In the past ten years, the basic automatic control based on classical control theory has been successfully applied to the grinding process, mainly including the following circuits: constant feeding and proportional water supply based on single-variable and single-loop mills, stable grinding concentration; based on cascade Controlled cyclone feed concentration or fixed value control of spiral classifier overflow concentration, based on segmented control of cyclone grading control. In general, China has stayed at the stage of constant feed ratio proportional water supply control in terms of grinding process control. The universal application of new grinding processes such as ABC and AB makes the semi-autogenous grinding machine become the core of control. China has made certain progress in semi-autogenous grinding and ball mill load monitoring. It has been proved by industrial tests that this technology can help the mining machine working potential. However, there is still a long way to go from stable commercial control software products, and the grinding process optimization control software products have been commercialized abroad.
In recent years, a lot of research work has been done abroad on the modeling, soft measurement and optimization control of grinding process. Roux et al. developed a load forecasting (MPC) principle and approximate dynamic programming, and extended the control time domain to the prediction time domain. A model predictive static programming method with reference instruction tracking type was developed. MPSP) has been successfully applied to single-stage closed-circuit grinding circuits. Celep so on refractory gold ore silver fine grinding mill was stirred parameters were optimized using a three-Box-Behnken experimental design in response surface methodology (RSM) combined with quadratic programming, operational parameters of the ultra-fine grinding ( Modeling and optimization of steel ball diameter, grinding time, ball-to-batch ratio, and stirring speed). This method is used to model some parameters of a copper sulfide wet ball mill system. Kapakyulu et al. used energy balance to establish a total heat transfer coefficient model as a function of load, mill speed, and cylinder/pad design, modeling the energy loss of the mill, and quantifying the total heat transfer coefficient of the cylinder. The energy loss of the barrel heat transfer is explained. Mitra drive data modeling techniques applied to lead zinc ore concentrator industrial grinding operation to predict an output variable circuit, key performance indicators (KPI) comprises a certain three grain size fractions, the percent solids and cyclic loading, the use of Feedforward neural networks (FNN) and recurrent neural networks (RNN) use a hybrid model based on physical and empirical methods to approximate the actual behavior of the plant. Jayasundara combines the discrete element method (DEM) with the commonly used wear model to predict the wear pattern of the agitator disc in the Aisha mill, which is beneficial to the design and improvement of the Aisha mill. Gunda introduced a method to monitor the axial mixing of the Aisha mill, using a balance of mass and energy along the length of the mill to establish a predictive model of slurry mixing in the mill. Capece combines the particle fragmentation model with the interaction between particles to derive a mechanical efficiency factor for nonlinear PBM, which is expected to improve the design, control and optimization of dry mills. Powell is sensitive to mill filling for the operation of the AG/SAG mill and develops grinding curves associated with performance specifications such as throughput, power and product size to achieve the best grinding operation. Throughput, power, and product granularity peak at different fill levels, and the establishment of a grinding curve helps the operator determine the optimal operating setpoint. McElroy predicts ion-ion (pp) impact energy through physically measurable variables outside the drum. Based on the discrete element (DEM) model, a soft-measurement model was developed for horizontal rotating drums to quantitatively predict pp impact energy. OM Alruiz et al. predict the amount of ore processing through geo- and ore-based models and help develop ore mining plans and plant maintenance plans to maximize ore processing. The model relates the hardness of the ore, the particle size of the flotation to the ore and the treatment amount of the grinding by a power-based simulation method. The hardness of the ore is tested by Bond performance index and JK of different regions or ore bodies. Heavy experiment or SMC experiment was obtained. Akira Sato et al. studied the wear rate of grinding media by DEM simulation. It was found that the wear rate is closely related to the rotation speed of the mill, the diameter of the steel ball and the filling rate, and the impact energy and wear of the steel ball calculated by DEM. There is also a certain correspondence between the rates. MS Powell et al. used the EDEM software to fit and predict the wear of the liner and the shape of the lift strip, and studied the impact energy in the mill through model simulation. The results show that combined with DEM modeling and simulation. The method can provide powerful help for the design and improvement of the liner. RY Yang et al. established and validated the DEM model of an ISA mill. The flow pattern, velocity, force field and power were used to describe the flow of materials in the mill. It was found that the damping coefficient of the particles had negligible effect on the fluidity of the material. However, the sliding friction coefficient, the rotational speed and the filler have a significant effect on the fluidity. M. Sri Raj Rajeswari and others successfully established a dynamic model of fluidized bed gas-solid two-phase flow using three-dimensional CFD, visualized the solid distribution motion inside the fluidized bed, and analyzed the solid feed rate, gas pressure, The effect of operating parameters such as grading speed on fluidized bed operation. A. Ebadnejad et al. established a semi-self-grinding model by response surface method. The model includes three main variables: steel ball size, medium filling rate and mixed filling rate. The model determines the relationship between these variables and the ore particle size D80. relationship. A. Remes et al. established a grinding process simulation model through experimental data, and found that the most effective modeling method is to define the input and output changes of each variable separately. Finally, the fuzzy model predictive control is realized. Optimization of feed rate and particle size distribution. Jian Tang et al. studied the soft measurement of mill load by algorithms such as FFT, KPCA, GA and LS-SVM, and verified it in laboratory-scale mill experiments. Mohammad Kor et al. used fuzzy logic and particle swarm optimization to model and optimize the laboratory mill, and compared the fuzzy algorithm with the regression algorithm to obtain the conditions for minimizing the wear of the liner. Augustine B. Makokha et al. studied the influence of module concentration and ore supply on grinding time by establishing a serial agitator die with dead zone. The influence of grinding concentration and filling rate on grinding time was obtained through analysis. Give the mine a greater conclusion. R. Ahmadi et al. established a method for quickly determining the Bond Index by grinding dynamics equation and material balance formula, and verified the effectiveness of this method through experiments.
In view of the characteristics of non-linearity, large lag, time-varying and random disturbance of grinding classification control system, domestically developed advanced detection instruments in recent years have solved the problem of real-time detection of key equipment and process parameters. Typical detection instruments Including grinding machine load monitoring system based on vibration principle, grinding sound ear based on spectrum analysis, particle size analysis system, etc.; and applying modern intelligent control technology to grinding process, predictive function control, fuzzy control, expert system and neural network control There are cases applied to industrial sites. Compared with traditional algorithm control effects, intelligent control algorithms can improve the processing capacity of the mill, stabilize and improve the classification quality. However, at present, China has not yet developed a sophisticated intelligent control software for grinding processes that is recognized by domestic and foreign counterparts.
Grinding control strategy research includes traditional control strategy and modern control strategy. Traditional control strategy research includes PID control, Smith control and decoupling control. The early grinding classification mainly adopts PID single loop control method, such as feeding and grading overflow. Control loops such as grinding concentration, predictive compensators are designed for time lag problems, and decoupling compensators are designed for variable coupling problems. With the development of computer technology and process simulation technology, a series of new modern control strategies have emerged, including fuzzy PID control, expert systems, model predictive control, neural networks and hybrid control strategies.
Domestic scholars have many researches and applications on fuzzy control of grinding process, mainly because fuzzy control does not need to establish accurate mathematical models of control objects, and only needs to summarize the experience and data of field operators into more perfect language rules. Control of grinding systems with nonlinear, time-varying, and hysteretic characteristics. For example, Cheng Heng combines fuzzy control and PID control. The whole control system not only has the adaptability of fuzzy control to parameters, but also has fast adjustment speed, and has the characteristics of no static difference and good stability of PID. Zou Jinhui introduced the basic principle of fuzzy control in combination with engineering practice. By measuring the mill current closely related to the load, the two-dimensional fuzzy controller is used to fuzzy control the set value of the ore supply, and the control algorithm for the set value of the ore is proposed. The control rules can effectively solve the problems of swelling and underloading of the mill and improve the working efficiency of the mill. Wu Guangyao uses the current method to detect the load of the ball mill, and uses the fuzzy self-adjusting PID method to control the ore supply. Huang Wei et al designed a two-input and two-output two-dimensional fuzzy controller. The fuzzy control method was used to obtain the set value of the ore and the amount of ore discharged, and to control the overflow concentration and the overflow particle size. First, the current of the ball mill is detected by the current transmitter, and after the fuzzy operation, the optimal set value of the current ore amount is obtained. In order to compensate for the ore feeding error of the mining machine, a feedback link for the ore is added, and the actual ore amount is detected by the electronic belt scale, and the set value obtained by the fuzzy calculation is compared, and the error is PID adjusted. Therefore, a cascade control system with fuzzy control and PID adjustment is constructed. The main loop of the system adopts the fuzzy control algorithm to optimize the set value of the ore. The secondary loop uses the PID controller to achieve stable ore feeding. Ma Yingxi constructed a fuzzy rule base based on historical data samples. Under the guidance of expert experience, a reasonable membership function and parameters were selected to establish a fuzzy control model for the grinding classification system. To solve the priority problem of fuzzy rules, the design was based on genetics. The fuzzy rule weight adjuster of the algorithm uses the weight sequence as the chromosome, sets the genetic group, crosses the mutated multiple generations of the chromosome, and uses the genetic selection of the gambling algorithm to make the chromosome evolve to the highest fitness direction, and finally optimizes. Weight sequence, which optimizes the adjustment of fuzzy rule weights. Wang Zhanlou selected three main parameters: mill sound, mill power, and sanding amount of a classifier. As the input of the fuzzy controller, these parameters are changing all the time, and the changes reflect the current working condition of the mill. The fuzzy controller performs fuzzy determination according to the change or change trend of the main parameters. For each change trend, the fuzzy controller will give a specific feeding principle, and then the PID controller adjusts the inverter control according to the principle of feeding. Feed the mining machine to achieve the purpose of accurate ore. Liu Qi proposed a fuzzy internal model control (FIMC) method to adjust the filter parameters in internal model control online by fuzzy logic inference. Zhou Ping proposed a multivariable fuzzy supervisory control (MFSC) method consisting of fuzzy supervisor, loop pre-set model and granularity predictor for a typical wet grinding circuit in the beneficiation process, using fuzzy intelligence and other intelligent techniques. The production process is supervised and the set values ​​of the bottom circuit are adjusted. Liang Lei uses a hydrocyclone for the grinding and grading process, adopts a fuzzy control system for the pumping pool level, and increases the integral link to improve the standard fuzzy control algorithm, so that the deviation of the setting and feedback and the degree of weight change of the deviation are different. The static error of the system is eliminated, and the simplified object model with the first-order inertia link with hysteresis is simulated. In order to effectively avoid the occurrence of swelling and underloading of the mill, to stabilize and improve the various working indexes of the mill, many researches have been carried out on the fuzzy PID grinding machine automatic control technology in China, and at the same time, the technology is in multiple metal mines. The application is carried out in the grinding classification system.
Intelligent techniques such as rule reasoning, expert systems, and artificial neural networks have received much attention. Using the rule-based reasoning (RBR) and statistical process control (SPC) techniques, Zhou Ping proposed an intelligent monitoring and control method for the mill load consisting of the SPC mechanism, the overload monitoring module and the supervisory controller. The supervisory controller automatically modifies the control loop. The set value is used to track the modified set value through the output of the control loop, so that the mill load is gradually away from the overload state, and an intelligent optimization control system for the grinding process that realizes the granularity index is constructed. For a typical two-stage closed-circuit grinding whole hematite ore mine unstable nature of the existence of large fluctuations in size, grinding performance indicators can not be measured online and the difficulty of establishing mathematical model and other issues, the control method is proposed based on data and knowledge, including case-based Inferred control loop pre-setting, grinding particle size dynamic neural network soft measurement and multivariable fuzzy dynamic regulator. Wang Yunfeng used the traditional RBF neural network and advanced fuzzy control technology to form an adaptive fuzzy control scheme, and simulated the control of mill load and grinding concentration. Aiming at the slow time-varying and nonlinear characteristics of the grinding classification process control, Zhao Hongwei proposed an adaptive fuzzy inference network model based on system identification. The fuzzy clustering method was used to systematically identify existing data samples and obtain fuzzy automatically. Based on the obtained fuzzy system and the corresponding fuzzy parameters, the adaptive fuzzy neural network inference system based on the Takage-Sugeno inference model is constructed, which is more adaptive and faster than traditional fuzzy neural networks. Qi has established an adaptive neuro-fuzzy inference system (ANFIS), and established a control model using ANFIS for different control objects in the grinding process. Ding Liang proposed a method based on fuzzy neural network control to study the ore supply of the mill, the slurry concentration of the sand pump pool, and the inlet pressure of the cyclone. The overflow particle size, cyclone operating pressure, and ore are selected. Quantity, sand pump pool slurry concentration, mill power, grinding electromechanical ear as test data, select the amount of ore, mill water after adding water, cyclone inlet pressure as the controlled variable, to achieve automatic control of ball mill feeding, sand pump Pool concentration control, control of cyclone overflow particle size. Based on fuzzy logic, Gao Zhichao constructed a fuzzy neural network with BP network as the structural framework, and discussed the application and research of fuzzy neural network based on fuzzy logic in the grinding intelligent control system of a mining company's concentrator. Liu Weifeng selected six parameters of the ore, liquid level, cyclone pressure, grinding concentration, supplemental water and ball mill current as input variables, and adopted a three-layer RBF neural network to predict the grinding grain size. To achieve the purpose of control of the grinding process, and use the MATLAB development platform to establish a prediction model, using industrial data to simulate the prediction model. Based on case-based reasoning technology, Dai Wei proposed an intelligent multivariable control method for closed-circuit grinding circuits, which is used to adjust the set value of the process control system. Based on the deviation between the expected and actual grinding grain size, comprehensive consideration of the mill load state is considered. Drive motor power and grinding information to adjust the set value of the new ore supply, cyclone feed concentration and feed flow. Based on case-based reasoning, rule-based reasoning, neural network and other intelligent methods, Zhao Dayong studied the intelligent operation optimization control method of grinding process consisting of control loop pre-set model, feedforward and feedback compensator and grinding grain size prediction model.
Grinding process optimization control method based on soft measurement, multivariable decoupling, model prediction and other technologies. Li Yong proposed a gray soft measurement method for the problem that the main parameters of the mill could not be detected online. The application of support vector machine (SVM) method was proposed for the real-time and effective detection of grinding productivity. Based on the on-line prediction of grinding-productivity and based on the soft-measurement model of grinding productivity and the mechanism model of energy consumption, aiming at improving grinding productivity and reducing production energy consumption, a comprehensive optimization control strategy for grinding process quality was proposed. The set value of the water replenishment is optimized by the improved grinding concentration control model, and then the cascade control is used to achieve the optimal tracking control of the grinding concentration. Zhao Dayong proposed a gain adaptive internal model control method, combining internal model control and object gain parameter identification algorithm to correct the internal model controller and controlled object model by onlinely identifying the gain parameters of the controlled object model. To reduce the effects of model parameters and various disturbances on the system. Dong Fei introduced a multi-model control strategy for multi-variable dynamic matrix control in a grinding control system. Based on different ore hardness, different step models were established for the ball milling process to predict the control of the grinding process. Robustness. Yang Shuliang decomposes the parameters that need to be controlled in the grinding process into the ore supply of the mill, the water supply of the mill, the concentration of the slurry into the cyclone, the working pressure of the cyclone, etc., to the overflow of the slurry The particle size and concentration are the final control targets, and the parameters such as the ore supply amount are controlled. In order to control the overflow concentration and fineness in the quality index interval, Ma Tianyu proposed a multi-model predictive control scheme considering the local steady-state economic target, and established a ball mill and classifier transfer matrix model based on the field database. Economic performance, the steady-state economic goal is embedded into the dynamic optimization objective function in the form of penalty function. To eliminate the influence of model mismatch caused by ball mill change, a multi-model switching strategy based on the changing rule is established. For successive rubbing process bauxite will mill simplified to a 3 input, a continuous mixing process output state are established spatial concentration of the prediction model according to the volume and material balance principles established grain size mass balance model different sets of mine proposed A weighted multi-model fineness prediction model based on weight coefficient optimization is proposed. A predictive control strategy based on multi-objective optimization structure with excavation concentration interval control and economic optimization is proposed. Guarantee the optimal addition of blanking and steel balls. Chen Xisong proposed a nonlinear multi-model control algorithm. Firstly, several sub-models and corresponding controllers are established near several equilibrium points. Then, the model matching degree is calculated online to adapt to the change of model parameters, and the final input of the controlled object. It is the weighted sum of the outputs of each controller and is applied to the control of the grinding graded nonlinear system. Combining the advantages of disturbance observer and model predictive control, Wang Hongchao combines on-line estimation based on disturbance observer with model predictive control to form an anti-disturbance multivariable composite control structure DOB-MPC to improve the control performance of grinding classification process. . Daiwei uses the data of grinding process and adopts neural network to propose a data-driven grinding process optimization control method consisting of loop pre-set value optimization, performance index estimation, optimization set value evaluation and grinding particle size soft measurement.
Based on model dynamic optimization, particle swarm optimization, multi-agent and other technical grinding optimization control methods. Ma Tianyu proposed a dynamic optimization control scheme. Firstly, the optimal control law of the controlled variable is obtained by the optimization calculation model, in order to eliminate the influence of steel ball wear, measurement error, ore particle size change and water pressure instability on the system. The feedback mechanism of quality index feedback is introduced. The feedback information is obtained by manual measurement. Finally, the fuzzy expert system compensates the control law obtained by optimizing the calculation model according to the deviation between the manual measurement and the expected index. Zhang Hongyan regards the grinding circuit in the beneficiation process as a nonlinear stochastic distribution system, and uses the improved PDF tracking control algorithm to control the PSD of the grinding products. By adjusting the newly added mineral amount of the mill at each sampling time, the PSD of the hydrocyclone overflow ore is close to the optimal PSD index required by the subsequent sorting process, and the kernel density estimation method is used to estimate the granularity PDF, and then based on The product granularity PDF and the target PDF error construct the performance index function, and optimize the performance index function through the PSO method to obtain the optimal control input for each sampling moment. It is difficult to control the drift of the optimal load point of the ball mill. He Xiaoqiao and others use the extreme value dynamic optimization method to dynamically optimize the optimal working point of the mill load. Ren Jinxia applied the improved particle swarm optimization PID scheme to the overflow concentration control of the grinding grading system, optimized the proportional, integral and differential coefficients, and simulated it with the simplified object model of second-order inertia plus lag. Ma Tianyu proposed an improved model of ball mill continuous grinding grain mass balance model (PBM) based on the conclusion of batch test and considering the working conditions. Based on the batch test results of some grain grades that do not meet the first grinding dynamics, the fracture rate is improved. Model, in order to eliminate the influence of interference, establish the LSSVM relationship model between the fracture rate model parameters and the various working conditions variables. Based on the improved model, PBM+MPC is used to realize the ball mill optimization control. Zhou Ping proposed MPC advanced feedback control based on improved disturbance observer (DOB) to control multivariable grinding operation. In the advanced feedback control of DOB-MPC, high-level optimization calculates the optimal working point by maximizing the profit function, and then transmits For the MPC layer, MPC acts as a pre-set controller to provide appropriate pre-set values ​​for the underlying basic feedback control system, and DOB acts as a compensator to dynamically compensate for the underlying control system based on observed disturbances and factory uncertainties. set. In order to design the upper loop setting system, an improved two-degree-of-freedom decoupling control method and a model approximation method based on multi-point step response matching for complex high-order multi-input and output time-delay systems are proposed. In the research of the optimization control system of the grinding process, Qi studied the grinding control system for multi-task decomposition, and divided several control links in the whole process of grinding production into several relatively independent sub-control tasks, such as feeding. Control tasks, liquid level control tasks, pressure control tasks, as well as grinding fault diagnosis and control decision-making tasks, and then integrate each sub-control task according to certain cooperation rules, and establish a model of grinding intelligent control system based on multi-agent. Multiple individual agents then complete the overall control tasks through negotiation and cooperation. Tie Ming has developed a distributed semi-physical simulation platform consisting of optimized computer and monitoring computer, controller, virtual execution and inspection equipment, and virtual grinding process objects. The virtual object simulation software is based on grinding grade. Process dynamic model, which is used for dynamic optimization control simulation of grinding process. Lu Shaowen has developed a grinding circuit simulator NEUSimMill, which is mainly used for testing and calibration of grinding process control systems, including advanced control systems such as integrated control. The simulator implements a dynamic ball mill grinding model to study process variables and product particle size distribution. Response to interference and control behavior.
Finished Razor ,Face Razor,Leaf Razor,Gillette Razors
NINGBO KAYDO PLASTIC CO.,LTD , https://www.kaydo-china.com