Dynamic response method and system of electric descending device based on fuzzy algorithm

Abstract

This invention provides a dynamic response method and system for an electric descending device based on a fuzzy algorithm. This method relates to the technical field of electric descending devices. The method comprises constructing a fuzzy control rule base by acquiring operating parameters such as descending speed and load weight; optimizing the control rules using a particle swarm optimization algorithm; inputting real-time operating parameters into the optimized rule base to obtain motor control signals; collecting motor output data to calculate dynamic response characteristics; and using a neural network predictor to generate compensation control variables to form optimized control signals. This invention improves the control accuracy and dynamic response speed of the electric descending device, and enhances the system's stability and reliability under different load conditions.

Description

Technical Field

[0001] The present invention relates to electric descending device technology, and in particular to a dynamic response method and system of an electric descending device based on a fuzzy algorithm. Background Art

[0002] Electric descenders are safety devices widely used in firefighting, aerial work, construction, and other fields. Their primary function is to ensure the safety of people and objects moving downward from a height by controlling the descent speed. With the advancement of industrial automation and intelligentization, the control technology of electric descenders has also been continuously upgraded, evolving from simple mechanical braking to electronic control systems. This is particularly true given the increasing demand for dynamic response capabilities in complex environments and under changing operating conditions.

[0003] Traditional electric descending control systems primarily utilize PID control or simple proportional control methods, adjusting the motor output torque using preset parameters to control the descending speed. While these control methods maintain relatively stable performance under fixed operating conditions, they struggle to adapt to varying load weights and environmental conditions. Sudden changes in load weight or external interference can stunt the system's response, making it difficult to adjust the motor output in a timely manner. This can lead to fluctuations in the descending speed and even dangerous shocks or loss of control.

[0004] The main drawbacks of existing electric descending control technology are: First, insufficient control accuracy. Most existing technologies use linear control strategies, which are difficult to cope with nonlinear operating conditions. In particular, under extremely light or heavy load conditions, control accuracy is significantly reduced, and a stable descending speed cannot be guaranteed. Second, dynamic response capability is limited. When the system faces sudden load changes or external interference, the traditional control method is slow to adjust, which is prone to speed fluctuations and shocks, increasing safety risks. Finally, there is a lack of adaptive optimization capabilities. Existing systems rarely have learning and self-optimization functions, and are unable to continuously improve control strategies based on historical operating data. This makes it difficult to improve long-term performance and requires frequent manual debugging and maintenance. Summary of the Invention

[0005] The embodiments of the present invention provide a dynamic response method and system for an electric descending device based on a fuzzy algorithm, which can solve the problems in the prior art.

[0006] A first aspect of an embodiment of the present invention provides a dynamic response method of an electric descending device based on a fuzzy algorithm, comprising:

[0007] The descent speed, load weight, and motor speed of the electric descender are obtained as working parameters, the descent speed and load weight are set as fuzzy control input variables, the motor output torque is set as the fuzzy control output variable, and a fuzzy control rule library is constructed;

[0008] Based on the particle swarm optimization algorithm, the rules in the fuzzy control rule base are quantitatively evaluated for fitness, and the fitness rules below a preset fitness threshold are screened out and then reorganized and mutated to generate a candidate rule set. The candidate rule set is input into a preset forward evaluation model for iterative optimization, and the rule optimization is completed when the overall fitness of the fuzzy control rule base reaches the preset fitness threshold;

[0009] Inputting the working parameters acquired in real time into the optimized fuzzy control rule library, obtaining a motor control signal through fuzzy reasoning operation, and outputting the motor control signal to the driving motor of the electric descending device;

[0010] Collecting actual output torque and motor speed data from the drive motor to calculate the dynamic response characteristics of the electric descending device;

[0011] Using historical response data to train a neural network dynamic response predictor, inputting the dynamic response characteristics into the neural network dynamic response predictor for real-time prediction, calculating a compensation control amount based on the prediction result, and superimposing the compensation control amount on the motor control signal to obtain an optimized control signal;

[0012] The electric descending device is driven to perform a descending action based on the optimized control signal, and the operation data is stored in a historical database for subsequent optimization.

[0013] Based on the particle swarm optimization algorithm, the rules in the fuzzy control rule base are quantitatively evaluated for fitness, and the fitness rules below the preset fitness threshold are screened out and then reorganized and mutated to generate a candidate rule set including:

[0014] Encoding the input and output membership function parameters of each rule in the fuzzy control rule base into a position vector in a particle swarm, initializing the velocity vector corresponding to the position vector, and constructing an initial particle swarm;

[0015] Obtaining corresponding fuzzy rules based on the initial particle swarm decoding, inputting the fuzzy rules into the controller for operation, collecting output errors and dynamic response data during the control process, and calculating rule fitness values ​​according to the output errors and the dynamic response data;

[0016] Record the position of the particle with the highest fitness value of the rule as the global optimal position, record the position corresponding to the historical optimal fitness of each particle as the individual optimal position, and calculate the particle velocity update amount based on the global optimal position and the individual optimal position;

[0017] The particle position is adjusted according to the particle velocity update amount, the rule fitness value corresponding to the updated particle position is compared with a preset fitness threshold, and the rules whose rule fitness value is lower than the preset fitness threshold are screened out as the rules to be optimized; a cross-recombination operation is performed on the rules to be optimized to generate new rules, and Gaussian mutation is applied to the new rules to obtain a candidate rule set.

[0018] Inputting the candidate rule set into a preset forward evaluation model for iterative optimization, and completing the rule optimization when the overall fitness of the fuzzy control rule base reaches the preset fitness threshold, includes:

[0019] Inputting the candidate rule set into a forward evaluation model, collecting the actual output value, expected output value, maximum output value, and steady-state output value during the control process, calculating a steady-state error index based on the actual output value and the expected output value, calculating an overshoot index based on the maximum output value and the steady-state output value, and obtaining a regulation time index for the output response to reach a steady-state range;

[0020] The steady-state error index, the overshoot index and the adjustment time index are weighted and summed with a preset weight coefficient to obtain a rule performance index, and the fitness value of each rule is calculated in combination with the rule activation frequency;

[0021] Calculating the fitness average of all rules in the candidate rule set to obtain an overall fitness value of the rule base, constructing a dynamic evaluation threshold according to the current number of iterations, and comparing the overall fitness value of the rule base with the dynamic evaluation threshold;

[0022] When the overall fitness value of the rule base is greater than the dynamic evaluation threshold, or the fitness improvement rate of consecutive iterations is less than the preset convergence threshold, the current rule base is output as the optimization result, otherwise the overall fitness value of the rule base is returned for the next round of iterative optimization.

[0023] Inputting the real-time acquired working parameters into the optimized fuzzy control rule base, and obtaining the motor control signal through fuzzy inference operation includes:

[0024] Performing fuzzy processing on the working parameters to obtain membership values, and inputting the membership values ​​into an optimized fuzzy control rule library for rule matching;

[0025] The rule activation strength is calculated based on the result of the rule matching, and the center of gravity method is used to perform a defuzzification operation on the rule activation strength to obtain a motor control signal.

[0026] Using historical response data to train a neural network dynamic response predictor, inputting the dynamic response characteristics into the neural network dynamic response predictor for real-time prediction, calculating a compensation control amount based on the prediction result, and superimposing the compensation control amount on the motor control signal to obtain an optimized control signal includes:

[0027] Collecting historical response data during the operation of the motor system, constructing training samples from the historical response data according to the corresponding relationship between input features and output responses, iteratively training a neural network dynamic response predictor using the training samples, optimizing network parameters of the neural network dynamic response predictor through forward calculation and error back propagation of the neural network dynamic response predictor, and completing the predictor training when the parameter prediction error is less than a preset error threshold;

[0028] Inputting the dynamic response characteristics into the trained neural network dynamic response predictor, obtaining a predicted response value of the motor system through forward calculation of the neural network dynamic response predictor, and comparing the predicted response value with an expected response value to obtain a response prediction error;

[0029] A compensation control amount is calculated based on the response prediction error, the compensation control amount is superimposed on the original motor control signal to obtain an optimized control signal, and the optimized control signal is output for motor control.

[0030] Obtaining a predicted response value of the motor system by forward calculation of the neural network dynamic response predictor, and comparing the predicted response value with the expected response value to obtain a response prediction error includes:

[0031] Inputting the dynamic response characteristics of the motor system into the neural network dynamic response predictor, and performing forward operations through hidden layer calculations and output layer calculations of the neural network dynamic response predictor to obtain a predicted response value;

[0032] An expected response value is obtained from the motor control system, and a difference calculation is performed between the predicted response value and the expected response value to obtain a response prediction error.

[0033] A second aspect of an embodiment of the present invention provides a dynamic response system of an electric descending device based on a fuzzy algorithm, comprising:

[0034] The first unit is used to obtain the descending speed, load weight and motor speed of the electric descending device as working parameters, set the descending speed and load weight as fuzzy control input variables, set the motor output torque as the fuzzy control output variable, and build a fuzzy control rule library;

[0035] The second unit is configured to perform a quantitative fitness evaluation on the rules in the fuzzy control rule base based on a particle swarm optimization algorithm, screen out fitness rules below a preset fitness threshold, reorganize and mutate them, generate a candidate rule set, input the candidate rule set into a preset forward evaluation model for iterative optimization, and complete the rule optimization when the overall fitness of the fuzzy control rule base reaches the preset fitness threshold;

[0036] The third unit is configured to input the working parameters acquired in real time into the optimized fuzzy control rule library, obtain a motor control signal through fuzzy reasoning operation, and output the motor control signal to the driving motor of the electric descending device;

[0037] The fourth unit is used to collect actual output torque and motor speed data from the driving motor and calculate the dynamic response characteristics of the electric descending device;

[0038] a fifth unit, configured to train a neural network dynamic response predictor using historical response data, input the dynamic response characteristics into the neural network dynamic response predictor for real-time prediction, calculate a compensation control amount based on the prediction result, and superimpose the compensation control amount on the motor control signal to obtain an optimized control signal;

[0039] The sixth unit is used to drive the electric descent control device to perform a descent action based on the optimized control signal, and store the operation data in a historical database for subsequent optimization.

[0040] According to a third aspect of an embodiment of the present invention, an electronic device is provided, including:

[0041] processor;

[0042] a memory for storing processor-executable instructions;

[0043] The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0044] According to a fourth aspect of an embodiment of the present invention, a computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the method described above is implemented.

[0045] The beneficial effects of this application are as follows:

[0046] By building a fuzzy control rule library that integrates the descent speed and load weight and optimizing the rules using the particle swarm algorithm, the intelligence and accuracy of the electric descent control system are achieved, effectively improving the stability and reliability of the system under complex working conditions.

[0047] A neural network dynamic response predictor is used to predict the descent process in real time and calculate the compensation control quantity, which realizes the precise control of the dynamic response characteristics of the electric descent device, significantly improves the smoothness of the descent process, and effectively eliminates the overshoot and oscillation phenomena in traditional control methods.

[0048] The continuous optimization mechanism based on the historical database enables the system to have self-learning capabilities and can adaptively adjust control parameters according to different working conditions, greatly improving the working efficiency and service life of the electric descending device under various load conditions, while reducing energy consumption and maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 Schematic diagram of the flow of a dynamic response method of an electric descending device based on a fuzzy algorithm according to an embodiment of the present invention;

[0050] Figure 2 This is a flow chart of the particle swarm optimization algorithm according to an embodiment of the present invention;

[0051] Figure 3 A histogram showing the performance comparison and analysis of the fuzzy control rule optimization method according to an embodiment of the present invention;

[0052] Figure 4 This is a flow chart of the neural network dynamic response prediction control according to an embodiment of the present invention. DETAILED DESCRIPTION

[0053] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0054] The following specific embodiments are used to describe the technical solution of the present invention in detail. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0055] Figure 1 FIG. 1 is a flow chart of a dynamic response method of an electric descending device based on a fuzzy algorithm according to an embodiment of the present invention. Figure 1 As shown, the method includes:

[0056] The descent speed, load weight, and motor speed of the electric descender are obtained as working parameters, the descent speed and load weight are set as fuzzy control input variables, the motor output torque is set as the fuzzy control output variable, and a fuzzy control rule library is constructed;

[0057] Based on the particle swarm optimization algorithm, the rules in the fuzzy control rule base are quantitatively evaluated for fitness, and the fitness rules below a preset fitness threshold are screened out and then reorganized and mutated to generate a candidate rule set. The candidate rule set is input into a preset forward evaluation model for iterative optimization, and the rule optimization is completed when the overall fitness of the fuzzy control rule base reaches the preset fitness threshold;

[0058] Inputting the working parameters acquired in real time into the optimized fuzzy control rule library, obtaining a motor control signal through fuzzy reasoning operation, and outputting the motor control signal to the driving motor of the electric descending device;

[0059] Collecting actual output torque and motor speed data from the drive motor to calculate the dynamic response characteristics of the electric descending device;

[0060] Using historical response data to train a neural network dynamic response predictor, inputting the dynamic response characteristics into the neural network dynamic response predictor for real-time prediction, calculating a compensation control amount based on the prediction result, and superimposing the compensation control amount on the motor control signal to obtain an optimized control signal;

[0061] The electric descending device is driven to perform a descending action based on the optimized control signal, and the operation data is stored in a historical database for subsequent optimization.

[0062] In an optional embodiment, the rules in the fuzzy control rule base are quantitatively evaluated for fitness based on a particle swarm optimization algorithm, and the fitness rules below a preset fitness threshold are screened out and then reorganized and mutated to generate a candidate rule set, including:

[0063] Encoding the input and output membership function parameters of each rule in the fuzzy control rule base into a position vector in a particle swarm, initializing the velocity vector corresponding to the position vector, and constructing an initial particle swarm;

[0064] Obtaining corresponding fuzzy rules based on the initial particle swarm decoding, inputting the fuzzy rules into the controller for operation, collecting output errors and dynamic response data during the control process, and calculating rule fitness values ​​according to the output errors and the dynamic response data;

[0065] Record the position of the particle with the highest fitness value of the rule as the global optimal position, record the position corresponding to the historical optimal fitness of each particle as the individual optimal position, and calculate the particle velocity update amount based on the global optimal position and the individual optimal position;

[0066] The particle position is adjusted according to the particle velocity update amount, the rule fitness value corresponding to the updated particle position is compared with a preset fitness threshold, and the rules whose rule fitness value is lower than the preset fitness threshold are screened out as the rules to be optimized; a cross-recombination operation is performed on the rules to be optimized to generate new rules, and Gaussian mutation is applied to the new rules to obtain a candidate rule set.

[0067] like Figure 2 As shown, the method includes:

[0068] The fuzzy control rule base contains several if-then rules, each of which contains membership functions for input and output variables. For example, in a temperature control system, the input variables might be the temperature error e and the error rate of change ec, while the output variable is the controlled variable u. The membership function for each variable can typically be represented as a triangular or trapezoidal function, with its shape and position determined by several parameters.

[0069] When encoding the input and output membership function parameters of each rule in the fuzzy control rule base as a particle position vector, real number encoding can be used. For example, if there are 10 rules, each containing two input variables and one output variable, and each variable has a membership function determined by three parameters, then the dimension of each particle's position vector is 10 × (2 + 1) × 3 = 90 dimensions. For the jth variable of the i-th rule, its membership function parameters can be represented as the (i-1) × (2 + 1) × 3 + (j-1) × 3 + 1 to (i-1) × (2 + 1) × 3 + j × 3 elements in the particle position vector.

[0070] The velocity vector can be randomly generated in the range [-Vmax, Vmax], where Vmax can be set to 20% of the search space. For example, if the membership function parameter range is [0, 100], then Vmax = 20. In a temperature control system, the population size can be set to 30, initializing 30 particles to form the initial particle swarm.

[0071] After the particle swarm is initialized, each particle's position vector must be decoded to obtain the corresponding fuzzy rule. Taking the temperature control system as an example, the decoding process maps the values ​​in the position vector back to the membership function parameters. If the values ​​of the first three elements of a particle's position vector are [20, 40, 60], they may represent the triangular membership function parameters for the first input variable of the first rule.

[0072] By inputting the decoded fuzzy rules into the controller for simulation, performance metrics can be collected during the control process. In a temperature control system, these metrics include the error e(t) between the setpoint and actual temperature, as well as dynamic response parameters such as overshoot, rise time, and settling time. A fitness function can be designed based on these metrics. For example, a time-weighted error integral metric can be used. For n sampling points within a control period T, the fitness value can be calculated as a weighted combination of the sum of squared errors and the dynamic performance metric.

[0073] Assume that in a certain iteration, the fitness value of the rule corresponding to the fifth particle is 85.6, which is the current highest fitness value. The position of this particle is recorded as the global optimal position gbest. At the same time, if the current fitness value of the eighth particle is 78.3, which is lower than its historical best fitness value of 79.1, its individual optimal position pbest remains unchanged. Conversely, if it is higher than the historical best, pbest is updated to the current position.

[0074] Particle velocity updates use an inertia weighting method. The new velocity consists of three components: the inertia term of the current velocity, the cognitive term for moving towards the individual optimal position, and the social term for moving towards the global optimal position. The inertia weight can be set to decrease linearly from 0.9 to 0.4, and the acceleration constant can be set to 2.0 and 2.0, respectively.

[0075] For example, for a particle in a certain dimension, the current velocity is 10, the current position is 35, the individual optimal position is 40, the global optimal position is 45, and the inertia weight is 0.8, then the new velocity may be calculated as 8 + random number × 2.0 × (40-35) + random number × 2.0 × (45-35) = 8 + 7 + 14 = 29.

[0076] Adjust the particle position based on the updated velocity. If the velocity in a dimension is 29 and the current position is 35, the new position is 35 + 29 = 64. If the position exceeds the boundary, it is truncated to the boundary value. The preset fitness threshold can be set to 80% of the global optimal fitness. For example, if the gbest corresponding fitness is 85.6, the threshold is 68.48.

[0077] In a given iteration, suppose the fitness of particles 12, 17, and 23 are 65.2, 62.8, and 67.1, respectively, all below the threshold of 68.48. These three rules are marked as pending optimization. When performing a crossover recombination operation on these rules, either a single-point crossover or a uniform crossover method can be used. For example, the position vectors of particles 12 and 17 intersect at position 45, generating two new particles.

[0078] Applying Gaussian mutation to the newly generated rules increases the diversity of the search space. Mutation adds a Gaussian random number with a mean of 0 and a decreasing standard deviation to the selected parameter. For example, if the value of the 50th element of a new particle is 58.3, a Gaussian random number with a standard deviation of 5 might be added, resulting in a mutated value of 61.5.

[0079] Through the cross-recombination and mutation operations described above, the resulting candidate rule set replaces the original low-fitness rules, forming a new generation of particle swarms. In temperature control system testing, the optimized fuzzy controller reduced overshoot from 15% to 8% and settling time from 12 seconds to 7 seconds, significantly improving control performance compared to the original controller.

[0080] In an optional embodiment, inputting the candidate rule set into a preset forward evaluation model for iterative optimization, and completing the rule optimization when the overall fitness of the fuzzy control rule base reaches the preset fitness threshold, comprises:

[0081] Inputting the candidate rule set into a forward evaluation model, collecting the actual output value, expected output value, maximum output value, and steady-state output value during the control process, calculating a steady-state error index based on the actual output value and the expected output value, calculating an overshoot index based on the maximum output value and the steady-state output value, and obtaining a regulation time index for the output response to reach a steady-state range;

[0082] The steady-state error index, the overshoot index and the adjustment time index are weighted and summed with a preset weight coefficient to obtain a rule performance index, and the fitness value of each rule is calculated in combination with the rule activation frequency;

[0083] Calculating the fitness average of all rules in the candidate rule set to obtain an overall fitness value of the rule base, constructing a dynamic evaluation threshold according to the current number of iterations, and comparing the overall fitness value of the rule base with the dynamic evaluation threshold;

[0084] When the overall fitness value of the rule base is greater than the dynamic evaluation threshold, or the fitness improvement rate of consecutive iterations is less than the preset convergence threshold, the current rule base is output as the optimization result, otherwise the overall fitness value of the rule base is returned for the next round of iterative optimization.

[0085] The candidate rule set is an initial set of fuzzy control rules generated from domain expert knowledge or historical data. It consists of multiple IF-THEN fuzzy rules. Each rule has an antecedent (the input variable and its fuzzy membership) and a consequent (the output variable and its fuzzy membership). For example, in a temperature control system, a rule might be expressed as "IF the temperature error is positive and large AND the temperature change rate is positive and small THEN the control output is neutral."

[0086] When candidate rule sets are fed into the forward evaluation model for evaluation, the system constructs a closed-loop control simulation environment. This environment includes a controlled plant model (e.g., a second-order system model), a fuzzy controller, and related parameter settings. For a temperature control system, the target temperature can be set to 25°C, the initial temperature to 20°C, the sampling period to 0.1 seconds, and the total simulation duration to 100 seconds.

[0087] During the simulation, the system collects control process data in real time. The actual output value refers to the actual temperature value of the control system at each sampling moment. For example, at t = 10 seconds, it might be 23.5°C. The expected output value refers to the system's target setpoint, in this case 25°C. The maximum output value refers to the peak temperature during the response, for example, 26.2°C. The steady-state output value refers to the temperature value after the system stabilizes, for example, 25.1°C.

[0088] Based on the collected data, the system calculates three key performance indicators. The steady-state error indicator is calculated by taking the absolute difference between the actual output value and the expected output value after the system stabilizes, such as:

[0089] |25.1-25|=0.1℃.

[0090] The overshoot specification is calculated as the difference between the maximum output value and the steady-state output value as a percentage of the steady-state output value, for example, (26.2 - 25.1) / 25.1 × 100% = 4.38%. The settling time specification is the time it takes for the output response to first enter and remain within the steady-state range (usually ±5% of the target value), for example, 12.5 seconds.

[0091] After the performance index calculation is completed, the system performs a rule performance evaluation. Assuming the preset weight coefficients are steady-state error 0.4, overshoot 0.3, and adjustment time 0.3, the rule performance index can be obtained by weighted summation. If the normalized steady-state error is 0.05, the overshoot is 0.15, and the adjustment time is 0.2, the rule performance index is:

[0092] 0.4×0.05+0.3×0.15+0.3×0.2=0.125.

[0093] Rule activation frequency refers to the ratio of the number of times a rule is triggered during the entire control process to the total number of control cycles. For example, if a rule is triggered 200 times in 1000 control cycles, its activation frequency is 0.2. Rules with higher activation frequencies have a greater impact on system control effectiveness and should be assigned higher weights.

[0094] The fitness value of each rule is calculated by a weighted combination of the rule performance index and the rule activation frequency. The calculation method is as follows: Rule fitness = (1-Rule performance index) × 0.7 + Rule activation frequency × 0.3. For example, if the rule performance index is 0.125 and the activation frequency is 0.2, the rule fitness is:

[0095] (1-0.125)×0.7+0.2×0.3=0.671.

[0096] The overall fitness value of the rule base is the arithmetic mean of the fitness values ​​of all rules in the candidate rule set. For example, if the rule set contains 25 rules with fitness values ​​of 0.671, 0.685, 0.592, and so on, the overall fitness value is the sum of all the rule fitness values ​​divided by 25. For example, the value might be 0.638.

[0097] The system establishes a dynamic evaluation threshold based on the current number of iterations. As the number of iterations increases, the evaluation threshold gradually increases, promoting continuous improvement in rule quality. For example, the formula: Dynamic Evaluation Threshold = Base Threshold + Number of Iterations × Growth Rate. If the base threshold is 0.5 and the growth rate is 0.01, when the number of iterations is 10, the dynamic evaluation threshold is 0.5 + 10 × 0.01 = 0.6.

[0098] There are two conditions for iterative termination: one is that the overall fitness value of the rule base is greater than the dynamic evaluation threshold; the other is that the fitness improvement rate of consecutive iterations is less than the preset convergence threshold. The fitness improvement rate is calculated as (current fitness - previous round fitness) / previous round fitness. For example, if the current fitness is 0.638 and the previous round fitness was 0.625, the improvement rate is (0.638-0.625) / 0.625 = 0.0208. If the preset convergence threshold is 0.01, the improvement rate is greater than the threshold and the iteration continues; if the improvement rate is less than 0.01 for three consecutive rounds, the iteration is terminated.

[0099] If the termination criteria are met, the system outputs the current rule base as the optimization result, which contains all optimized fuzzy control rules. These rules can be directly applied to the actual control system. If the termination criteria are not met, the system returns the overall fitness value of the rule base for the next round of iterative optimization, which may include rule modification, deletion of redundant rules, or generation of new rules to further improve the quality of the rule base.

[0100] Through the above optimization process, the system can effectively improve the quality of fuzzy control rules and optimize the performance of the control system. It is suitable for various industrial scenarios that require precise control, such as temperature control, pressure control, speed control and other fields.

[0101] Figure 3 This is a bar chart showing the performance comparison and analysis of the fuzzy control rule optimization method according to an embodiment of the present invention:

[0102] The figure shows the performance comparison data of three different control schemes (traditional fuzzy control, genetic algorithm optimized fuzzy control, and this technical solution) under five test scenarios. In terms of adaptability to complex working conditions, the traditional solution achieved 65.4%, the genetic algorithm optimized solution improved to 78.6%, and the present solution reached the highest level at 92.7%. In the test of external disturbance resistance, the three schemes achieved 58.7%, 72.4%, and 89.5%, respectively. The control accuracy index showed that the traditional solution achieved 71.2%, which increased to 83.5% after genetic algorithm optimization, and the present solution further increased to 94.6%. In the convergence speed test, the traditional solution achieved 62.5%, the genetic algorithm optimized solution improved to 76.2%, and the present solution reached 88.7%. In terms of computational efficiency, the traditional solution performed the best, reaching 88.3%, while the genetic algorithm optimized solution dropped to 74.8%, and the present solution maintained a high level of 83.2%. Overall, this technical solution significantly outperformed the other two solutions in all test scenarios except computational efficiency. The improvement in control accuracy was particularly significant, increasing by 23.4 percentage points compared to the traditional solution, demonstrating its clear performance advantage in practical applications. While slightly lower in computational efficiency than the traditional solution, it still maintains an acceptable high level. This slight loss in efficiency is traded for significant performance improvements in other areas, demonstrating its overall application value.

[0103] In an optional embodiment, the operating parameters acquired in real time are input into the optimized fuzzy control rule base, and the motor control signal is obtained through fuzzy inference operation, including:

[0104] Performing fuzzy processing on the working parameters to obtain membership values, and inputting the membership values ​​into an optimized fuzzy control rule library for rule matching;

[0105] The rule activation strength is calculated based on the result of the rule matching, and the center of gravity method is used to perform a defuzzification operation on the rule activation strength to obtain a motor control signal.

[0106] The sensor network collects various operating parameters of the current working environment. These parameters may include temperature, humidity, pressure, speed, position, and other physical quantities. For example, in a temperature control system, if the temperature sensor collects a real-time ambient temperature of 28.5°C and the target temperature is set at 25.0°C, the temperature deviation is 3.5°C, and the temperature change rate is 0.2°C / minute. This collected raw data is filtered and normalized by the data preprocessing module to eliminate noise and standardize the data format.

[0107] Fuzzification of preprocessed operating parameters is a key step in implementing fuzzy control. The goal of fuzzification is to convert precise numerical values ​​into membership values ​​on fuzzy sets. Taking temperature deviation as an example, the system divides the temperature deviation range into seven fuzzy subsets: "negative large," "negative medium," "negative small," "zero," "positive small," "positive medium," and "positive large." For a measured temperature deviation of 3.5°C, the system calculates, using the triangular membership function, that its membership to "positive small" is 0.3, its membership to "positive medium" is 0.7, and its membership to all other fuzzy subsets is 0. Similarly, for a temperature change rate of 0.2°C / minute, the system calculates its membership to "zero" to be 0.6 and its membership to "positive small" to be 0.4.

[0108] These membership values ​​are then input into the optimized fuzzy control rule base for rule matching. A fuzzy control rule base consists of a series of if-then rules that describe the mapping between inputs and outputs. For example, a rule base might include rules such as: "If the temperature deviation is positive and the temperature change rate is zero, then the motor control signal is negative and neutral," or "If the temperature deviation is positive and the temperature change rate is small, then the motor control signal is large and negative," etc. In practical applications, the rule base may contain dozens or even hundreds of rules to cover every possible input combination.

[0109] Each rule's antecedent (IF part) is checked in turn to see if it matches the fuzzy membership value of the current input. For the temperature control example above, the system finds that two rules are activated: Rule 1's antecedent "temperature deviation is positive and the temperature change rate is zero" matches the current input, and Rule 2's antecedent "temperature deviation is positive and the temperature change rate is small" also matches the current input.

[0110] Based on the rule matching results, the system further calculates the rule activation strength. Rule activation strength indicates the degree of influence a rule has on the final control output. In fuzzy control, the MIN operation (minimum value) is often used to calculate rule activation strength. For Rule 1, the membership degree of the antecedent "temperature deviation is neutral" is 0.7, and the membership degree of "temperature change rate is zero" is 0.6. Taking the minimum value, the activation strength of Rule 1 is 0.6. Similarly, the activation strength of Rule 2 is MIN(0.7, 0.4) = 0.4.

[0111] After calculating the rule activation strength, the system uses the center of gravity method to defuzzify the rule activation strengths and obtain the final motor control signal. The center of gravity method is a commonly used defuzzification method. Its basic idea is to truncate the consequent (then portion) of each activated rule according to its activation strength. The centroid of all fuzzy sets after truncation is then calculated as the precise control output value.

[0112] Assume that the fuzzy set corresponding to the consequent of Rule 1, "The motor control signal is negative medium," has a triangular membership function distribution on the output universe [-100, 100], centered at -50. The fuzzy set corresponding to the consequent of Rule 2, "The motor control signal is negative large," has a center at -80. The system first truncates the "negative medium" fuzzy set according to the activation strength of Rule 1, 0.6, to form a flat-top trapezoid with a height of 0.6. Similarly, the "negative large" fuzzy set is truncated according to the activation strength of Rule 2, 0.4. The system then calculates the centroid of the region formed by the two truncated fuzzy sets.

[0113] Assuming the calculated center of mass position is -60°, the resulting motor control signal is -60°. This value indicates that the motor should operate in reverse at 60% power to reduce system temperature. The system converts this control signal into a voltage or current signal suitable for the motor driver. For example, -60° can be mapped to a -7.5V voltage signal and sent to the motor driver to execute the corresponding action.

[0114] In practice, this process is repeated at high frequency. The system continuously collects the latest operating parameters and calculates the latest control signals through fuzzy reasoning, thereby achieving real-time and precise control of the motor. This approach can achieve good control results even when the system model is not accurate or has nonlinear or time-varying characteristics.

[0115] In an optional embodiment, using historical response data to train a neural network dynamic response predictor, inputting the dynamic response characteristics into the neural network dynamic response predictor for real-time prediction, calculating a compensation control amount based on the prediction result, and superimposing the compensation control amount on the motor control signal to obtain an optimized control signal includes:

[0116] Collecting historical response data during the operation of the motor system, constructing training samples from the historical response data according to the corresponding relationship between input features and output responses, iteratively training a neural network dynamic response predictor using the training samples, optimizing network parameters of the neural network dynamic response predictor through forward calculation and error back propagation of the neural network dynamic response predictor, and completing the predictor training when the parameter prediction error is less than a preset error threshold;

[0117] Inputting the dynamic response characteristics into the trained neural network dynamic response predictor, obtaining a predicted response value of the motor system through forward calculation of the neural network dynamic response predictor, and comparing the predicted response value with an expected response value to obtain a response prediction error;

[0118] A compensation control amount is calculated based on the response prediction error, the compensation control amount is superimposed on the original motor control signal to obtain an optimized control signal, and the optimized control signal is output for motor control.

[0119] like Figure 4 As shown, the method includes:

[0120] Acquire historical response data during the motor system's operation. This historical response data includes, but is not limited to, time-varying data on motor parameters such as speed, torque, position, current, and voltage. Specifically, this data can be collected by installing various sensors in the motor system, such as Hall sensors, current sensors, and position encoders. The data acquisition frequency can be set to 200 Hz, meaning data is collected every 5 milliseconds. By running the motor system continuously for 24 hours, approximately 17.28 million data samples can be obtained.

[0121] The collected historical response data undergoes preprocessing, including data cleaning and normalization. Data cleaning removes outliers, such as data points outside the rated motor parameter range. Normalization maps each parameter to the [0, 1] range, improving neural network training efficiency. After preprocessing, training samples are constructed based on the correspondence between input features and output responses. Input features may include motor control signals, load status, and historical response status. The output response is the corresponding motor response value, such as speed and position.

[0122] A neural network dynamic response predictor was constructed. This predictor uses a fully connected neural network structure, consisting of an input layer, hidden layers, and an output layer. The number of nodes in the input layer is equal to the input feature dimension. For example, if the control signal, load status, and the response status at the previous three moments are selected as input, the number of nodes is 5. The hidden layer adopts a two-layer structure, with 32 nodes in the first layer and 16 nodes in the second layer. The number of nodes in the output layer is equal to the prediction target dimension. For example, if only the speed is predicted, the number of nodes is 1. The ReLU function is used as the activation function to avoid the vanishing gradient problem.

[0123] The constructed training samples were used to iteratively train the neural network dynamic response predictor. The training process employed batch gradient descent, with a batch size of 128 and an initial learning rate of 0.01. A learning rate decay strategy was used, with the learning rate decreasing by a factor of 0.9 every 50 epochs. The mean squared error (MSE) loss function was used to calculate the difference between the predicted and actual values. The network parameters were updated using a backpropagation algorithm, using the Adam optimizer with parameters β1 set to 0.9, β2 set to 0.999, and ε set to 1e-8. During training, the prediction error on the validation set was monitored. Training was terminated when the validation set error change was less than a preset threshold of 0.0001 for 10 consecutive training epochs, or when the maximum number of training epochs, 500, was reached.

[0124] During actual use, the dynamic response characteristics of the motor system are input into the trained neural network dynamic response predictor in real time. These dynamic response characteristics include the current control signal, load status, and response status at the previous moments. Through the neural network's forward calculation, the predicted motor system response value for the next moment is obtained. During the forward calculation process, the input data undergoes the same normalization processing as during training, and then passes through each layer of the network for calculation, ultimately outputting the predicted value. For example, if the input motor control signal is 50% of the rated value, the load is 30% of the rated value, and the speeds at the previous three moments are 1000 rpm, 1050 rpm, and 1080 rpm, respectively, the predictor may output a predicted speed value of 1100 rpm for the next moment.

[0125] The predicted response value is compared with the expected response value to calculate the response prediction error. The expected response value is the target response value set according to the control requirements, such as a target speed of 1200 rpm. The response prediction error is the difference between the expected response value and the predicted response value, which is 100 rpm in the above example.

[0126] The compensation control variable is calculated based on the response prediction error. This can be calculated using a proportional control method: multiplying the response prediction error by a compensation coefficient. The compensation coefficient can be set based on the motor system characteristics. For example, for speed control, a compensation coefficient of 0.2 can correspond to a compensation control variable of 20% of the rated value. A segmented compensation strategy can also be used: a smaller compensation coefficient is used when the error is small, such as 0.1 when the error is less than 50 rpm; a larger compensation coefficient is used when the error is larger, such as 0.3 when the error is greater than 100 rpm.

[0127] The calculated compensation control variable is superimposed on the original motor control signal to obtain the optimized control signal. This superposition can be simple addition. For example, if the original control signal is 50% of the rated value and the compensation control variable is 20% of the rated value, the optimized control signal is 70% of the rated value. Alternatively, a weighted average method can be used. For example, if the original control signal is weighted 0.7 and the compensation control variable is weighted 0.3, the optimized control signal is 50% × 0.7 + 20% × 0.3 = 41% of the rated value.

[0128] The optimized control signal is output to the motor driver for actual motor control. After receiving the optimized control signal, the motor driver generates a corresponding PWM waveform to control the conduction of the power devices, thereby adjusting the motor's speed, torque, or position. Simultaneously, the actual motor response data is collected and used as input for the next moment's prediction, forming a closed-loop control system.

[0129] By using this method, a neural network is used to predict the dynamic response characteristics of the motor system, optimizing motor control. This effectively reduces performance degradation caused by control lag and improves the dynamic response and stability of the motor system. Experiments show that this method reduces the overshoot of the motor system in the step response from 15% to 5%, reduces the settling time from 200ms to 80ms, and reduces the steady-state error from ±2% to ±0.5%.

[0130] In an optional embodiment, obtaining a predicted response value of the motor system by forward calculation of the neural network dynamic response predictor, and comparing the predicted response value with the expected response value to obtain a response prediction error comprises:

[0131] Inputting the dynamic response characteristics of the motor system into the neural network dynamic response predictor, and performing forward operations through hidden layer calculations and output layer calculations of the neural network dynamic response predictor to obtain a predicted response value;

[0132] An expected response value is obtained from the motor control system, and a difference calculation is performed between the predicted response value and the expected response value to obtain a response prediction error.

[0133] The neural network dynamic response predictor uses a multilayer perceptron architecture, consisting of an input layer, hidden layers, and an output layer. It is used to predict the dynamic response characteristics of a motor system under specific control signals. The dynamic response characteristics of a motor system include parameters such as speed, torque, current, and position, which can be used to characterize the motor system's operating state.

[0134] The forward computational process of the neural network dynamic response predictor begins by inputting the motor system's dynamic response characteristics into the predictor. Specifically, the motor system's current state parameters, such as a speed of 1500 rpm, a torque of 2.5 Nm, and a phase current of 3.2 A, are passed as input vectors to the neural network's input layer. Control commands, such as a voltage command of 310 V, a current command of 5.0 A, and a PWM duty cycle of 0.75, can also be input.

[0135] After the input data enters the neural network, it first passes through the hidden layer calculation. In this embodiment, the hidden layer adopts a two-layer structure, the first hidden layer contains 64 neurons, and the second hidden layer contains 32 neurons. During the hidden layer calculation, the input data is dot-producted with the weights of each neuron, and then the bias value is added, and then processed by the activation function. The activation function uses the ReLU function. When the input is positive, the output is equal to the input value, and when the input is negative, the output is zero. For example, for the first neuron in the first hidden layer, if the dot product result of the input vector and the weight vector plus the bias value is 2.35, the output value after processing by the ReLU function is still 2.35; if the calculated result is -0.5, the output is 0.

[0136] The data processed by the first hidden layer serves as the input to the second hidden layer, where it undergoes the same weight calculation and activation function. The output of the second hidden layer is then passed to the output layer for further calculation. The number of neurons in the output layer depends on the number of motor response parameters to be predicted. For example, when predicting speed and torque, the output layer contains two neurons. The output layer uses a linear activation function, directly outputting the dot product of the hidden layer input and the output layer weight plus the bias value, without any nonlinear transformation.

[0137] After completing the forward calculation, the neural network outputs a predicted response value. For example, the neural network might predict that the motor speed will reach 1850 rpm and the torque will reach 3.2 Nm in 0.1 seconds. These predicted values ​​reflect the neural network's estimate of the future state of the motor system.

[0138] After obtaining the predicted response value, the expected response value must be obtained from the motor control system. The expected response value is the target state that the motor should reach based on control requirements. For example, the motor speed should reach 1800 rpm and the torque should reach 3.0 Nm after 0.1 seconds. The expected response value can be obtained from the control system's reference model, user settings, or the output of the upper-level controller.

[0139] The difference between the predicted response and the expected response is calculated to determine the response prediction error. Using the previous example, the speed prediction error is 1850 rpm - 1800 rpm = 50 rpm, and the torque prediction error is 3.2 Nm - 3.0 Nm = 0.2 Nm. These errors quantify the deviation between the predicted result and the expected target.

[0140] The response prediction error can be used to subsequently adjust control parameters. For example, if the predicted speed is higher than the expected speed, the control system can reduce the voltage output; if the predicted torque is higher than the expected torque, the current command value can be reduced. In this way, the system can adjust control parameters in advance to achieve precise control of the motor's dynamic response.

[0141] The weight parameters of the neural network predictor are obtained through offline training using a large amount of historical operating data. This training data includes the motor's input control signals and corresponding output responses under different operating conditions. For example, operating data is collected under different loads (0.5 Nm, 1.0 Nm, 1.5 Nm, etc.) and different target speeds (500 rpm, 1000 rpm, 1500 rpm, etc.), recording the correspondence between control inputs and system responses. The network weights are then optimized using a backpropagation algorithm to minimize the error between the predicted output and the actual response.

[0142] The hardware platform used for the neural network dynamic response predictor can be an embedded processor, such as a 32-bit microcontroller with a 200MHz clock speed, 256KB of RAM, and 2MB of flash memory. To improve computational efficiency, the network parameters can be quantized into 16-bit fixed-point values, reducing storage requirements and computational overhead. The forward computation cycle can be set to 1ms to match the control cycle of the motor control system.

[0143] This method is applicable to various motor systems, including DC motors, AC induction motors, and permanent magnet synchronous motors. Adapting to each motor type requires simply adjusting the neural network's input features and network structure. For example, for a permanent magnet synchronous motor, the input features can be expanded to include rotor position angle and dq-axis currents; for a DC motor, the input features can be simplified to armature current and speed as the primary input features.

[0144] A second aspect of an embodiment of the present invention provides a dynamic response system of an electric descending device based on a fuzzy algorithm, comprising:

[0145] The first unit is used to obtain the descending speed, load weight and motor speed of the electric descending device as working parameters, set the descending speed and load weight as fuzzy control input variables, set the motor output torque as the fuzzy control output variable, and build a fuzzy control rule library;

[0146] The second unit is configured to perform a quantitative fitness evaluation on the rules in the fuzzy control rule base based on a particle swarm optimization algorithm, screen out fitness rules below a preset fitness threshold, reorganize and mutate them, generate a candidate rule set, input the candidate rule set into a preset forward evaluation model for iterative optimization, and complete the rule optimization when the overall fitness of the fuzzy control rule base reaches the preset fitness threshold;

[0147] The third unit is configured to input the working parameters acquired in real time into the optimized fuzzy control rule library, obtain a motor control signal through fuzzy reasoning operation, and output the motor control signal to the driving motor of the electric descending device;

[0148] The fourth unit is used to collect actual output torque and motor speed data from the driving motor and calculate the dynamic response characteristics of the electric descending device;

[0149] a fifth unit, configured to train a neural network dynamic response predictor using historical response data, input the dynamic response characteristics into the neural network dynamic response predictor for real-time prediction, calculate a compensation control amount based on the prediction result, and superimpose the compensation control amount on the motor control signal to obtain an optimized control signal;

[0150] The sixth unit is used to drive the electric descent control device to perform a descent action based on the optimized control signal, and store the operation data in a historical database for subsequent optimization.

[0151] According to a third aspect of an embodiment of the present invention, an electronic device is provided, including:

[0152] processor;

[0153] a memory for storing processor-executable instructions;

[0154] The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0155] According to a fourth aspect of an embodiment of the present invention, a computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the method described above is implemented.

[0156] The present invention may be a method, an apparatus, a system and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for executing various aspects of the present invention.

[0157] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. The dynamic response method of electric descending device based on fuzzy algorithm is characterized by: include: The descent speed, load weight, and motor speed of the electric descender are obtained as working parameters, the descent speed and load weight are set as fuzzy control input variables, the motor output torque is set as the fuzzy control output variable, and a fuzzy control rule library is constructed; Based on the particle swarm optimization algorithm, the rules in the fuzzy control rule base are quantitatively evaluated for fitness, and the fitness rules below a preset fitness threshold are screened out and then reorganized and mutated to generate a candidate rule set. The candidate rule set is input into a preset forward evaluation model for iterative optimization, and the rule optimization is completed when the overall fitness of the fuzzy control rule base reaches the preset fitness threshold; Inputting the working parameters acquired in real time into the optimized fuzzy control rule library, obtaining a motor control signal through fuzzy reasoning operation, and outputting the motor control signal to the driving motor of the electric descending device; Collecting actual output torque and motor speed data from the drive motor to calculate the dynamic response characteristics of the electric descending device; Using historical response data to train a neural network dynamic response predictor, inputting the dynamic response characteristics into the neural network dynamic response predictor for real-time prediction, calculating a compensation control amount based on the prediction result, and superimposing the compensation control amount on the motor control signal to obtain an optimized control signal; The electric descending device is driven to perform a descending action based on the optimized control signal, and the operation data is stored in a historical database for subsequent optimization.

2. The method according to claim 1, characterized in that Based on the particle swarm optimization algorithm, the rules in the fuzzy control rule base are quantitatively evaluated for fitness, and the fitness rules below the preset fitness threshold are screened out and then reorganized and mutated to generate a candidate rule set including: Encoding the input and output membership function parameters of each rule in the fuzzy control rule base into a position vector in a particle swarm, initializing the velocity vector corresponding to the position vector, and constructing an initial particle swarm; Obtaining corresponding fuzzy rules based on the initial particle swarm decoding, inputting the fuzzy rules into the controller for operation, collecting output errors and dynamic response data during the control process, and calculating rule fitness values ​​according to the output errors and the dynamic response data; Record the position of the particle with the highest fitness value of the rule as the global optimal position, record the position corresponding to the historical optimal fitness of each particle as the individual optimal position, and calculate the particle velocity update amount based on the global optimal position and the individual optimal position; The particle position is adjusted according to the particle velocity update amount, the rule fitness value corresponding to the updated particle position is compared with a preset fitness threshold, and the rules whose rule fitness value is lower than the preset fitness threshold are screened out as the rules to be optimized; a cross-recombination operation is performed on the rules to be optimized to generate new rules, and Gaussian mutation is applied to the new rules to obtain a candidate rule set.

3. The method according to claim 1, characterized in that Inputting the candidate rule set into a preset forward evaluation model for iterative optimization, and completing the rule optimization when the overall fitness of the fuzzy control rule base reaches the preset fitness threshold, includes: Inputting the candidate rule set into a forward evaluation model, collecting the actual output value, expected output value, maximum output value, and steady-state output value during the control process, calculating a steady-state error index based on the actual output value and the expected output value, calculating an overshoot index based on the maximum output value and the steady-state output value, and obtaining a regulation time index for the output response to reach a steady-state range; The steady-state error index, the overshoot index and the adjustment time index are weighted and summed with a preset weight coefficient to obtain a rule performance index, and the fitness value of each rule is calculated in combination with the rule activation frequency; Calculating the fitness average of all rules in the candidate rule set to obtain an overall fitness value of the rule base, constructing a dynamic evaluation threshold according to the current number of iterations, and comparing the overall fitness value of the rule base with the dynamic evaluation threshold; When the overall fitness value of the rule base is greater than the dynamic evaluation threshold, or the fitness improvement rate of consecutive iterations is less than the preset convergence threshold, the current rule base is output as the optimization result, otherwise the overall fitness value of the rule base is returned for the next round of iterative optimization.

4. The method according to claim 1, wherein Inputting the real-time acquired working parameters into the optimized fuzzy control rule base, and obtaining the motor control signal through fuzzy inference operation includes: Performing fuzzy processing on the working parameters to obtain membership values, and inputting the membership values ​​into an optimized fuzzy control rule library for rule matching; The rule activation strength is calculated based on the result of the rule matching, and the center of gravity method is used to perform a defuzzification operation on the rule activation strength to obtain a motor control signal.

5. The method according to claim 1, characterized in that Using historical response data to train a neural network dynamic response predictor, inputting the dynamic response characteristics into the neural network dynamic response predictor for real-time prediction, calculating a compensation control amount based on the prediction result, and superimposing the compensation control amount on the motor control signal to obtain an optimized control signal includes: Collecting historical response data during the operation of the motor system, constructing training samples from the historical response data according to the corresponding relationship between input features and output responses, iteratively training a neural network dynamic response predictor using the training samples, optimizing network parameters of the neural network dynamic response predictor through forward calculation and error back propagation of the neural network dynamic response predictor, and completing the predictor training when the parameter prediction error is less than a preset error threshold; Inputting the dynamic response characteristics into the trained neural network dynamic response predictor, obtaining a predicted response value of the motor system through forward calculation of the neural network dynamic response predictor, and comparing the predicted response value with an expected response value to obtain a response prediction error; A compensation control amount is calculated based on the response prediction error, the compensation control amount is superimposed on the original motor control signal to obtain an optimized control signal, and the optimized control signal is output for motor control.

6. The method according to claim 5, characterized in that Obtaining a predicted response value of the motor system by forward calculation of the neural network dynamic response predictor, and comparing the predicted response value with the expected response value to obtain a response prediction error includes: Inputting the dynamic response characteristics of the motor system into the neural network dynamic response predictor, and performing forward operations through hidden layer calculations and output layer calculations of the neural network dynamic response predictor to obtain a predicted response value; An expected response value is obtained from the motor control system, and a difference calculation is performed between the predicted response value and the expected response value to obtain a response prediction error.

7. A dynamic response system of an electric descending device based on a fuzzy algorithm, used to implement the method according to any one of claims 1 to 6, characterized in that: include: The first unit is used to obtain the descending speed, load weight and motor speed of the electric descending device as working parameters, set the descending speed and load weight as fuzzy control input variables, set the motor output torque as the fuzzy control output variable, and build a fuzzy control rule library; The second unit is configured to perform a quantitative fitness evaluation on the rules in the fuzzy control rule base based on a particle swarm optimization algorithm, screen out fitness rules below a preset fitness threshold, reorganize and mutate them, generate a candidate rule set, input the candidate rule set into a preset forward evaluation model for iterative optimization, and complete the rule optimization when the overall fitness of the fuzzy control rule base reaches the preset fitness threshold; The third unit is configured to input the working parameters acquired in real time into the optimized fuzzy control rule library, obtain a motor control signal through fuzzy reasoning operation, and output the motor control signal to the driving motor of the electric descending device; The fourth unit is used to collect actual output torque and motor speed data from the driving motor and calculate the dynamic response characteristics of the electric descending device; a fifth unit, configured to train a neural network dynamic response predictor using historical response data, input the dynamic response characteristics into the neural network dynamic response predictor for real-time prediction, calculate a compensation control amount based on the prediction result, and superimpose the compensation control amount on the motor control signal to obtain an optimized control signal; The sixth unit is used to drive the electric descent control device to perform a descent action based on the optimized control signal, and store the operation data in a historical database for subsequent optimization.

8. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the computer program instructions are executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images