring the hard indicators of the system, including:

[0009] Step 1: boundary test of various control parameters of the controller under specified performance conditions;

[0010] Step 2: integrate various system performance indicators and set the system fitness function;

[0011] Step 3: Set the range of control parameters to be optimized and initialize the population based on the system processor performance;

[0012] Step 4: Using the symbiotic optimization algorithm, the control parameters are optimized online intelligently based on the fitness function value;

[0013] Step 5: When the symbiotic optimization algorithm meets the termination conditions, the control parameters after comprehensive performance indicators can be obtained, and the optimized parameters are directly assigned to the controller to complete the online intelligent optimization of the control parameters of the digital accelerometer servo control system.

[0014] In step 1, the digital accelerometer head is placed at zero position, and the microprocessor is programmed to make the controller output positive and negative excitations of the same amplitude, and the excitation amplitude is gradually increased until the head capacitance reaches the boundary, and the head capacitance value is measured, that is, the capacitance boundary value, that is, the range of change of the system control object (capacitance);

[0015] Specify the system indicators to be debugged, including hard indicators and optimization indicators. Hard indicators include pulse response, and optimization indicators include system overshoot, rise time, and control object balance time.

[0016] Gradually increase the proportional coefficient until the controlled object shows equal amplitude oscillation. The proportional coefficient at this time is recorded as the upper boundary kp of the proportional coefficient. max , gradually reduce the proportional coefficient until the impulse response theory can reach the maximum amplitude just meeting the predetermined index, and record the proportional coefficient at this time as the lower boundary of the proportional coefficient kp min ;

[0017] The maximum amplitude of the pulse response theory is calculated based on the output value of the capacitor stable section and the capacitor boundary value under a 1g (1 gravitational acceleration) step. The formula is as follows:

[0018] pulse=cap / step_1g

[0019] Among them, pulse is the maximum amplitude that can be reached in pulse response theory, cap is the capacitance boundary value, and step_1g is the output value of the capacitance stable section under the 1g amplitude step response test;

[0020] On the basis of the lower boundary of the proportional coefficient, gradually increase the integral coefficient until the capacitance balance time meets the predetermined index under the 1g amplitude step response test, record the integral coefficient ki at this time, and take its 10% disturbance range as the parameter variation range, that is, ki min =0.9*ki,ki max =1.1*ki;

[0021] The lower boundary of the proportional coefficient and the critical integral coefficient ki min On this basis, gradually increase the differential coefficient until the system overshoot and rise time of the controller output response under the 1g amplitude step response test meet the predetermined indicators, and record the differential coefficient at this time as the lower boundary of the differential coefficient kd min ; On the upper boundary of the proportional coefficient and the critical integral coefficient ki max On this basis, gradually increase the differential coefficient until the system overshoot and rise time of the controller output response under the 1g amplitude step response test meet the predetermined indicators or the system oscillates, and record the differential coefficient at this time as the upper boundary of the differential coefficient kd max .

[0022] In step 2, when the control parameters have met the hard indicators of impulse response, the constructed fitness function includes dynamic indicators including system overshoot, rise time, and control object balance time, and static indicators of controller output noise under system zero input. According to the influence of the control rate of the control algorithm itself on each indicator, each indicator is weighted and combined to obtain a fitness function that meets actual needs. Under system zero input, the controller output noise is affected by the differential parameter. The larger the differential parameter, the greater the output noise. Therefore, in order to ensure the zero-position output accuracy of the accelerometer, it needs to be optimized.

[0023] In step 2, before optimizing the proportional coefficient, integral coefficient, and differential coefficient, the signal flow simplification rule is first used to simplify the digital accelerometer closed-loop system as an open-loop system with an input of u and a corresponding output of y after the given proportional coefficient, integral coefficient, and differential coefficient; u is the fixed input of the closed-loop system, and y is the response output of the closed-loop system; a 1g unit step and 0 input are performed on u, and the 1g unit step response time is consistent with the 0 input response time. The closed-loop system response output is collected, and the controller output overshoot, rise time rise_time, and balance time equi_time of the controlled object under the 1g unit step are calculated, and the indicator fitness evaluation is performed to obtain the system fitness function, and the formula is as follows:

[0024]

[0025] in:

[0026] u_d(k) is the closed-loop constant 1g unit step input sequence;

[0027] y_d(k) is the output sequence of the closed-loop system 1g unit step response;

[0028] u_s(k) is the closed-loop constant-value 0 input sequence;

[0029] y_s(k) is the closed-loop system zero-input response output sequence;

[0030] fc is the main control frequency of the closed-loop system;

[0031] α and β are the weighted coefficients of the indicators for the response to 1g unit step and 0 input in the fitness function, that is, the dynamic response and static response of the system are weighed. By changing α and β, the emphasis of the system on the dynamic response indicator and the static response indicator can be changed; γ is the weighted coefficient of the overshoot and rise time of the dynamic response performance in the fitness function. These two dynamic performance indicators are directly related to the proportional coefficient and differential coefficient in the controller. The increase of the proportional coefficient increases the overshoot and the rise time decreases. The increase of the differential coefficient reduces the overshoot and the rise time. That is, the weighted term corresponding to γ ​​is used to weigh the proportional action and the differential action. Here, the predetermined indicators are taken as an overshoot of 20% and a rise time of 50 control cycles; μ is the weighted coefficient of the balance time of the dynamic response performance control object in the fitness function, which is directly related to the integral coefficient. The integral coefficient increases and the balance time decreases. Here, the predetermined indicator is taken as the balance time of 1000 control cycles.

[0032] In step 3, the optimization range of the control parameters to be optimized is set according to the boundaries of the control parameters tested in step 1, and the appropriate population size is set in combination with the performance of the processor itself;

[0033] Parameter range to be optimized: Proportional coefficient kp∈(kp min ,kp max ); integral coefficient ki∈(ki min ,ki max ); differential coefficient kd∈(kd min ,kd max );

[0034] Population initialization:

[0035] X=Xmin+rand(eco_size,D)*(Xmax-Xmin)

[0036] Xmax=[kp max ,ki max ,kd max ]

[0037] Xmin=[kp min ,ki min ,kd min ]

[0038] Among them, X is the parameter population, eco_size is the initial population size, D is the number of parameters to be optimized, which is 3 here, and rand is a random number between 0 and 1.

[0039] In step 4, based on the decrease of the fitness function value constructed in step 2, and taking the range of each control parameter set in step 3 as the space to be optimized, the symbiotic optimization algorithm is used to optimize the control parameters through the steps of mutualism, partial symbiosis, and parasitism.

[0040] In step 4, the specific steps of the symbiotic biological optimization algorithm are:

[0041] Mutual Benefit Symbiosis Formula:

[0042]

[0043] Among them, X i , X j ∈X, represents the individuals of both symbiotic parties before updating, X i_new , X j_new represents the individuals of both symbiotic parties after the update, X best ∈X is the best individual in the current population, BF 1 , BF 2 is the mutual benefit coefficient, which is a random number of 1 or 2. MV is the average of the two individuals in the symbiosis. rand is a random number between 0 and 1. X best -MV*BF 1 , X best -MV*BF 2 In the population search process, it means that two individuals work together to get closer to the optimal individual;

[0044] Finally, compare the fitness function values ​​of the two individuals before and after the update, and retain the best one;

[0045] Partial symbiosis formula:

[0046] X i_new =X i +rand*(X best -X j )

[0047] Among them, rand is a random number between 0 and 1, and X best -X j It means that in the process of population search, it reflects the help provided by individual j to individual i, so that individual i gradually approaches the optimal individual;

[0048] Finally, compare the fitness function values ​​of individual i before and after the update, and retain the best one;

[0049] Parasitic formula:

[0050]

[0051] Among them, ceil is the rounding function, D is the number of parameters to be optimized, which is 3 here, rand is a random number between 0 and 1, and X dmax , X dmin Represents individual X i The maximum and minimum values ​​of the dth dimension range, d is 1, 2 or 3, X i_new (d) represents individual X i The change value of the d-th dimension parameter;

[0052] Finally, compare X i_new The fitness function value of another random individual is selected and retained.

[0053] In step five, when the fitness function value of the symbiotic optimization algorithm in step four is less than the specified value or the number of population iterations reaches the specified value, the parameter intelligent optimization is terminated. At this time, the population individual corresponding to the minimum value of the fitness function is the optimal PID control parameter obtained by weighing various system indicators.

[0054] When the fitness function J reaches the minimum value, it means that the proportional coefficient, integral coefficient and differential coefficient obtained by the symbiotic algorithm are the optimal control parameters for balancing the dynamic and static response indicators of the system. The digital accelerometer servo control system after online parameter optimization can be obtained by directly applying the optimal control parameters to the closed-loop system controller.

[0055] Compared with the prior art, the present invention has the following beneficial effects:

[0056] 1. The optimization method proposed in the present invention relies on the control characteristics of the PID control algorithm itself, and can test the upper and lower limits of each control parameter under the control parameters of the digital accelerometer servo control system to meet the hard indicators of the pulse response, thereby providing a parameter optimization range for subsequent performance optimization.

[0057] 2. The method proposed in the present invention can optimize the control parameters by comprehensively considering the dynamic response, system overshoot, rise time, control object balance time and static noise required by the system, thereby improving the comprehensive performance and applicability of the digital accelerometer system.

[0058] 3. The method proposed in the present invention adopts a symbiotic biological intelligent optimization algorithm. This method has the characteristics of no need to add parameters manually and completely relying on the system itself to find the best solution. In addition, since the algorithm has a variety of population screening steps with different mechanisms, it has the characteristics of large-scale search and local rapid convergence, which greatly improves the efficiency of parameter optimization.

[0059] 4. The method proposed in the present invention can write an automated program based on the flow chart to perform online parameter optimization, and the parameter optimization depends on the actual system, abandons the system model identification process, greatly simplifies the parameter optimization steps, improves the efficiency of control parameter optimization, and reduces the experience requirements and workload of on-site debugging personnel and the risk of unstable system operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 It is a flow chart of an online intelligent optimization method for control parameters of a digital accelerometer servo control system in a specific embodiment of the present invention;

[0061] Figure 2 This is a flow chart of the proportional coefficient boundary test in a specific implementation manner of the present invention;

[0062] Figure 3 This is a flow chart of the integral coefficient boundary test in a specific implementation manner of the present invention;

[0063] Figure 4 This is a flow chart of differential coefficient boundary test in a specific implementation manner of the present invention;

[0064] Figure 5 This is a flow chart of optimizing the control parameters using an intelligent algorithm in a specific implementation manner of the present invention. DETAILED DESCRIPTION

[0065] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and are not used to limit the scope of the present invention.

[0066] An online intelligent optimization method for control parameters of a digital accelerometer servo control system is used to optimize online to obtain the optimal PID control parameters that balance dynamic and static indicators while ensuring the hard indicators of the system, such as Figure 1 As shown, including:

[0067] Step 1: boundary test of various control parameters of the controller under specified performance conditions;

[0068] Step 2: integrate various system performance indicators and set the system fitness function;

[0069] Step 3: Set the range of control parameters to be optimized and initialize the population based on the system processor performance;

[0070] Step 4: Using the symbiotic optimization algorithm, the control parameters are optimized online intelligently based on the fitness function value;

[0071] Step 5: When the symbiotic optimization algorithm meets the termination conditions, the control parameters after comprehensive performance indicators can be obtained, and the optimized parameters are directly assigned to the controller to complete the online intelligent optimization of the control parameters of the digital accelerometer servo control system.

[0072] In step 1, the digital accelerometer head is placed at zero position, and the microprocessor is programmed to make the controller output positive and negative excitations of the same amplitude, and the excitation amplitude is gradually increased until the head capacitance reaches the boundary, and the head capacitance value is measured, that is, the capacitance boundary value, that is, the range of change of the system control object (capacitance);

[0073] Specify the system indicators to be debugged, including hard indicators and optimization indicators. Hard indicators include pulse response, and optimization indicators include system overshoot, rise time, and control object balance time.

[0074] like Figure 2 As shown, gradually increase the proportional coefficient until the controlled object shows equal amplitude oscillation. The proportional coefficient at this time is recorded as the upper boundary kp of the proportional coefficient. max , gradually reduce the proportional coefficient, and perform a 1g amplitude step response test until the pulse response theory can reach the maximum amplitude that just meets the predetermined index. The proportional coefficient at this time is recorded as the lower boundary of the proportional coefficient kp min ;

[0075] The maximum amplitude of the pulse response theory is calculated based on the output value of the capacitor stable section and the capacitor boundary value under a 1g (1 gravitational acceleration) step. The formula is as follows:

[0076] pulse=cap / step_1g

[0077] Among them, pulse is the maximum amplitude that can be reached in pulse response theory, cap is the capacitance boundary value, and step_1g is the output value of the capacitance stable section under the 1g amplitude step response test;

[0078] like Figure 3 As shown, based on the lower boundary of the proportional coefficient, the integral coefficient is gradually increased until the capacitance balance time meets the predetermined index under the 1g amplitude step response test, and the integral coefficient ki is recorded at this time, and its 10% disturbance range is taken as the parameter variation range, that is, ki min =0.9*ki,ki max =1.1*ki;

[0079] like Figure 4 As shown, the lower boundary of the proportional coefficient and the critical integral coefficient ki min On this basis, gradually increase the differential coefficient until the system overshoot and rise time of the controller output response under the 1g amplitude step response test meet the predetermined indicators, and record the differential coefficient at this time as the lower boundary of the differential coefficient kd min ; On the upper boundary of the proportional coefficient and the critical integral coefficient ki max On this basis, gradually increase the differential coefficient until the system overshoot and rise time of the controller output response under the 1g amplitude step response test meet the predetermined indicators or the system oscillates, and record the differential coefficient at this time as the upper boundary of the differential coefficient kd max .

[0080] In step 2, when the control parameters have met the hard indicators of impulse response, the constructed fitness function includes dynamic indicators including system overshoot, rise time, and control object balance time, and static indicators of controller output noise under system zero input. According to the influence of the control rate of the control algorithm itself on each indicator, each indicator is weighted and combined to obtain a fitness function that meets actual needs. Under system zero input, the controller output noise is affected by the differential parameter. The larger the differential parameter, the greater the output noise. Therefore, in order to ensure the zero-position output accuracy of the accelerometer, it needs to be optimized.

[0081] In step 2, before optimizing the proportional coefficient, integral coefficient, and differential coefficient, the signal flow simplification rule is first used to simplify the digital accelerometer closed-loop system as an open-loop system with an input of u and a corresponding output of y after the given proportional coefficient, integral coefficient, and differential coefficient; u is the fixed input of the closed-loop system, and y is the response output of the closed-loop system; a 1g unit step and 0 input are performed on u, and the 1g unit step response time is consistent with the 0 input response time. The closed-loop system response output is collected, and the controller output overshoot, rise time rise_time, and balance time equi_time of the controlled object under the 1g unit step are calculated, and the indicator fitness evaluation is performed to obtain the system fitness function, and the formula is as follows:

[0082]

[0083] in:

[0084] u_d(k) is the closed-loop constant 1g unit step input sequence;

[0085] y_d(k) is the output sequence of the closed-loop system 1g unit step response;

[0086] u_s(k) is the closed-loop constant-value 0 input sequence;

[0087] y_s(k) is the closed-loop system zero-input response output sequence;

[0088] fc is the main control frequency of the closed-loop system;

[0089] α and β are the weighted coefficients of the indicators for the response to 1g unit step and 0 input in the fitness function, that is, the dynamic response and static response of the system are weighed. By changing α and β, the emphasis of the system on the dynamic response indicator and the static response indicator can be changed; γ is the weighted coefficient of the overshoot and rise time of the dynamic response performance in the fitness function. These two dynamic performance indicators are directly related to the proportional coefficient and differential coefficient in the controller. The increase of the proportional coefficient increases the overshoot and the rise time decreases. The increase of the differential coefficient reduces the overshoot and the rise time. That is, the weighted term corresponding to γ ​​is used to weigh the proportional action and the differential action. Here, the predetermined indicators are taken as an overshoot of 20% and a rise time of 50 control cycles; μ is the weighted coefficient of the balance time of the dynamic response performance control object in the fitness function, which is directly related to the integral coefficient. The integral coefficient increases and the balance time decreases. Here, the predetermined indicator is taken as the balance time of 1000 control cycles.

[0090] In step 3, the optimization range of the control parameters to be optimized is set according to the boundaries of the control parameters tested in step 1, and the appropriate population size is set in combination with the performance of the processor itself;

[0091] Parameter range to be optimized: Proportional coefficient kp∈(kp min ,kp max ); integral coefficient ki∈(ki min ,ki max ); differential coefficient kd∈(kd min ,kd max );

[0092] Population initialization:

[0093] X=Xmin+rand(eco_size,D)*(Xmax-Xmin)

[0094] Xmax=[kp max,ki max ,kd max ]

[0095] Xmin=[kp min ,ki min ,kd min ]

[0096] Among them, X is the parameter population, eco_size is the initial population size, D is the number of parameters to be optimized, which is 3 here, and rand is a random number between 0 and 1.

[0097] In step 4, based on the decrease of the fitness function value constructed in step 2, and taking the range of each control parameter set in step 3 as the space to be optimized, the symbiotic optimization algorithm is used to optimize the control parameters through the steps of mutualism, partial symbiosis, and parasitism.

[0098] In step 4, if Figure 5 As shown, the specific steps of the symbiotic biological optimization algorithm are:

[0099] Mutual Benefit Symbiosis Formula:

[0100]

[0101] Among them, X i , X j ∈X, represents the individuals of both symbiotic parties before updating, X i_new , X j_new represents the individuals of both symbiotic parties after the update, X best ∈X is the best individual in the current population, BF 1 , BF 2 is the mutual benefit coefficient, which is a random number of 1 or 2. MV is the average of the two individuals in the symbiosis. rand is a random number between 0 and 1. X best -MV*BF 1 , X best -MV*BF 2 In the population search process, it means that two individuals work together to get closer to the optimal individual;

[0102] Finally, compare the fitness function values ​​of the two individuals before and after the update, and retain the best one;

[0103] Partial symbiosis formula:

[0104] X i_new =X i +rand*(X best -X j )

[0105] Among them, rand is a random number between 0 and 1, and X best -X jIt means that in the process of population search, it reflects the help provided by individual j to individual i, so that individual i gradually approaches the optimal individual;

[0106] Finally, compare the fitness function values ​​of individual i before and after the update, and retain the best one;

[0107] Parasitic formula:

[0108]

[0109] Among them, ceil is the rounding function, D is the number of parameters to be optimized, which is 3 here, rand is a random number between 0 and 1, and X dmax , X dmin Represents individual X i The maximum and minimum values ​​of the dth dimension range, d is 1, 2 or 3, X i_new (d) represents individual X i The change value of the d-th dimension parameter;

[0110] Finally, compare X i_new The fitness function value of another random individual is selected and retained.

[0111] In step five, when the fitness function value of the symbiotic optimization algorithm in step four is less than the specified value or the number of population iterations reaches the specified value, the parameter intelligent optimization is terminated. At this time, the population individual corresponding to the minimum value of the fitness function is the optimal PID control parameter obtained by weighing various system indicators.

[0112] When the fitness function J reaches the minimum value, it means that the proportional coefficient, integral coefficient and differential coefficient obtained by the symbiotic algorithm are the optimal control parameters for balancing the dynamic and static response indicators of the system. The digital accelerometer servo control system after online parameter optimization can be obtained by directly applying the optimal control parameters to the closed-loop system controller.

[0113] The specific implementation mode of the present invention takes a certain type of digital quartz flexible accelerometer closed-loop control system as an example. The system adopts a digital PID controller integrated in a microprocessor, and the overall structure of the system conforms to the characteristics of a second-order system, and is suitable for the parameter boundary test algorithm rules of step one and the intelligent algorithm optimization of step four of the present invention.

[0114] In this embodiment, the debugging result is:

[0115] Step 1: The results of the boundary test of each control parameter of the controller are:

[0116] Test kp∈(0.016,0.125); ki∈(0.54,0.66); kd∈(0.00001,0.00005);

[0117] Step 2: integrate various system performance indicators and set the system fitness function as follows:

[0118]

[0119] In the formula, α, β are 5, γ is 10, μ is 1000, and the control main frequency fc is 100KHz;

[0120] Step 3: The control parameters to be optimized and the population initialization debugging results are:

[0121] max_iter=100;% number of iterations

[0122] D=3;%Parameter dimension to be optimized

[0123] eco_size=20;% population size

[0124] Xmin=[0.016 0.54 0.00001];% parameter lower limit

[0125] Xmax=[0.125 0.66 0.00005];%parameter upper limit

[0126] X=Xmin+rand(eco_size,D)*(Xmax-Xmin); % Population initialization

[0127] Among them, max_iter is the cutoff condition for the number of iterations of the optimization algorithm, the parameter dimension D is 3, namely the proportional coefficient, integral coefficient, differential coefficient, and the population size eco_size is 50;

[0128] In step five, based on Figure 1 , Figure 5 Process, when the fitness function J reaches the specified number of population iterations, it means that the PID control parameters obtained by the symbiotic algorithm are the optimal PID control parameters after comprehensive dynamic performance indicators and static noise indicators. The optimized parameters can directly meet the engineering indicator requirements of the digital accelerometer servo control system for stable operation as well as pulse response and step response.

[0129] In addition, it should be understood that after reading the above description of the present invention, those skilled in the art may make various changes or modifications to the present invention, and these equivalent forms also fall within the scope defined by the claims attached to this application.

Claims

1. A method for online intelligent optimization of control parameters of a digital accelerometer servo control system, characterized in that: include: Step 1: boundary test of various control parameters of the controller under specified performance conditions; Step 2: integrate various system performance indicators and set the system fitness function; Step 3: Set the range of control parameters to be optimized and initialize the population based on the system processor performance; Step 4: Using the symbiotic optimization algorithm, the control parameters are optimized online intelligently based on the fitness function value; Step 5: When the symbiotic optimization algorithm meets the termination conditions, the control parameters after comprehensive performance indicators can be obtained, and the optimized parameters are directly assigned to the controller to complete the online intelligent optimization of the control parameters of the digital accelerometer servo control system.

2. The method for online intelligent optimization of control parameters of a digital accelerometer servo control system according to claim 1, characterized in that: In step 1, the digital accelerometer head is placed at zero position, and the microprocessor is programmed to make the controller output positive and negative excitations of the same amplitude, and the excitation amplitude is gradually increased until the head capacitance reaches the boundary, and the head capacitance value is measured, that is, the capacitance boundary value, which is also the range of change of the system control object; Specify the system indicators to be debugged, including hard indicators and optimization indicators. Hard indicators include pulse response, and optimization indicators include system overshoot, rise time, and control object balance time. Gradually increase the proportional coefficient until the controlled object shows equal amplitude oscillation. The proportional coefficient at this time is recorded as the upper boundary kp of the proportional coefficient. max , gradually reduce the proportional coefficient until the impulse response theory can reach the maximum amplitude just meeting the predetermined index, and record the proportional coefficient at this time as the lower boundary of the proportional coefficient kp min ; The maximum amplitude that can be achieved in pulse response theory is calculated based on the output value of the capacitor stable section and the capacitor boundary value under a 1g step. The formula is as follows: pulse=cap / step_1g Among them, pulse is the maximum amplitude that can be reached in pulse response theory, cap is the capacitance boundary value, and step_1g is the output value of the capacitance stable section under the 1g amplitude step response test; On the basis of the lower boundary of the proportional coefficient, gradually increase the integral coefficient until the capacitance balance time meets the predetermined index under the 1g amplitude step response test, record the integral coefficient ki at this time, and take its 10% disturbance range as the parameter variation range, that is, ki min =0.9*ki,ki max =1.1*ki; The lower boundary of the proportional coefficient and the critical integral coefficient ki min On this basis, gradually increase the differential coefficient until the system overshoot and rise time of the controller output response under the 1g amplitude step response test meet the predetermined indicators, and record the differential coefficient at this time as the lower boundary of the differential coefficient kd min ; On the upper boundary of the proportional coefficient and the critical integral coefficient ki max On this basis, gradually increase the differential coefficient until the system overshoot and rise time of the controller output response under the 1g amplitude step response test meet the predetermined indicators or the system oscillates, and record the differential coefficient at this time as the upper boundary of the differential coefficient kd max .

3. The method for online intelligent optimization of control parameters of a digital accelerometer servo control system according to claim 2, characterized in that: In step 2, when the control parameters have met the hard indicators of pulse response, the constructed fitness function includes dynamic indicators including system overshoot, rise time, and control object equilibrium time, and static indicators of controller output noise under zero input of the system. According to the influence of the control rate of the control algorithm itself on each indicator, a weighted combination of each indicator is performed to obtain a fitness function that meets actual needs.

4. The method for online intelligent optimization of control parameters of a digital accelerometer servo control system according to claim 3, characterized in that: In step 2, before optimizing the proportional coefficient, integral coefficient, and differential coefficient, the signal flow simplification rule is first used to simplify the digital accelerometer closed-loop system as an open-loop system with input u and corresponding output y after given proportional coefficients, integral coefficients, and differential coefficients; The u is the constant input of the closed-loop system, and the y is the response output of the closed-loop system; a 1g unit step and a 0 input are performed on the u, and the 1g unit step response time is kept consistent with the 0 input response time. The closed-loop system response output is collected, and the controller output overshoot, rise time rise_time and equilibrium time equi_time of the controlled object under the 1g unit step are calculated, and the indicator fitness evaluation is performed to obtain the system fitness function, and the formula is as follows: in: u_d(k) is the closed-loop constant 1g unit step input sequence; y_d(k) is the output sequence of the closed-loop system 1g unit step response; u_s(k) is the closed-loop constant-value 0 input sequence; y_s(k) is the closed-loop system zero-input response output sequence; fc is the main control frequency of the closed-loop system; α and β are the weighted coefficients of the indicators for 1g unit step and 0 input response in the fitness function, that is, the trade-off between the dynamic response and static response of the system; γ is the weighted coefficient for the overshoot and rise time of the dynamic response performance in the fitness function, which is used to weigh the proportional action and the differential action; μ is the weighted coefficient for the equilibrium time of the dynamic response performance control object in the fitness function, which is directly related to the integral coefficient.

5. The method for online intelligent optimization of control parameters of a digital accelerometer servo control system according to claim 2, characterized in that: In step 3, the optimization range of the control parameters to be optimized is set according to the boundaries of the control parameters tested in step 1, and the appropriate population size is set in combination with the performance of the processor itself; Parameter range to be optimized: Proportional coefficient kp∈(kp min ,kp max ); Integration coefficient ki∈(ki min ,ki max ); differential coefficient kd∈(kd min ,kd max ); Population initialization: X=Xmin+rand(eco_size,D)*(Xmax-Xmin) Xmax=[kp max ,ki max ,kd max ] Xmin=[kp min ,ki min ,kd min ] Among them, X is the parameter population, eco_size is the initial population size, D is the number of parameters to be optimized, which is 3 here, and rand is a random number between 0 and 1.

6. The method for online intelligent optimization of control parameters of a digital accelerometer servo control system according to claim 5, characterized in that: In step 4, based on the decrease in the fitness function value constructed in step 2, and taking the range of each control parameter set in step 3 as the space to be optimized, the symbiotic optimization algorithm is used to optimize the control parameters through the steps of mutualism, partial symbiosis, and parasitism.

7. The method for online intelligent optimization of control parameters of a digital accelerometer servo control system according to claim 6, characterized in that: In step 4, the specific steps of the symbiotic biological optimization algorithm are: Mutual Benefit Symbiosis Formula: Among them, X i , X j ∈X, represents the individuals of both symbiotic parties before updating, X i_new , X j_new represents the individuals of both symbiotic parties after the update, X best ∈X is the optimal individual of the current population, BF1 and BF2 are the mutual benefit coefficients, which are random numbers 1 or 2, and MV is the average of the individuals of both symbiotic parties; rand is a random number between 0 and 1; X best -MV*BF1,X best -MV*BF2 indicates that two individuals work together to get closer to the optimal individual during the population search process; Finally, compare the fitness function values ​​of the two individuals before and after the update, and retain the best one; Partial symbiosis formula: X i_new =X i +rand*(X best -X j ) Among them, rand is a random number between 0 and 1, and X best -X j It means that in the process of population search, it reflects the help provided by individual j to individual i, so that individual i gradually approaches the optimal individual; Finally, compare the fitness function values ​​of individual i before and after the update, and retain the best one; Parasitic formula: Among them, ceil is the rounding function, D is the number of parameters to be optimized, which is 3 here, rand is a random number between 0 and 1, and X dmax , X dmin Represents individual X i The maximum and minimum values ​​of the dth dimension range, d is 1, 2 or 3, X i_new (d) represents individual X i The change value of the d-th dimension parameter; Finally, compare X i_new The fitness function value of another random individual is selected and retained.

8. The method for online intelligent optimization of control parameters of a digital accelerometer servo control system according to claim 5, characterized in that: In step five, when the fitness function value of the symbiotic optimization algorithm in step four is less than the specified value or the number of population iterations reaches the specified value, the parameter intelligent optimization is terminated. At this time, the population individual corresponding to the minimum value of the fitness function is the optimal PID control parameter obtained by weighing various system indicators.

Images