Mineral separation ball milling parameter adjustment control system and method based on artificial intelligence

Abstract

The invention discloses an artificial intelligence-based beneficiation ball milling parameter adjustment control system and method, the system comprises a field control layer and a local control management layer, the field control layer comprises concentration, liquid level, pressure and opening detection equipment controlled by a PLC controller, and the local control management layer comprises a server, a backup server and a computer operation terminal. The computer operation terminal comprises a data display module, an artificial intelligence parameter adjustment module and a historical data query module; signals of various equipment modification values are sent to a PLC, the PLC sends a control signal to a production execution unit, mineral separation ball-milling operation parameter data is selected as an input quantity of a model, and the quality of a mineral ball-milling process product is taken as an optimization target output quantity. According to the invention, automatic adjustment of mineral separation ball-milling parameters can be realized, the workload of field personnel is simplified, the stability of production control is improved, and a better control effect on the mineral separation ball-milling production process is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of mineral processing automatic control, and in particular to an artificial intelligence-based mineral processing ball mill parameter adjustment control system and method. Background Art

[0002] At present, the ball milling process of most domestic mineral processing plants still requires a lot of manual intervention, and the adjustment process is a trial-and-error method that relies entirely on manual experience. This method has the disadvantages of large parameter deviation, long adjustment time, and strong dependence on the actual experience of operators. Although a few domestic mineral processing plants have achieved automatic control of the mineral processing ball milling process, the grinding effect is not ideal due to the reliance on a single parameter to judge the state of the mill. The system generally has defects such as large control errors, serious parameter fluctuations, and poor product stability. Therefore, frequent manual intervention is required, and the dependence on manual labor has not been fundamentally changed. In summary, in order to eliminate the disadvantages brought about by frequent manual intervention, improve product quality, and achieve the goal of energy saving and consumption reduction, it is very important to develop and research advanced intelligent mineral processing ball mill parameter adjustment and control systems. Summary of the invention

[0003] The purpose of the present invention is to provide an artificial intelligence-based mineral processing ball mill parameter adjustment control system and method to solve the problems raised by the above-mentioned background technology.

[0004] To achieve the above object, the present invention provides the following technical solutions:

[0005] An artificial intelligence-based mineral processing ball mill parameter adjustment and control system, comprising a field control layer and a local control management layer, wherein the field control layer comprises concentration, liquid level, pressure, and opening detection equipment controlled by a PLC controller, and the local control management layer comprises a server, a backup server, and a computer operation terminal. The field control layer transmits each independent detection data to a database of the server through the PLC controller, and the server is connected to the computer operation terminal, and the computer operation terminal comprises a data display module, an artificial intelligence parameter adjustment module, and a historical data query module;

[0006] The data display module is used to display the real-time data of the mineral processing ball mill parameters from the PLC controller; the historical data query module is used to query the historical data of the mineral processing ball mill parameters, and generate a model training set after data cleaning; the artificial intelligence parameter adjustment module uses the historical numerical training set of the mineral processing ball mill parameters as input, trains the mineral processing ball mill artificial intelligence parameter adjustment model, and corrects it online in real time to modify the upper and lower limit values ​​of the production rules.

[0007] Furthermore, the historical data is the historical production data collected from the mineral processing ball mill site, and the real-time data is the real-time operation data collected from the mineral processing ball mill site, including the cyclone feed pressure value, the ball mill feed amount value, the ball mill feed speed value, the feed concentration value, the cyclone feed pump box liquid level value, and the cyclone feed valve opening value.

[0008] The present invention provides another technical solution: an artificial intelligence-based ore dressing ball mill parameter adjustment control method, comprising the following steps:

[0009] S1: Sampling the historical data of the adjustment of the production process parameters of the ore dressing ball mill, cleaning the data and generating the model training set, and using this data as the initial input data of the model at the current moment;

[0010] S2: Use fuzzy neural network to train the real-time sampled data and establish a prediction model for the parameter adjustment and control of the ore dressing ball mill;

[0011] S3: Optimize the prediction model to achieve optimal control of the parameters of the ore dressing ball mill production process.

[0012] Furthermore, the prediction model in S2 is as follows:

[0013]

[0014] Among them, u(k)=(u1(k),u2(k),…,u m (k)) represents the control quantity of the ball mill production optimization system, y(k) = (y1(k), y2(k),…,y n (k)) represents the target output of the mineral processing ball mill production state, d represents the output delay of the mineral processing ball mill production system, and p and q represent the number of input and output of the mineral processing ball mill production process system.

[0015] Furthermore, the fuzzy neural network structure in S2 has four layers:

[0016] Each neuron in the input layer represents an input variable of the prediction model, denoted by X1, X2, …, X r The operation control variables u(k) of the ball mill and the associated first p-order output y(k) are represented in turn;

[0017] Each input variable X in the membership function layer i There are u membership functions μ ij (j=1,2…,u), all are Gaussian membership functions:

[0018]

[0019] where μ ij is x iThe j-th membership function of ij and σ ij x i The center and width of the j-th Gaussian function, u is the number of membership functions;

[0020] The jth rule R in the fuzzy rule layer j The output is:

[0021]

[0022] Each node in the output layer represents an output variable that is a weighted sum of the input signals:

[0023]

[0024] Where y is the optimized target output value that characterizes the ball milling production status, w j The result parameter.

[0025] Furthermore, the method for optimizing control in S3 is:

[0026] S301: Obtain the output value of the ore dressing ball mill production system through the prediction model for optimization:

[0027] At the current sampling time k, the output estimate of the ore dressing ball mill production system is obtained by the established prediction model through the past input and current input u(k) of the ore dressing ball mill production system

[0028] The output estimate of the ore dressing ball mill production system can be obtained by the input u(k+1) of the ore dressing ball mill production system to be optimized and the past input / output

[0029] Assume that the prediction deviation at time k is Correction with deviation Get the correction amount

[0030] S302: Optimize the input values ​​of the ore dressing ball mill production system:

[0031] According to the optimal control quantity u of the ore dressing ball mill production, the objective function of the optimal control quantity u of the ore dressing ball mill production is determined as follows:

[0032]

[0033] Where minJ is the quality of the product of the ball milling process as the optimization objective function, and the product quality data includes ball milling output, ball milling grade, and system stability as the evaluation criteria, y ir is the reference trajectory of the characteristic state output of the i-th ore dressing ball mill production, obtained by solving the objective function of the ore dressing ball mill production optimization, y ipis the predicted output of the corresponding i-th beneficiation ball mill production characteristic state output after feedback correction, m and n are the dimensions of input and output respectively, q i and λ j is the weighting coefficient;

[0034] The minimum value of the above objective function is obtained through online real-time rolling optimization using the particle swarm algorithm, and the optimal control quantity u(k+1) is obtained. The optimal control quantity u(k+1) is applied to the ore dressing ball mill production system for optimal control.

[0035] Compared with the prior art, the present invention has the following beneficial effects:

[0036] The artificial intelligence-based mineral processing ball mill parameter adjustment control system and method of the present invention uses a mature fuzzy neural network model for data training based on historical data of the parameter adjustment process, and adjusts the structure and number of parameters of the algorithm model in real time to ensure the optimal quality of the mineral processing ball mill product in the current process. The present invention aims at the large delay characteristics of the mineral processing ball mill production process, can realize automatic adjustment of mineral processing ball mill parameters, simplify the workload of on-site personnel, improve the stability of production control, and has a good control effect on the mineral processing ball mill production process. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 It is a schematic diagram of the system framework of the present invention;

[0038] Figure 2 It is a block diagram of the internal modules of the computer operation terminal of the present invention;

[0039] Figure 3 It is the flow chart of the fuzzy neural network algorithm of the present invention. DETAILED DESCRIPTION

[0040] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0041] See also Figure 1-2An artificial intelligence-based mineral processing ball mill parameter adjustment and control system provided in an embodiment of the present invention comprises a field control layer and a local control management layer, wherein the field control layer comprises concentration, liquid level, pressure and opening detection equipment controlled by a PLC controller, and the local control management layer comprises a server, a backup server and a computer operation terminal, and the field control layer transmits each independent detection data to a database of the server through a PLC controller, and the server is connected to the computer operation terminal, and the computer operation terminal comprises a data display module, an artificial intelligence parameter adjustment module and a historical data query module; wherein the data display module is used to display the real-time data of the mineral processing ball mill parameters from the PLC controller; the historical data query module is used to query the historical data of the mineral processing ball mill parameters, and generate a model training set after data cleaning; the artificial intelligence parameter adjustment module uses the historical numerical training set of the mineral processing ball mill parameters as input, trains an artificial intelligence parameter adjustment model for the mineral processing ball mill, and corrects it online in real time, so as to modify the upper and lower limit values ​​of the production rules. The system sends signals of various equipment modification values ​​to the PLC, and the PLC sends control signals to the production execution unit. The ore dressing ball mill operating parameter data is selected as the input of the model, and the quality of the ore ball mill process product is used as the optimization target output. The product quality data includes ball mill output, ball mill grade, system stability, etc. as evaluation criteria.

[0042] Based on the above system, an embodiment of the present invention further provides a method for adjusting and controlling parameters of a mineral processing ball mill based on artificial intelligence, comprising the following steps:

[0043] S1: Sampling the historical data of the production process parameter adjustment of the ore dressing ball mill, cleaning the data and generating the model training set, and using this data as the initial input data of the model at the current moment; among them, historical data: historical production data collected from the ore dressing ball mill site, including cyclone feed pressure value, ball mill feed amount value, ball mill feed speed value, feed concentration value, cyclone feed pump box liquid level value, cyclone feed valve opening value, etc. Real-time operation data: real-time operation data collected from the ore dressing ball mill site, including cyclone feed pressure value, ball mill feed amount value, ball mill feed speed value, feed concentration value, cyclone feed pump box liquid level value, cyclone feed valve opening value, etc. After cleaning, the data is directly imported into the model. Adjustment experience: It is reflected in the adjustment process of the three parameter values ​​of the upper and lower limits of concentration, the upper and lower limits of pressure, and the upper and lower limits of liquid level, with the purpose of enabling the system to quickly reach a stable operating state.

[0044] S2: Algorithm model: fuzzy neural network, use fuzzy neural network to train real-time sampled data and establish a prediction model for ball mill parameter adjustment control; specifically, the prediction model is as follows:

[0045]

[0046] Among them, u(k)=(u1(k),u2(k),…,u m (k)) represents the control quantity of the ball mill production optimization system, y(k) = (y1(k), y2(k),…,y n (k)) represents the target output of the mineral processing ball mill production state, d represents the output delay of the mineral processing ball mill production system, and p and q represent the number of input and output of the mineral processing ball mill production process system.

[0047] The fuzzy neural network structure has four layers:

[0048] Each neuron in the input layer represents an input variable of the prediction model, denoted by X1, X2, …, X r The operation control variables u(k) of the ball mill and the associated first p-order output y(k) are represented in turn;

[0049] Each input variable X in the membership function layer i There are u membership functions μ ij (j=1,2…,u), all are Gaussian membership functions:

[0050]

[0051] where μ ij is x i The j-th membership function of ij and σ ij x i The center and width of the j-th Gaussian function, u is the number of membership functions;

[0052] The jth rule R in the fuzzy rule layer j The output is:

[0053]

[0054] Each node in the output layer represents an output variable that is a weighted sum of the input signals:

[0055]

[0056] Where y is the optimized target output value that characterizes the ball milling production status, w j The result parameter

[0057] S3: Optimize the prediction model to achieve optimal control of the parameters of the ore dressing ball mill production process. The specific optimization control method is as follows:

[0058] S301: Obtain the output value of the ore dressing ball mill production system through the prediction model for optimization:

[0059] At the current sampling time k, the output estimate of the ore dressing ball mill production system is obtained by the established prediction model through the past input and current input u(k) of the ore dressing ball mill production system

[0060] The output estimate of the ore dressing ball mill production system can be obtained by the input u(k+1) of the ore dressing ball mill production system to be optimized and the past input / output

[0061] Assume that the prediction deviation at time k is Correction with deviation Get the correction amount

[0062] S302: Optimize the input values ​​of the ore dressing ball mill production system:

[0063] According to the optimal control quantity u of the ore dressing ball mill production, the objective function of the optimal control quantity u of the ore dressing ball mill production is determined as follows:

[0064]

[0065] Where minJ is the quality of the product of the ball milling process as the optimization objective function, and the product quality data includes ball milling output, ball milling grade, and system stability as the evaluation criteria, y ir is the reference trajectory of the characteristic state output of the i-th ore dressing ball mill production, obtained by solving the objective function of the ore dressing ball mill production optimization, y ip is the predicted output of the corresponding i-th beneficiation ball mill production characteristic state output after feedback correction, m and n are the dimensions of input and output respectively, q i and λ j is the weighting coefficient;

[0066] The minimum value of the above objective function is obtained through online real-time rolling optimization using the particle swarm algorithm, and the optimal control quantity u(k+1) is obtained. The optimal control quantity u(k+1) is applied to the ore dressing ball mill production system for optimal control.

[0067] See also Figure 3 In order to further verify and explain the above control method, the following fuzzy neural network specific algorithm process is also provided:

[0068] First, the historical data of ball mill parameters are obtained and the historical parameters of ball mill process are sampled and the abnormal data are cleaned to form a training set. In order to improve the calculation rate of the model, the historical data of ball mill are first trained offline to obtain the prediction model of ball mill in the initial state.

[0069] Secondly, initialize the system state, prediction model and number of parameters. Take the generated model training data set as the input sample, and define the number of parameters and the rules for this production.

[0070] Secondly, online real-time training model: at sampling time k, based on the prediction model of the initial state of the beneficiation ball mill established offline, the fuzzy neural network is used to train the current sampling data, and the beneficiation ball mill optimization prediction model is corrected online in real time. The Mahalanobis distance value, error value, and overall error proportion value of the current target are solved.

[0071] Then, according to the size of the solution value, the determined beneficiation ball mill optimization control quantity u(k) and the system output y(k+d) are obtained through the fuzzy neural network prediction model. Assume the ball mill control quantity u(k+1) to be optimized as the position vector, bring it into the prediction model, and obtain the estimated output of the system at time k+1 Correct the estimated output through deviation and obtain the fitness function F, then select "Delete current rule", "Adjust parameters", and "Adjust upper and lower limits of rules".

[0072] Finally, combined with the operating status of the ball mill and the comprehensive quality of the products in the ore dressing ball milling process, a suitable algorithm model is determined, that is, the optimal control quantity u(k+1) is used in the ore dressing ball milling system to control the ore dressing ball milling process in real time.

[0073] In summary: the present invention provides a mineral processing ball mill parameter adjustment control system and method based on artificial intelligence, which uses a mature fuzzy neural network model for data training based on historical data of the parameter adjustment process, and adjusts the structure and number of parameters of the algorithm model in real time to ensure that the quality of the mineral processing ball mill product (slurry particle size) of the current process is optimal. The present invention aims at the large delay characteristics of the mineral processing ball mill production process, can realize the automatic adjustment of mineral processing ball mill parameters, simplify the workload of on-site personnel, and improve the stability of production control, and has a good control effect on the mineral processing ball mill production process.

[0074] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. An artificial intelligence-based mineral processing ball mill parameter adjustment and control system, characterized in that: It includes a field control layer and a local control management layer. The field control layer includes concentration, liquid level, pressure, and opening detection equipment controlled by a PLC controller. The local control management layer includes a server, a backup server, and a computer operation terminal. The field control layer transmits each independent detection data to a database of the server through the PLC controller. The server is connected to the computer operation terminal. The computer operation terminal includes a data display module, an artificial intelligence parameter adjustment module, and a historical data query module. The data display module is used to display the real-time data of the mineral processing ball mill parameters from the PLC controller; the historical data query module is used to query the historical data of the mineral processing ball mill parameters, and generate a model training set after data cleaning; the artificial intelligence parameter adjustment module uses the historical numerical training set of the mineral processing ball mill parameters as input, trains the mineral processing ball mill artificial intelligence parameter adjustment model, and corrects it online in real time to modify the upper and lower limit values ​​of the production rules.

2. The artificial intelligence-based mineral processing ball mill parameter adjustment and control system according to claim 1, characterized in that: The historical data are historical production data collected from the mineral processing ball mill site, and the real-time data are real-time operation data collected from the mineral processing ball mill site, including cyclone feed pressure value, ball mill feed amount value, ball mill feed speed value, feed concentration value, cyclone feed pump box liquid level value, and cyclone feed valve opening value.

3. A method for adjusting and controlling parameters of a mineral processing ball mill based on artificial intelligence, which is implemented based on the method for adjusting and controlling parameters of a mineral processing ball mill based on artificial intelligence according to claim 1, and is characterized in that: The following steps are involved: S1: Sampling the historical data of the adjustment of the production process parameters of the ore dressing ball mill, cleaning the data and generating the model training set, and using this data as the initial input data of the model at the current moment; S2: Use fuzzy neural network to train the real-time sampled data and establish a prediction model for the parameter adjustment and control of the ore dressing ball mill; S3: Optimize the prediction model to achieve optimal control of the parameters of the ore dressing ball mill production process.

4. The method for adjusting and controlling parameters of a mineral processing ball mill based on artificial intelligence according to claim 3, characterized in that: The prediction model in S2 is as follows: Among them, u(k)=(u1(k),u2(k),…,u m (k)) represents the control quantity of the ball mill production optimization system, y(k) = (y1(k), y2(k),…,y n (k)) represents the target output of the mineral processing ball mill production state, d represents the output delay of the mineral processing ball mill production system, and p and q represent the number of input and output of the mineral processing ball mill production process system.

5. The method for adjusting and controlling parameters of a mineral processing ball mill based on artificial intelligence according to claim 4, characterized in that: The fuzzy neural network structure in S2 has four layers: Each neuron in the input layer represents an input variable of the prediction model, denoted by X1, X2, …, X r The operation control variables u(k) of the ball mill and the associated first p-order output y(k) are represented in turn; Each input variable X in the membership function layer i There are u membership functions μ ij (j=1,2…,u), all are Gaussian membership functions: where μ ij is x i The j-th membership function of ij and σ ij x i The center and width of the j-th Gaussian function, u is the number of membership functions; The jth rule R in the fuzzy rule layer j The output is: Each node in the output layer represents an output variable that is a weighted sum of the input signals: Where y is the optimized target output value that characterizes the ball milling production status, w j The result parameter.

6. The method for adjusting and controlling parameters of a mineral processing ball mill based on artificial intelligence according to claim 5, characterized in that: The method for optimizing control in S3 is: S301: Obtain the output value of the ore dressing ball mill production system through the prediction model for optimization: At the current sampling time k, the output estimate of the ore dressing ball mill production system is obtained by the established prediction model through the past input and current input u(k) of the ore dressing ball mill production system The output estimate of the ore dressing ball mill production system can be obtained by the input u(k+1) of the ore dressing ball mill production system to be optimized and the past input / output Assume that the prediction deviation at time k is Correction with deviation Get the correction amount S302: Optimize the input values ​​of the ore dressing ball mill production system: According to the optimal control quantity u of the ore dressing ball mill production, the objective function of the optimal control quantity u of the ore dressing ball mill production is determined as follows: Where minJ is the quality of the product of the ball milling process as the optimization objective function, and the product quality data includes ball milling output, ball milling grade, and system stability as the evaluation criteria, y ir is the reference trajectory of the characteristic state output of the i-th ore dressing ball mill production, obtained by solving the objective function of the ore dressing ball mill production optimization, y ip is the predicted output of the corresponding i-th beneficiation ball mill production characteristic state output after feedback correction, m and n are the dimensions of input and output respectively, q i and λ j is the weighting coefficient; The minimum value of the above objective function is obtained through online real-time rolling optimization using the particle swarm algorithm, and the optimal control quantity u(k+1) is obtained. The optimal control quantity u(k+1) is applied to the ore dressing ball mill production system for optimal control.

Images