Rotary kiln rotating speed detection and intelligent control method and device

Abstract

The invention relates to rotating speed detection and intelligent control of a rotary kiln, and discloses an intelligent rotating speed detection and control method and device of the rotary kiln, the head part and the tail part of the rotary kiln are respectively provided with a rotating speed adjusting device 1 and a rotating speed adjusting device 2, and are correspondingly provided with an acceleration sensor 1 and an acceleration sensor 2; a current sensor 1 and a current sensor 2 are connected to the rotating speed adjusting device 1 and the rotating speed adjusting device 2, a rotating speed sensor is further arranged on the rotary kiln, and the rotary kiln rotating synchronous control system is jointly formed and used for synchronously adjusting rotation of the rotary kiln. Detection feedback values are obtained after rotating speed and current detection values pass through a parameter detection module, three parameters which interact, influence and correlate with one another of the rotating speed, the current and the acceleration are synchronously and accurately adjusted through a rotating speed controller, and coordination, synchronism and robustness of rotating speed control of the rotary kiln are improved.

Description

Technical Field

[0001] The present invention relates to the field of rotational speed detection and intelligent control of rotary kilns, and particularly to an intelligent detection and control method and device for the rotational speed of a rotary kiln. Background Art

[0002] The existing methods for controlling the rotational speed of rotary kilns mainly adjust the rotational speed of rotary kilns through simple means such as PID controllers, PI controllers, and single rotary kiln adjustment devices. There are problems such as low control accuracy and easy skew, making it difficult to control the rotational speed at the target value and achieve stable control, which will affect the production efficiency of rotary kilns. On the other hand, it also seriously affects the production safety of rotary kilns. At the same time, with the development of big data and intelligent control technologies, the development of rotational speed detection and intelligent control of rotary kilns towards intensification and intelligence has become an inevitable trend. Summary of the Invention

[0003] Object of the Invention: Aiming at the problems in the background art, the present invention provides an intelligent detection and control method and device for the rotational speed of a rotary kiln, which can more accurately detect and control the rotational speed of the rotary kiln, ensure the production safety of the rotary kiln, and improve production efficiency.

[0004] Technical Solution: The present invention discloses an intelligent detection and control method for the rotational speed of a rotary kiln, which includes the following steps:

[0005] Step 1: Rotational speed adjustment devices 1 and 2 are respectively arranged at the head and tail of the rotary kiln to synchronously adjust the rotation of the rotary kiln, and acceleration sensors 1 and 2 are correspondingly arranged to detect the acceleration values of the head and tail of the rotary kiln; current sensors 1 and 2 are connected to the rotational speed adjustment devices 1 and 2 to respectively detect the current values of the rotational speed adjustment devices 1 and 2, and a rotational speed sensor is further arranged on the rotary kiln to detect the rotational speed of the rotary kiln;

[0006] Step 2: The values detected by the rotational speed sensor, current sensor 2, and current sensor 1 are respectively output as the detection feedback values of the speed sensor, current sensor 2, and current sensor 1 after passing through parameter detection modules 1 - 3;

[0007] Step 3: A rotational speed control system for the rotary kiln is constructed, including a rotational speed controller, differential unit 2, differential unit 3, parameter detection modules 1 - 3, GWO's NARX neural network model 3 - 4, GWO's fuzzy recursive neural network model - PID controller 1 - 4, GWO's NARX neural network model - PID controller 2 - 3, and GWO's fuzzy recursive neural network model - GWO's BiGRU neural network model 3 - 4;

[0008] The rotational speed target value and the rotational speed detection feedback value output by the parameter detection module 1 are used as the inputs of the rotational speed controller. The errors and the rate of change of errors between the output of the rotational speed controller and the output of the acceleration sensor 1 are used as the inputs of the GWO fuzzy recurrent neural network model-PID controller 2. The current change rate target value is 0, and the difference and the rate of change of the difference between the current change rate target value and the output of the differential unit 3 are used as the inputs of the GWO fuzzy recurrent neural network model-PID controller 1. The outputs of the GWO fuzzy recurrent neural network model-PID controller 1, the GWO fuzzy recurrent neural network model-PID controller 2, the GWO NARX neural network model-PID controller 2, and the GWO NARX neural network model-PID controller 3 are used as the inputs of the GWO fuzzy recurrent neural network model-GWO BiGRU neural network model 3. The output of the GWO fuzzy recurrent neural network model-GWO BiGRU neural network model 3 is used as the input of the rotational speed adjustment device 1;

[0009] The errors and the rate of change of errors between the output of the rotational speed controller and the output of the acceleration sensor 2 are used as the inputs of the GWO fuzzy recurrent neural network model-PID controller 4. The current change rate target value is 0 and the difference and the rate of change of the difference between it and the output of the differential unit 2 are used as the inputs of the GWO fuzzy recurrent neural network model-PID controller 3. The outputs of the GWO fuzzy recurrent neural network model-PID controller 3, the GWO fuzzy recurrent neural network model-PID controller 4, the GWO NARX neural network model-PID controller 2, and the GWO NARX neural network model-PID controller 3 are used as the inputs of the GWO fuzzy recurrent neural network model-GWO BiGRU neural network model 4. The output of the GWO fuzzy recurrent neural network model-GWO BiGRU neural network model 4 is used as the input of the rotational speed adjustment device 2;

[0010] The difference between the outputs of the acceleration sensor 1 and the acceleration sensor 2 and the change rate of the difference are used as the inputs of the GWO's NARX neural network model - PID controller 2. The difference between the outputs of the current sensor 1 and the current sensor 2 and the change rate of the difference are used as the inputs of the GWO's NARX neural network model - PID controller 3. The outputs of the acceleration sensor 1 and the acceleration sensor 2 are respectively used as the inputs of the GWO's NARX neural network model 3 and the GWO's NARX neural network model 4. The output of the rotational speed sensor is used as the input of the GWO's NARX neural network model 3, the GWO's NARX neural network model 4, and the parameter detection module 1. The outputs of the GWO's NARX neural network model 3 and the GWO's NARX neural network model 4 are used as the inputs of the parameter detection module 1. The outputs of the current sensor 1 and the current sensor 2 are respectively used as the inputs of the parameter detection module 3 and the parameter detection module 2. The outputs of the parameter detection module 3 and the parameter detection module 2 are respectively used as the inputs of the differential unit 3 and the differential unit 2.

[0011] Furthermore, the structures of the parameter detection modules 1 - 3 are the same, and each includes the GWO's NARX neural network model - GWO's BiGRU neural network model, the GWO's BiGRU neural network model 1, and the GWO's fuzzy recurrent neural network model - GWO's BiGRU neural network model 1. The output of the rotational speed sensor or the current sensor 2 or the current sensor 1 is respectively used as the input of the GWO's NARX neural network model - GWO's BiGRU neural network model, the GWO's BiGRU neural network model 1, and the GWO's fuzzy recurrent neural network model - GWO's BiGRU neural network model 1. The output of the GWO's NARX neural network model - GWO's BiGRU neural network model is used as the input of the GWO's fuzzy recurrent neural network model - GWO's BiGRU neural network model 1. The output of the GWO's BiGRU neural network model 1 is used as the input of the GWO's fuzzy recurrent neural network model - GWO's BiGRU neural network model 1. The output of the GWO's fuzzy recurrent neural network model - GWO's BiGRU neural network model 1 is used as the detection feedback value of the output of the rotational speed sensor or the current sensor 2 or the current sensor 1.

[0012] Furthermore, the rotational speed controller includes a differential unit 1, the GWO's NARX neural network model - PID controller 1, the GWO's NARX neural network model 1 - 2, the GWO's fuzzy recurrent neural network model - GWO's BiGRU neural network model 2, and the GWO's BiGRU neural network model 2;

[0013] The rotational speed target value serves as the input of the differential unit 1. The difference between the rotational speed target value and the rotational speed detection feedback value and the change rate of the difference are used as the inputs of the GWO's NARX neural network model - PID controller 1. The outputs of the differential unit 1, the GWO's NARX neural network model - PID controller 1, and the GWO's fuzzy recursive neural network model - GWO's BiGRU neural network model 2 are used as the inputs of the GWO's NARX neural network model 1. The sum of the output of the GWO's NARX neural network model 1 and the output of the GWO's NARX neural network model - PID controller 1 is used as the input of the GWO's BiGRU neural network model 2. The rotational speed detection feedback value is used as the input of the GWO's NARX neural network model 2. The difference between the rotational speed detection feedback value and the output of the GWO's NARX neural network model 2 and the change rate of the difference, and the difference between the rotational speed target value and the rotational speed detection feedback value and the change rate of the difference are used as the inputs of the GWO's fuzzy recursive neural network model - GWO's BiGRU neural network model 2. The sum of the output of the differential unit 1, the output of the GWO's BiGRU neural network model 2, and the output of the GWO's fuzzy recursive neural network model - GWO's BiGRU neural network model 2 is used as the output of the rotational speed controller.

[0014] Further, the structures of the GWO's fuzzy recursive neural network model - PID controller 1 - 4 are the same. They are all composed of the GWO's fuzzy recursive neural network model in series with the PID controller. The three parameters output by the GWO's fuzzy recursive neural network model are used as the inputs of the PID controller to achieve online adjustment of the parameters of the PID controller. The GWO's fuzzy recursive neural network model is a fuzzy recursive neural network model with optimized parameters by the GWO algorithm.

[0015] Further, the structures of the GWO's fuzzy recursive neural network model - GWO's BiGRU neural network model 1, the GWO's fuzzy recursive neural network model - GWO's BiGRU neural network model 2, and the GWO's fuzzy recursive neural network model - GWO's BiGRU neural network model 3 - 4 are the same. They are all composed of the GWO's fuzzy recursive neural network model in series with the GWO's BiGRU neural network model. The output of the GWO's fuzzy recursive neural network model is used as the input of the GWO's BiGRU neural network model. The GWO's fuzzy recursive neural network model and the GWO's BiGRU neural network model are respectively fuzzy recursive neural network models and BiGRU neural network models with optimized parameters by the GWO algorithm.

[0016] Further, the structures of the GWO-based NARX neural network model - PID controllers 1 - 3 are the same, all of which are in series connection of the GWO-based NARX neural network model and the PID controller, with the output of the GWO-based NARX neural network model as the input of the PID controller. The GWO-based NARX neural network model is to optimize the parameters of the NARX neural network model by the GWO algorithm.

[0017] Further, the differential unit 1 takes the rotational speed target values at multiple sampling times as the input, and determines the change rate of the rotational speed target by dividing the difference between the rotational speed targets at adjacent sampling times by the sampling time.

[0018] Further, after obtaining the current feedback value of the current sensor 2 output by the parameter detection module 2, the differential unit 2 determines the current change rate of the current sensor 2 by dividing the difference between the current feedback value and the feedback value at the previous sampling moment by the sampling time; after obtaining the current feedback value of the current sensor 1 output by the parameter detection module 3, the differential unit 3 determines the current change rate of the current sensor 1 by dividing the difference between the current feedback value and the feedback value at the previous sampling moment by the sampling time.

[0019] The present invention also discloses a rotary kiln rotational speed detection and intelligent control device, including an STM32 microprocessor, an acceleration sensor, a current sensor, a rotational speed sensor, a rotational speed regulating device, and a field control terminal. The acceleration sensor, the current sensor, and the rotational speed sensor are respectively connected to the STM32 microprocessor through corresponding conditioning circuits; communication between the STM32 microprocessor and the field control terminal is realized through a USB interface. The STM32 microprocessor realizes data acquisition and preprocessing. The field control terminal is an industrial control computer, and a rotary kiln rotational speed control system is set in the field control terminal to implement the above-mentioned rotary kiln rotational speed detection and intelligent control method to monitor the rotational speed of the rotary kiln.

[0020] Beneficial effects:

[0021] 1. The neural network model of the rotary kiln parameters optimized by the GWO optimization algorithm can effectively explore the search space of the neural network model of the rotary kiln parameters, find the global optimal solution of the neural network model of the rotary kiln parameters, and can quickly converge to the optimal solution of the neural network model of the rotary kiln parameters, with a relatively high convergence speed. Therefore, the parameters of the BiGRU neural network model, NARX neural network model, and fuzzy recursive neural network model of the rotary kiln parameters are randomly generated and updated by the grey wolf optimization algorithm, so as to more quickly obtain their optimal solutions and further improve the accuracy and robustness of their prediction and control of the rotary kiln parameters.

[0022] Second, the BiGRU neural network model of GWO can capture the time-series features of the rotary kiln parameters bidirectionally, and can simultaneously capture the bidirectional dependence relationship between the historical parameters and future parameter changes of the rotary kiln, improving the utilization efficiency of the rotary kiln parameter features and the accuracy and robustness of the prediction of the BiGRU neural network model of GWO.

[0023] Third, the input of the NARX neural network model of GWO has the time-delay orders of the input and output, which can better describe the influence of the time-series parameters of the rotary kiln parameters on the NARX neural network model of GWO. The NARX neural network model of GWO can better describe the characteristics of the time series of the rotary kiln parameter changes. An NARX neural network model is established, and the GWO algorithm is used to optimize the weights and thresholds of the NARX neural network model to achieve the optimization of the NARX neural network model of GWO. The parameters in the operation process of the rotary kiln are used as the input of the NARX neural network model of GWO to accurately predict the rotary kiln parameters at the next moment, improving the accuracy, robustness and reliability of predicting and controlling the rotary kiln parameters.

[0024] Fourth, the fuzzy recurrent neural network model of GWO combines the learning ability of the neural network and the interpretability of the rule-based fuzzy system, and is widely used in the modeling of nonlinear systems. Its core is to add a recurrent link in the fuzzy neural network rule layer, and its function is to timely and dynamically feedback and save the rotary kiln parameter information, forming a recurrent unit structure that effectively covers the fuzzy recurrent neural network model of GWO. It can be seen from this that the output of the fuzzy recurrent neural network model of GWO is jointly determined by the current rotary kiln parameters of the fuzzy recurrent neural network model of GWO, the input values of the previous rotary kiln parameters, and the output rotary kiln parameter values of the previous fuzzy recurrent neural network model of GWO. Therefore, the fuzzy recurrent neural network model of GWO can be used to handle the prediction and control problems of inaccurate and uncertain rotary kiln parameters, improving the accuracy, robust characteristics and reliability of predicting and controlling the rotary kiln parameters.

[0025] Fifth, the NARX neural network model of GWO can better describe the characteristics of the time series of the rotary kiln parameter changes. Therefore, the NARX neural network model-PID controller of GWO uses the three parameters output by the NARX neural network model of GWO to tune the parameters of the PID controller, improving the anti-interference ability, robustness and the stability of the rotary kiln speed of the NARX neural network model-PID controller of GWO.

[0026] 6. The output of the fuzzy recurrent neural network model using GWO features is determined by the current rotary kiln parameters, the input values of the previous rotary kiln parameters, and the output values of the previous rotary kiln parameters of the GWO-based fuzzy recurrent neural network model. The GWO-based fuzzy recurrent neural network model-PID controller tunes the PID controller parameters using the three parameters output by the GWO-based fuzzy recurrent neural network model, improving the anti-interference ability, robustness, and stability of the rotary kiln speed of the GWO-based fuzzy recurrent neural network model-PID controller.

[0027] 7. The rotary kiln speed control system consists of a speed controller, speed adjustment device 1, and speed adjustment device 2 to form a rotary kiln rotation synchronization control system. By jointly and precisely adjusting the three parameters of speed, current, and acceleration of the adjustment device, which interact, influence, and are related to each other, the coordination, synchronization, and robustness of the rotary kiln speed control are improved.

[0028] 8. The GWO-based NARX neural network model-PID controller 1, GWO-based NARX neural network model-PID controller 2, and GWO-based NARX neural network model-PID controller 3 of the rotary kiln speed control system use the delay characteristics of the GWO-based NARX neural network model to keep the PID parameters stable. The GWO-based fuzzy recurrent neural network model-PID controller 1-4 uses the adaptability and robustness of the GWO-based fuzzy recurrent neural network model to achieve the fast self-adaptability of the PID parameters. The interaction, influence, and complementarity between the GWO-based NARX neural network model-PID controller 1-3 and the GWO-based fuzzy recurrent neural network model-PID controller 1-4 keep the speed adjustment device stable, fast, and adaptive, improving the robustness and accuracy of the rotary kiln speed control.

[0029] 9. The speed controller and parameter detection module 1 constitute the overall control of the rotary kiln speed. The GWO-based fuzzy recurrent neural network model-PID controller 1-2, parameter detection module 3, and GWO-based fuzzy recurrent neural network model-GWO's BiGRU neural network model 3 achieve the control of speed adjustment device 1. The GWO-based fuzzy recurrent neural network model-PID controller 3-4, parameter detection module 2, and GWO-based fuzzy recurrent neural network model-GWO's BiGRU neural network model 4 achieve the control of speed adjustment device 2. In this way, the overall control of the rotary kiln speed and the individual control of the two adjustment devices are combined, giving full play to the respective advantages of the overall and partial controls. The overall control remains stable, and the partial control is fast, achieving the combination of stability, robustness, and rapidity of the rotary kiln control.

[0030] X. GWO's NARX neural network model - PID controller 2 - 3 realizes the balanced control of two speed regulation devices 1 - 2, prevents the rotary kiln from skewing, and keeps the rotary kiln running stably. Description of the Drawings

[0031] Figure 1 It is a schematic structural diagram of the parameter detection module in the embodiment of the present invention;

[0032] Figure 2 It is a schematic structural diagram of the speed controller in the embodiment of the present invention;

[0033] Figure 3 It is a schematic structural diagram of the rotary kiln speed control system in the embodiment of the present invention;

[0034] Figure 4 It is a schematic structural diagram of the rotary kiln speed detection and intelligent control device in the embodiment of the present invention. Detailed Embodiment

[0035] The following further describes the present invention with reference to the drawings.

[0036] The embodiment of the present invention provides a method and device for rotary kiln speed detection and intelligent control, which specifically includes the following steps:

[0037] I. Design of the Parameter Detection Module

[0038] In this embodiment, the parameter detection module 1, parameter detection module 2, and parameter detection module 3 have the same structure, including the GWO's NARX neural network model - GWO's BiGRU neural network model, GWO's BiGRU neural network model 1, and GWO's fuzzy recursive neural network model - GWO's BiGRU neural network model 1. The structural diagram of the parameter detection module is shown in Figure 1 as shown.

[0039] 1. Design of the GWO's BiGRU Neural Network Model

[0040] The advantages of the BiGRU neural network model of GWO are mainly reflected in its processing ability and computational efficiency for the time-series data of rotary kiln parameters. The BiGRU neural network model of GWO can capture both past and future information of the time-series data of rotary kiln parameters, and utilize the past and future information of rotary kiln parameters to better capture the time-series characteristics of rotary kiln parameters. This enables the BiGRU neural network model of GWO to perform excellently when dealing with rotary kiln parameters with complex dependencies. The gating mechanism in the BiGRU neural network model of GWO can effectively ensure the stability of long-term dependent data of rotary kiln parameters, thus avoiding the problem of loss or attenuation of rotary kiln parameter information in long sequences. When capturing the time-series characteristics of rotary kiln parameters, the BiGRU neural network model of GWO can improve computational efficiency while maintaining the accuracy of predicting rotary kiln parameters. The BiGRU neural network model of GWO combines the forward and backward propagation of rotary kiln parameters, and simultaneously considers the past and future time-series data of rotary kiln parameters, more effectively capturing the dependencies in the time series of rotary kiln parameters and significantly improving the time-series data modeling ability of rotary kiln parameters. The current hidden state of the BiGRU neural network model of GWO is jointly determined by the outputs of the forward and backward propagation hidden layers and the input at the current moment.

[0041] The formula of the BiGRU neural network model of GWO is as follows:

[0042]

[0043] x t represents the input rotary kiln parameters, w t is the output weight of the hidden layer of the GRU neural network model for the forward propagation of rotary kiln parameter information at time t, v t is the output weight of the hidden layer of the GRU neural network model for the backward propagation of rotary kiln parameter information at time t; b t represents the bias corresponding to the hidden

[0044] →← state of the BiGRU neural network model of GWO at time t, h t-1 represents the output result of the forward propagation hidden state of rotary kiln parameter information, h t-1 represents the output result of the backward propagation hidden layer state of rotary kiln parameter information, and GRU() is the gated recurrent unit.

[0045] The BiGRU neural network model of GWO is the BiGRU neural network model optimized by the grey wolf algorithm. There are parameters in the BiGRU neural network model that can be optimized and selected. The parameters that have the greatest impact on the BiGRU neural network model are the learning rate, the maximum number of iterations, and the number of neurons in the hidden layer in turn. If the learning rate is too small, it may prolong the training cycle of the BiGRU neural network model; if it is too large, it may hinder the convergence of the BiGRU neural network model. The number of neurons in the BiGRU neural network model directly affects the learning ability of the model and the complexity of the network. Too many nodes in the BiGRU neural network model will increase the training time of the BiGRU neural network model, and too few will damage the performance of the BiGRU neural network model.

[0046] The present invention uses the grey wolf optimization algorithm to randomly generate and update the parameter combination of the BiGRU neural network model, and can more quickly obtain the optimal solution of the parameters of the BiGRU neural network model. There are convergence factors and information feedback mechanisms that can be adaptively adjusted in the grey wolf optimization algorithm, which can achieve a balance between local optimization and global search. The grey wolf population is composed of four different levels of grey wolves, namely α wolves, β wolves, δ wolves, and ω wolves. The α wolf is the optimal solution in the optimization algorithm; the β wolf is the sub-optimal solution in the optimization algorithm; the δ wolf obeys the orders of the α wolf and the β wolf; the ω wolf is an ordinary member and updates its position around the α wolf, β wolf, and δ wolf. The hunting process of grey wolves includes surrounding the prey, chasing the prey, and attacking the prey. The specific steps to optimize the BiGRU neural network model by improving the grey wolf algorithm are as follows:

[0047] (1) Establish a BiGRU neural network model. Determine the neural network structure and initialize the weights and thresholds;

[0048] (2) Initialize the relevant parameters of GWO. Set parameters such as the population size of the wolf pack, the maximum number of iterations, and the upper and lower bounds;

[0049] (3) Calculate the fitness value, sort it from large to small, and update the positions and parameters of all grey wolves;

[0050] (4) Obtain the error between the training samples and the test samples and the position of the leading wolf α corresponding thereto;

[0051] (5) Determine whether the set error is satisfied or the maximum number of iterations is reached. If not, repeat steps (3)-(5) until the condition is satisfied;

[0052] (6) Feed back the parameters of the BiGRU neural network model with the optimal solution to the BiGRU neural network model.

[0053] In the grey wolf optimization (GWO) algorithm for optimizing the BiGRU neural network model, the number of grey wolves is 20, and the maximum number of iterations is 50. The optimization ranges of the number of hidden layer neurons, the maximum number of training times, and the learning rate of the GWO algorithm for the BiGRU neural network model are [20, 150], [150, 300], and [0.001, 0.01], respectively.

[0054] 2. GWO's NARX neural network model - Design of GWO's BiGRU neural network model

[0055] GWO's NARX neural network model - GWO's BiGRU neural network model is formed by connecting GWO's NARX neural network model and GWO's BiGRU neural network model in series, with the output of GWO's NARX neural network model serving as the input of GWO's BiGRU neural network model. GWO's NARX neural network model is a non - linear autoregressive network with external input and belongs to the regression neural network. The commonly used NARX neural network includes an input layer, the time - delay order of the input layer, a hidden layer, an output layer, and the time - delay order of the output layer.

[0056] GWO's NARX neural network model can be expressed as:

[0057] y(t)=f[y(t - 1),y(t - 2),…,y(t - n y ),u(t),u(t - 1),u(t - 2),…,u(t - n u )] (4)

[0058] u(t) and y(t) are the input and output of GWO's NARX neural network model at time t, n u and n y are the time - delay orders of the input and output of the NARX neural network model respectively, and f is the non - linear function fitted by GWO's NARX neural network model. Since the input of GWO's NARX neural network model has time - delay orders of input and output, it can better describe the influence of the time - series parameters of the rotary kiln parameters on GWO's NARX neural network model. On the one hand, GWO's NARX neural network model can follow the continuity of the change of rotary kiln parameters, thus inferring the change trend of rotary kiln parameters; on the other hand, through the processing of the historical data of the rotary kiln and considering the influence of the input factors of the rotary kiln, it is convenient to establish a dynamic system model for the rotary kiln parameters affected by multiple factors through GWO's NARX neural network model. GWO's NARX neural network model is used to predict the rotary kiln parameters, and it can more comprehensively consider the various factors that the rotary kiln parameters may be affected during the operation of the rotary kiln.

[0059] The NARX neural network model of GWO is to optimize the NARX neural network model using the GWO algorithm. The methods and steps for optimizing the NARX neural network model using the GWO algorithm refer to the design method of optimizing the BiGRU neural network model using the GWO algorithm.

[0060] 3. The fuzzy recurrent neural network model of GWO - The design of the BiGRU neural network model of GWO

[0061] The fuzzy recurrent neural network model of GWO - The BiGRU neural network model 1 of GWO is the fuzzy recurrent neural network model of GWO - the BiGRU neural network model of GWO, which is the series connection of the fuzzy recurrent neural network model of GWO and the BiGRU neural network model of GWO. The output of the fuzzy recurrent neural network model of GWO is used as the input of the BiGRU neural network model of GWO.

[0062] The core of the fuzzy recurrent neural network model of GWO is to add a recurrent link in the rule layer of the fuzzy neural network. Its function is to timely and dynamically feedback and save the rotary kiln parameter information, forming a recursive unit structure that effectively covers the fuzzy recurrent neural network model of GWO. It can be seen that the output of the fuzzy recurrent neural network model of GWO is jointly determined by the current rotary kiln parameters of the fuzzy recurrent neural network model of GWO, the input values of the previous rotary kiln parameters, and the rotary kiln parameter values output by the previous fuzzy recurrent neural network model of GWO. Therefore, the fuzzy recurrent neural network model of GWO can be used to handle the prediction and control problems of imprecise and uncertain rotary kiln parameters and has good robustness.

[0063] The basic idea of the fuzzy recurrent neural network model of GWO is to transform the complex global nonlinear problem of rotary kiln parameters into a simple local linear problem of rotary kiln parameters through IF - THEN rules, improving the accuracy, robustness, and reliability of rotary kiln parameter prediction and control. The fuzzy recurrent neural network model of GWO has a four - layer structure, namely the input layer, membership layer, rule layer, and output layer. The input - output relationships of each layer of the network are as follows:

[0064] (1) Input layer, the output of each node in the input layer is the input rotary kiln parameter value;

[0065] (2) Membership layer, each node in the membership layer represents a membership function, and the Gaussian function is selected as the membership function;

[0066] (3) Rule layer, each node in the rule layer is represented by Π, which means multiplying the input rotary kiln parameter signals;

[0067] (4) Output layer, this layer is represented by Σ, which means summing all the input rotary kiln parameter signals.

[0068] The fuzzy recurrent neural network model of GWO optimizes the parameters of the fuzzy recurrent neural network model by the GWO algorithm. The method for the GWO algorithm to optimize the parameters of the fuzzy recurrent neural network model refers to the design method of the BiGRU neural network model of GWO in the present invention.

[0069] II. Design of speed controller

[0070] The speed controller includes a differential unit 1, a NARX neural network model - PID controller 1 of GWO, a NARX neural network model 1 - 2 of GWO, a fuzzy recurrent neural network model - BiGRU neural network model 2 of GWO, and a BiGRU neural network model 2 of GWO. The NARX neural network model - PID controller 1 of GWO is the NARX neural network model - PID controller of GWO. The output of the NARX neural network model of GWO serves as the three parameters Kp, Ki, and Kd of the PID controller. The output of the NARX neural network model of GWO is used as the input of the PID controller, thereby realizing the online adjustment of the parameters of the PID controller and improving the robustness, accuracy, and reliability of the NARX neural network model - PID controller of GWO.

[0071] The incremental output of the NARX neural network model - PID controller of GWO is:

[0072] Δu(k) = k p [e(k) - e(k - 1)] + k i e(k) + k d [e(k) - 2e(k - 1) + e(k - 2)] (5)

[0073] The NARX neural network model - PID controller of GWO uses the three parameters output by the NARX neural network model of GWO to tune the parameters of the PID controller. The basic idea of the NARX neural network model of GWO to tune the PID controller is to adopt an adaptive adjustment method to adjust the three parameters of the PID controller in real - time and online, so as to optimize the control of the characteristics of the rotary kiln speed system to the greatest extent. The three - parameter vector formula of the PID controller of its algorithm is:

[0074]

[0075] The NARX neural network model of GWO can better describe the characteristics of the time series of the rotary kiln parameter changes. Therefore, the NARX neural network model - PID controller of GWO uses the three parameters output by the NARX neural network model of GWO to tune the parameters of the PID controller, improving the anti - interference ability, robustness, and stability of the rotary kiln speed of the NARX neural network model - PID controller of GWO.

[0076] The structure of the fuzzy recursive neural network model of GWO - the BiGRU neural network model 2 of GWO is the same as that of the fuzzy recursive neural network model of GWO - the BiGRU neural network model 1 of GWO, and it is designed with reference to the above - mentioned design method. The design methods of other models in the rotational speed controller refer to the relevant design steps of the parameter detection module of the present invention. The structure diagram of the rotational speed controller is shown in Figure 2 shown.

[0077] III. Design of Rotary Kiln Rotational Speed Control System

[0078] The rotary kiln rotational speed control system includes a rotational speed controller, parameter detection modules 1 - 3, the fuzzy recursive neural network model of GWO - PID controller 1 - 4, the NARX neural network model of GWO - PID controller 2 - 3, the fuzzy recursive neural network model of GWO - the BiGRU neural network model 3 - 4, differential unit 2 - 3, and the NARX neural network model of GWO 3 - 4.

[0079] The structures of the fuzzy recursive neural network model of GWO - PID controller 1 - 4 are the same. The three parameters output by the fuzzy recursive neural network model of GWO serve as the three parameters Kp, Ki, and Kd of the PID controller. The output of the fuzzy recursive neural network model of GWO serves as the input of the PID controller, so as to realize the online adjustment of the parameters of the PID controller, improve the robustness, accuracy, and reliability of the fuzzy recursive neural network model of GWO - PID controller. The design method of the fuzzy recursive neural network model of GWO - PID controller refers to the design method of the NARX neural network model of GWO - PID controller of the present invention. The NARX neural network model of GWO - PID controller 2 - 3 has the same model structure as the above - mentioned NARX neural network model of GWO - PID controller 1, and it is designed with reference to the design of the NARX neural network model of GWO - PID controller 1. The fuzzy recursive neural network model of GWO - the BiGRU neural network model 3 - 4 refers to the design of the above - mentioned fuzzy recursive neural network model of GWO - the BiGRU neural network model 1 - 2. The NARX neural network model of GWO 3 - 4 refers to the design of the above - mentioned NARX neural network model of GWO 1 - 2, that is, the design of the NARX neural network model of GWO. The design methods of other models refer to the design steps and methods of the relevant models of this patent. The structure diagram of the rotary kiln rotational speed control system is shown in Figure 3 shown.

[0080] IV. Design of Rotary Kiln Rotational Speed Detection and Intelligent Control Device

[0081] The rotary kiln speed detection and intelligent control device includes an STM32 microprocessor, an acceleration sensor, a current sensor, a speed sensor, a speed regulation device, and a field control terminal. Each sensor is connected to the STM32 microprocessor through a corresponding conditioning circuit. Communication between the STM32 microprocessor and the field control terminal is achieved through a USB interface. The software of the STM32 microprocessor mainly realizes data acquisition and preprocessing. The field control terminal is an industrial control computer, and there is a rotary kiln speed control system in the field control terminal to monitor the speed of the rotary kiln. The software is designed using the C language, with high compatibility, greatly improving the work efficiency of software design and development, and enhancing the reliability, readability, and portability of the program code.

[0082] The field monitoring terminal is an industrial control computer. The field monitoring terminal mainly realizes the collection of rotary kiln parameters and the monitoring of the speed. The main functions of the field monitoring terminal are the setting of rotary kiln communication parameters, the analysis and data management of rotary kiln parameters, and the rotary kiln speed control system, to achieve the monitoring of rotary kiln parameters and speed control, the query of rotary kiln historical data, the analysis of rotary kiln data, the alarm of rotary kiln data, personnel management, and the operation log of the rotary kiln, etc. The structure diagram of the rotary kiln speed detection and intelligent control device is shown in Figure 4 as follows.

[0083] The above embodiments are only used to illustrate the technical concept and characteristics of the present invention, and the purpose is to enable those who are familiar with this technology to understand the content of the present invention and implement it accordingly, and it cannot be used to limit the protection scope of the present invention. Any equivalent transformation or modification made according to the spirit and essence of the present invention should be covered within the protection scope of the present invention.

Claims

1. A rotary kiln speed detection and intelligent control method, characterized in that: The steps include: Step 1: A speed regulating device 1 and a speed regulating device 2 are respectively arranged at the head and the tail of the rotary kiln for synchronously regulating the rotation of the rotary kiln, and an acceleration sensor 1 and an acceleration sensor 2 are correspondingly arranged for detecting the acceleration values ​​of the head and the tail of the rotary kiln; the speed regulating device 1 and the speed regulating device 2 are connected with a current sensor 1 and a current sensor 2 for detecting the current values ​​of the speed regulating device 1 and the speed regulating device 2 respectively, and the rotary kiln is also provided with a speed sensor for detecting the speed of the rotary kiln; Step 2: The values ​​detected by the speed sensor, the current sensor 2, and the current sensor 1 are respectively outputted through the parameter detection modules 1-3 to obtain the detection feedback values ​​of the speed sensor, the current sensor 2, and the current sensor 1; Step 3: construct a rotary kiln speed control system, including a speed controller, a differential unit 2, a differential unit 3, a parameter detection module 1-3, a NARX neural network model of GWO 3-4, a fuzzy recursive neural network model of GWO-PID controller 1-4, a NARX neural network model of GWO-PID controller 2-3 and a fuzzy recursive neural network model of GWO-BiGRU neural network model of GWO 3-4; The speed target value and the speed detection feedback value output by the parameter detection module 1 are used as the input of the speed controller, the error and the error change rate between the speed controller output and the acceleration sensor 1 output are used as the input of the fuzzy recursive neural network model-PID controller 2 of GWO, the current change rate target value is 0, and the difference and the change rate of the difference between the current change rate target value and the output of the differential unit 3 are used as the input of the fuzzy recursive neural network model-PID controller 1 of GWO, the outputs of the fuzzy recursive neural network model-PID controller 1 of GWO, the fuzzy recursive neural network model-PID controller 2 of GWO, the NARX neural network model-PID controller 2 of GWO and the NARX neural network model-PID controller 3 of GWO are used as the input of the fuzzy recursive neural network model-GWO BiGRU neural network model 3 of GWO, and the output of the fuzzy recursive neural network model-GWO BiGRU neural network model 3 of GWO is used as the input of the speed regulating device 1; The error and the error change rate between the output of the speed controller and the output of the acceleration sensor 2 are used as the input of the fuzzy recursive neural network model-PID controller 4 of GWO, the current change rate target value is 0 and the difference between it and the output of the differential unit 2 and the change rate of the difference are used as the input of the fuzzy recursive neural network model-PID controller 3 of GWO, the outputs of the fuzzy recursive neural network model-PID controller 3 of GWO, the fuzzy recursive neural network model-PID controller 4 of GWO, the NARX neural network model-PID controller 2 of GWO and the NARX neural network model-PID controller 3 of GWO are used as the input of the fuzzy recursive neural network model-GWO BiGRU neural network model 4 of GWO, and the output of the fuzzy recursive neural network model-GWO BiGRU neural network model 4 of GWO is used as the input of the speed regulating device 2; The difference between the output of acceleration sensor 1 and the output of acceleration sensor 2 and the rate of change of the difference are used as the input of GWO's NARX neural network model-PID controller 2, the difference between the output of current sensor 1 and the output of current sensor 2 and the rate of change of the difference are used as the input of GWO's NARX neural network model-PID controller 3, the outputs of acceleration sensor 1 and acceleration sensor 2 are used as the inputs of GWO's NARX neural network model 3 and GWO's NARX neural network model 4 respectively, the speed sensor output is used as the input of GWO's NARX neural network model 3, GWO's NARX neural network model 4 and parameter detection module 1, the outputs of GWO's NARX neural network model 3 and GWO's NARX neural network model 4 are used as the input of parameter detection module 1, the outputs of current sensor 1 and current sensor 2 are used as the inputs of parameter detection module 3 and parameter detection module 2 respectively, and the outputs of parameter detection module 3 and parameter detection module 2 are used as the inputs of differential unit 3 and differential unit 2 respectively.

2. A rotary kiln speed detection and intelligent control method according to claim 1, characterized in that: The parameter detection modules 1-3 have the same structure, and all include GWO's NARX neural network model-GWO's BiGRU neural network model, GWO's BiGRU neural network model 1 and GWO's fuzzy recursive neural network model-GWO's BiGRU neural network model 1, the speed sensor or current sensor 2 or current sensor 1 output is respectively used as the input of GWO's NARX neural network model-GWO's BiGRU neural network model, GWO's BiGRU neural network model 1 and GWO's fuzzy recursive neural network model-GWO's BiGRU neural network model 1, the output of GWO's NARX neural network model-GWO's BiGRU neural network model is used as the input of GWO's fuzzy recursive neural network model-GWO's BiGRU neural network model 1, the output of GWO's BiGRU neural network model 1 is used as the input of GWO's fuzzy recursive neural network model-GWO's BiGRU neural network model 1, and the output of GWO's fuzzy recursive neural network model-GWO's BiGRU neural network model 1 is used as the detection feedback value output by the speed sensor or current sensor 2 or current sensor 1.

3. A rotary kiln speed detection and intelligent control method according to claim 1, characterized in that: The speed controller includes a differential unit 1, a NARX neural network model of GWO-PID controller 1, a NARX neural network model of GWO 1-2, a fuzzy recursive neural network model of GWO-BiGRU neural network model 2 of GWO and a BiGRU neural network model 2 of GWO; The speed target value is used as the input of the differential unit 1, the difference between the speed target value and the speed detection feedback value and the rate of change of the difference are used as the input of the NARX neural network model-PID controller 1 of GWO, the output of the differential unit 1, the NARX neural network model-PID controller 1 of GWO and the fuzzy recurrent neural network model-BiGRU neural network model 2 of GWO are used as the input of the NARX neural network model 1 of GWO, the sum of the output of the NARX neural network model 1 of GWO and the output of the NARX neural network model-PID controller 1 of GWO is used as the input of the BiGRU neural network model 2 of GWO, the speed detection feedback value is used as the input of the NARX neural network model 2 of GWO, the difference and the rate of change of the difference between the speed detection feedback value and the output of the NARX neural network model 2 of GWO and the difference and the rate of change of the difference between the speed target value and the speed detection feedback value are used as the input of the fuzzy recurrent neural network model-BiGRU neural network model 2 of GWO, and the sum of the output of the differential unit 1, the output of the BiGRU neural network model 2 of GWO and the output of the fuzzy recurrent neural network model-BiGRU neural network model 2 of GWO is used as the output of the speed controller.

4. A rotary kiln speed detection and intelligent control method according to claim 1, characterized in that: The GWO fuzzy recursive neural network model-PID controller 1-4 has the same structure, that is, the GWO fuzzy recursive neural network model is connected in series with the PID controller, and the three parameters output by the GWO fuzzy recursive neural network model are used as the input of the PID controller to realize online adjustment of the parameters of the PID controller; the GWO fuzzy recursive neural network model is a GWO algorithm to optimize the fuzzy recursive neural network model parameters.

5. A rotary kiln speed detection and intelligent control method according to claim 2 or 3, characterized in that: The structures of the fuzzy recursive neural network model of GWO-BiGRU neural network model 1 of GWO, the fuzzy recursive neural network model of GWO-BiGRU neural network model 2 of GWO, and the fuzzy recursive neural network model of GWO-BiGRU neural network model 3-4 of GWO are the same, that is, the fuzzy recursive neural network model of GWO and the BiGRU neural network model of GWO are connected in series, and the output of the fuzzy recursive neural network model of GWO is used as the input of the BiGRU neural network model of GWO. The fuzzy recursive neural network model of GWO and the BiGRU neural network model of GWO are the parameters of the fuzzy recursive neural network model and the BiGRU neural network model optimized by the GWO algorithm respectively.

6. A rotary kiln speed detection and intelligent control method according to claim 3, characterized in that: The structures of the GWO NARX neural network model-PID controllers 1-3 are the same, that is, the GWO NARX neural network model is connected in series with the PID controller, the GWO NARX neural network model output is used as the input of the PID controller, and the GWO NARX neural network model is a GWO algorithm to optimize the NARX neural network model parameters.

7. A rotary kiln speed detection and intelligent control method according to claim 3, characterized in that: The differential unit 1 takes the speed target values ​​at multiple sampling times as input, and divides the speed target difference between adjacent sampling times by the sampling time to determine the speed target change rate.

8. A rotary kiln speed detection and intelligent control method according to claim 1, characterized in that: After the differential unit 2 obtains the current feedback value of the current sensor 2 output by the parameter detection module 2, the difference between the current feedback value and the feedback value at the previous sampling moment is divided by the sampling time to determine the current change rate of the current sensor 2; after the differential unit 3 obtains the current feedback value of the current sensor 1 output by the parameter detection module 3, the difference between the current feedback value and the feedback value at the previous sampling moment is divided by the sampling time to determine the current change rate of the current sensor 1.

9. A rotary kiln speed detection and intelligent control device, characterized in that: The invention comprises an STM32 microprocessor, an acceleration sensor, a current sensor, a rotation speed sensor, a rotation speed regulating device and a field control end, wherein the acceleration sensor, the current sensor and the rotation speed sensor are respectively connected to the STM32 microprocessor through corresponding conditioning circuits; the communication between the STM32 microprocessor and the field control end is realized through a USB interface, the STM32 microprocessor realizes data acquisition and preprocessing, the field control end is an industrial control computer, and a rotary kiln rotation speed control system is arranged in the field control end to realize the rotary kiln rotation speed detection and intelligent control method described in any one of claims 1 to 8, and monitor the rotation speed of the rotary kiln.

Images