Diamond synthesis process intelligent optimization control method for large-cavity cubic press

Abstract

The invention belongs to the technical field of diamond large-cavity synthesis, and particularly relates to an intelligent optimization control method for a diamond synthesis process of a large-cavity cubic press. The method comprises the steps of firstly collecting parameter indexes and ranges of pressure and temperature in a diamond synthesis process, then building an improved PID controller, predicting and adjusting a proportionality coefficient through a neural network, improving anti-integral saturation adjustment integral, and adaptively adjusting a differential time constant according to system uncertainty. And optimizing PID parameters by using an improved genetic algorithm, introducing an elitism strategy, and adaptively adjusting crossover and mutation probabilities. And finally, optimized controller parameters are obtained to carry out intelligent control on the pressure and the temperature. Compared with the prior art, the pressure and the temperature can be accurately controlled, the control precision and the system adaptability are improved, the diamond synthesis quality and efficiency are guaranteed, and the problems that a traditional control method is insufficient in control precision and poor in self-adaptive capacity are solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of diamond large cavity synthesis, and particularly relates to an intelligent optimization control method for the diamond synthesis process of a large cavity six-sided press. Background Art

[0002] In the field of diamond large cavity synthesis, a large cavity six-sided press is a commonly used device. When synthesizing diamonds, the precise control of pressure and temperature plays a decisive role in the synthesis effect. Currently, the existing control methods are difficult to meet the growing demand for high-quality diamond production. On the one hand, conventional control means cannot accurately and real-time sense the dynamic changes of pressure and temperature during the synthesis process, resulting in poor control accuracy. The diamonds synthesized are uneven in quality, affecting product performance and application range. On the other hand, in the face of complex and changeable synthesis environments and process requirements, traditional methods lack adaptability and intelligent adjustment capabilities, and cannot optimize control parameters in a timely manner, resulting in low production efficiency, serious energy waste, and increased production costs. Summary of the Invention

[0003] In view of the technical problems existing in the above background art, the present invention proposes an intelligent optimization control method for the diamond synthesis process of a large cavity six-sided press.

[0004] To achieve the above object, the technical solution adopted by the present invention is as follows, including the following steps:

[0005] S1. First, collect the pressure and temperature parameter indicators and their ranges during the diamond synthesis process;

[0006] S2. Build an improved PID controller, adjust the proportion based on a neural network, construct a multi-layer perceptron as a prediction neural network, predict the value at the next moment and then calculate the prediction deviation where r(k + 1) is the set value at the next moment, and the proportion coefficient adjustment amount ΔK p (k) is: where α is the adjustment factor, and the adjusted proportion coefficient K p (k) = K p0 + ΔK p (k), where K p0 is the initial proportion coefficient;

[0007] S3. Adjust the integral by improving anti-integral saturation, introduce an integral limiting coefficient β, and when the control output u(k) reaches the saturation value, judge the relationship between e(k) and the change direction of the control output: where K i is the integral coefficient, and T z is the sampling period;

[0008] S4. Then, perform adaptive differential adjustment to calculate the system uncertainty index δ(k) based on the deviation between the estimated state and the actual measurement value. where Y(k), represent the estimated state and the actual measurement value respectively, and the differential time constant T d (k) is adaptively adjusted according to the uncertainty index δ(k): T d (k) = T d0 ·(1 + γ·δ(k)), where T d0 is the initial differential time constant and γ is the adjustment coefficient. Then, the output D(k) of the differential link is:

[0009] S5. Obtain the optimal improved PID controller parameters through an improved genetic algorithm optimization method.

[0010] S6. Finally, obtain the controller parameters to perform intelligent control on the pressure and temperature during the diamond synthesis process.

[0011] Preferably, in step S2, the neural network model is constructed by selecting a multi-layer perceptron as the prediction neural network. The input layer receives the measurement values y(k - 1), y(k - 2), y(k - 3),..., y(k - n) and the corresponding deviations e(k - 1), e(k - 2), e(k - 3),..., e(k - n) at the past n moments, and the output layer outputs the predicted value for the next moment. In the forward propagation, the output z 1 (k) of the first hidden layer is: z 1 (k) = σ(W 1 ·[y(k - 1),..., y(k - n), e(k - 1),..., e(k - n)] T + b 1 ) where W 1 , b 1 are the weight matrix and bias of the first hidden layer respectively, σ is the activation function, and the subsequent output layer is: z i (k) = σ(W i ·z i-1 (k)+ b i ) i = 2, 3, then the output layer is where W 4 , b 4 are the weight matrix and bias of the output layer.

[0012] Preferably, the neural network has one input layer, three hidden layers, and one output layer.

[0013] Preferably, in step S4, the estimated state is obtained by solving the state equation, and its state equation is: wherein is the estimated state vector, is the estimated output, A, B, and C are system matrices, and L is the observer gain matrix.

[0014] Preferably, based on the above improvements, the formula for the optimized PID control output u(k) is: u(k) = K p (k)·e(k) + I(k) + D(k), where e(k) is the deviation at time k.

[0015] Preferably, during the evolution of the improved genetic algorithm in step S5, an elite retention strategy is introduced. The top 5% of individuals with the highest fitness in each generation are directly retained in the next generation. At the same time, the crossover probability and mutation probability are adaptively adjusted according to the generation number of the algorithm. From generation 0 to 20, the crossover probability is set to 0.9 and the mutation probability is set to 0.05 to accelerate the search speed of the algorithm and expand the search range. As the generation number increases, from generation 21 to 30, the crossover probability is reduced to 0.6 and the mutation probability is 0.02 to focus more on local search, and finally the parameters of the improved PID controller are obtained.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are as follows. In terms of the control method, an improved PID controller is built, and the proportional coefficient is adjusted based on neural network prediction, which can accurately follow the changes in pressure and temperature and make a rapid response; the anti-integral saturation measure is improved to avoid excessive accumulation of the integral term and improve the control accuracy; the adaptive differential adjustment changes the differential time constant according to the system uncertainty, enhancing the adaptability of the system to complex environments. In terms of parameter optimization, an improved genetic algorithm is adopted to perform global search by simulating natural evolution, an elite retention strategy is introduced, and the crossover and mutation probabilities are adaptively adjusted to quickly screen out the optimal PID parameters suitable for the dynamic characteristics of the synthesis process, thereby ensuring the quality and efficiency of diamond synthesis. Detailed implementation manners

[0017] In order to more clearly understand the above objects, features, and advantages of the present invention, the present invention will be further described below with reference to the embodiments. It should be noted that, without conflict, the embodiments of the present application and the features in the embodiments may be combined with each other.

[0018] In the following description, many specific details are set forth in order to fully understand the present invention. However, the present invention may be implemented in other ways different from those described herein. Therefore, the present invention is not limited by the specific embodiments disclosed in the following specification.

[0019] Embodiment. In the field of diamond large cavity synthesis, the large cavity six - sided top press is a commonly used device. The diamond synthesis process has extremely high requirements for the precise control of pressure and temperature. Any slight deviation may lead to a decline in the quality of synthesized diamonds. Existing control methods have problems such as insufficient control accuracy and poor adaptability. Therefore, the present invention proposes an intelligent optimization control method for the diamond synthesis process of a large cavity six - sided top press.

[0020] Adopt the new structure of the 900 - type large press and the synthesis block combination. Use the 220 - type anvil, with an oblique angle of 41.5, the width of the hypotenuse of the anvil is 20, the side length of the top surface of the anvil is 80, and it is paired with a 95 - mm synthesis block. The 95 - mm synthesis block takes the maximum negative tolerance. The wall thickness of the pyrophyllite block is 10.5, the thickness of the liner is 5.5, the thickness of the sealing edge is 14, and it is combined with a dolomite ring of 7.5 and a pyrophyllite ring of 9. In order to achieve precise control of the diamond synthesis process, it is first necessary to collect the pressure and temperature parameter indicators and their ranges during the synthesis process. These parameters are the inputs of the control system and directly affect the adjustment and optimization of the controller.

[0021] Then, considering that the proportional - integral - derivative (PID) controller uses a fixed coefficient in its proportional link and cannot be adjusted in real - time according to the dynamic changes in the synthesis process. When facing rapid fluctuations in pressure and temperature, the response lags and large deviations are easily generated. The integral link lacks an effective anti - integral saturation mechanism. When the control output reaches saturation, the integral term continues to accumulate, resulting in serious overshoot of the system and affecting the synthesis accuracy and stability. The derivative link has poor adaptability to system uncertainties and is difficult to adjust the derivative time constant according to the actual situation, easily amplifying noise interference and reducing the control effect. In contrast, the improved PID controller of the present invention has significant advantages. By predicting and adjusting the proportional coefficient through a neural network, it can accurately follow the changing trends of pressure and temperature, make a rapid response, and reduce deviations. The improved anti - integral saturation measure effectively avoids excessive accumulation of the integral term, ensures the stable operation of the system, and improves the control accuracy. The adaptive derivative adjustment dynamically changes the derivative time constant according to system uncertainties. While suppressing noise, it can timely capture the small changes in pressure and temperature, enhance the system's adaptability to complex synthesis environments, comprehensively improve the control performance, and ensure the quality and efficiency of diamond synthesis. Build an improved PID controller, adjust the proportion based on a neural network, construct a multi - layer perceptron as the prediction neural network. The neural network has an input layer, three hidden layers, and an output layer. Select a multi - layer perceptron as the prediction neural network. The input layer receives the measured values y(k - 1), y(k - 2), y(k - 3)..., y(k - n) and the corresponding deviations e(k - 1), e(k - 2), e(k - 3)..., e(k - n) at the past n moments, and the output layer outputs the predicted value for the next moment. In the forward propagation, the output z 1 (k) is: z 1 (k) = σ(W1 · [y(k - 1),..., y(k - n), e(k - 1),..., e(k - n)] T + b 1 ) where W 1 , b 1 are the weight matrix and bias of the first hidden layer respectively, σ is the activation function, and the subsequent output layer is: z i (k) = σ(W i · z i-1 (k) + b i ) If i = 2, 3, then the output layer is where W 4 , b 4 are the weight matrix and bias of the output layer. After predicting the value at the next moment , calculate the prediction deviation where r(k + 1) is the set value at the next moment, and the proportional coefficient adjustment amount ΔK p (k) is:[[]] where α is the adjustment factor, then the adjusted proportional coefficient K p (k) = K p0 + ΔK p (k), where K p0 is the initial proportional coefficient. Then, adjust the integral by improving the anti - integral saturation. Introduce the integral limiter coefficient β. When the control output u(k) reaches the saturation value, judge the relationship between e(k) and the change direction of the control output: where K i is the integral coefficient, and T z is the sampling period. Finally, perform adaptive differential adjustment. Calculate the system uncertainty index δ(k) according to the deviation between the estimated state and the actual measured value where Y(k), represent the estimated state and the actual measured value respectively. The differential time constant T d (k) is adaptively adjusted according to the uncertainty index δ(k): T d (k) = T d0 · (1 + γ· δ(k)), where T d0 is the initial differential time constant, and γ is the adjustment coefficient. Then, the output D(k) of the differential link is: where the estimated state is obtained by solving the state equation, and its state equation is: where is the estimated state vector, is the estimated output, A, B, C are system matrices, L is the observer gain matrix, is the first - order derivative of the state vector with respect to time. Combining the above improvements, the formula for the optimized PID control output u(k) is: u(k) = Kp (k)·e(k)+I(k)+D(k), where e(k) is the deviation at time k.

[0022] Next, considering the traditional methods for determining PID parameters, such as the trial-and-error method, which rely on manual experience, consume a large amount of time and effort, and it is difficult to find the optimal solution, and have poor adaptability in the face of complex and changing synthesis environments. The improved genetic algorithm has many advantages. It simulates the natural evolution process and performs global search in the parameter space through selection, crossover, and mutation operations. It can quickly screen out better solutions from a large number of possible parameter combinations and avoid falling into local optima. The elite retention strategy is introduced in the improved genetic algorithm to ensure that the excellent parameters of each generation are retained; the crossover probability and mutation probability are adaptively adjusted to speed up the search speed in the early stage of the algorithm and focus on local optimization in the later stage. Finally, the obtained PID parameters can better match the dynamic characteristics of the synthesis process. Initialize the population. Extract 30% of the control parameter combinations from the historical synthesis data as part of the initial individuals, and randomly generate the remaining 70% of the individuals to form an initial population of 100 individuals. Use the tournament selection method. Randomly select 5 individuals from the population each time, and select the individual with the best fitness to enter the next generation. The fitness function is comprehensively determined according to indicators such as the pressure and temperature control accuracy, stability, and diamond synthesis quality in the synthesis process. In the evolutionary generations 0 - 20, set the crossover probability to 0.9 and the mutation probability to 0.05 to speed up the search speed of the algorithm and expand the search range; starting from the 21st generation, reduce the crossover probability to 0.6 and the mutation probability to 0.02 to focus more on local search. Directly retain the individuals ranked in the top 5% of fitness in each generation to the next generation to ensure that excellent individuals are not eliminated during the evolution process. After 30 generations of iteration, obtain the optimal parameters of the improved PID controller.

[0023] Finally, obtain the controller parameters to intelligently control the pressure and temperature in the diamond synthesis process.

[0024] The above is only a preferred embodiment of the present invention, and it is not a limitation of the present invention in other forms. Any person skilled in the art may use the disclosed technical content to make changes or modifications into equivalent embodiments with equivalent changes and apply them to other fields. However, as long as it does not depart from the technical solution content of the present invention, any simple modification, equivalent change, and modification made to the above embodiments based on the technical essence of the present invention still belong to the protection scope of the technical solution of the present invention.

Claims

1. An intelligent optimization control method for diamond synthesis process of a large cavity six-sided top press, characterized in that: The following steps are involved: S1. First, collect the pressure and temperature parameter indicators and the range of the parameter indicators during the diamond synthesis process; S2. Build an improved PID controller, adjust the ratio based on the neural network, build a multilayer perceptron as a prediction neural network, and predict the value at the next moment Then, calculate the prediction deviation Among them, r(k+1) is the set value at the next moment, and the proportional coefficient adjustment amount ΔK p (k) is: Where α is the adjustment factor, then the adjusted proportional coefficient K p (k) = K p0 +ΔK p (k), where K p0 is the initial proportionality factor; S3. By improving the anti-integral saturation, the integral is adjusted and the integral limit coefficient β is introduced. When the control output u(k) reaches the saturation value, the relationship between e(k) and the direction of change of the control output is determined: Where K i is the integration coefficient, T z is the sampling period; S4, then perform adaptive differential adjustment, and calculate the uncertainty index δ(k) of the system according to the deviation between the estimated state and the actual measured value. Where Y(k), Represent the estimated state and the actual measured value respectively, and the differential time constant T d (k) Adaptive adjustment based on uncertainty indicator δ(k): T d (k) = T d0 ·(1+γ·δ(k)), where T d0 is the initial differential time constant, γ is the adjustment coefficient, then the output D(k) of the differential link is: S5, obtaining the optimal improved PID controller parameters by improving the genetic algorithm optimization method; S6. Finally, controller parameters are obtained to intelligently control the pressure and temperature during the diamond synthesis process.

2. The intelligent optimization control method for diamond synthesis process of a large cavity six-sided top press according to claim 1 is characterized in that: The neural network model in step S2 is constructed by selecting a multilayer perceptron as a prediction neural network, the input layer receives the measured values ​​y(k-1), y(k-2), y(k-3)..., y(kn) and the corresponding deviations e(k-1), e(k-2), e(k-3)..., e(kn) at the past n moments, and the output layer outputs the predicted value at the next moment The output z1(k) of the first hidden layer in the forward propagation is: z1(k) = σ(W1 · [y(k-1), ..., y(kn), e(k-1), ..., e(kn)] T +b1) where W1, b1 are the weight matrix and bias of the first hidden layer, σ is the activation function, and the subsequent output layer is: z i (k)=σ(W i ·z i-1 (k)+b i )i=2,3, then the output layer is Where W4, b4 are the weight matrix and bias of the output layer.

3. The intelligent optimization control method for diamond synthesis process of a large cavity six-sided top press according to claim 2 is characterized in that: The neural network has one input layer, three hidden layers and one output layer.

4. The intelligent optimization control method for diamond synthesis process of a large cavity six-sided top press according to claim 1 is characterized in that: The estimated state in step S4 is obtained by solving the state equation, which is: in To estimate the state vector, is the estimated output, A, B, C are the system matrices, and L is the observer gain matrix.

5. The intelligent optimization control method for diamond synthesis process of a large cavity six-sided top press according to claim 1 is characterized in that: Based on the above improvements, the formula for the optimized PID control output u(k) is: u(k)=K p (k)·e(k)+I(k)+D(k), where e(k) is the deviation at time k.

6. The intelligent optimization control method for diamond synthesis process of a large cavity six-sided top press according to claim 1 is characterized in that: The step S5 improves the genetic algorithm in the evolution process, introduces an elite retention strategy, and directly retains the top 5% individuals in fitness in each generation to the next generation. At the same time, the crossover probability and mutation probability are adaptively adjusted according to the evolutionary algebra of the algorithm. From the evolutionary algebra 0 to 20, the crossover probability is set to 0.9 and the mutation probability is set to 0.05 to speed up the search speed of the algorithm and expand the search range; as the evolutionary algebra increases, from the 21st generation to the 30th generation, the crossover probability is reduced to 0.6 and the mutation probability is reduced to 0.02, focusing more on local search, and finally obtaining improved PID controller parameters.

Images