Flotation column process intelligent control method and system based on tailing ash content online feedback

Abstract

The invention relates to a flotation column process intelligent control method and system based on tailing ash content online feedback, belongs to the technical field of flotation, and solves the problem of poor stability of a flotation column process in the prior art. The method comprises the specific steps that working condition data related to the tailing ash content in the flotation column production process are collected in real time; based on state variable data meeting working condition requirements in the collected working condition data, utilizing the trained prediction control model to obtain predicted control variable data; in response to the fact that the tailing ash value read in real time is qualified, an execution mechanism is used for executing regulation and control according to the predicted control variable data; in response to the fact that the tailing ash value is unqualified, the predicted control variable data and the tailing ash value are utilized to carry out control variable optimization based on a quadratic programming optimization algorithm, and an optimal regulation and control sequence is obtained; performing regulation and control according to the optimal regulation and control sequence by utilizing an execution mechanism; and in the production process, cyclic adjustment and optimization are carried out by utilizing a predictive control model and an optimization algorithm, so that the stability of a control strategy is realized.

Description

Technical Field

[0001] The present invention relates to the field of flotation technology, and in particular to an intelligent control method and system for a flotation column process based on online feedback of tailings ash content. Background Art

[0002] Flotation process is the main technology for sorting fine-grained minerals. It has been widely used in mineral processing plants. In coal sorting, it is mainly used to improve the quality and reduce the ash content of fine-grained coal slime. With the application of mechanized and intelligent mining technology, the proportion of fine-grained coal slime in the raw coal for washing is getting higher and higher, which leads to a large amount of coal slime in coal preparation plants, reduced floatability of coal slime and reduced economic benefits of coal preparation plants. According to the properties of the equipment, the flotation process can be divided into flotation machine and flotation column process. The flotation column process is more suitable for the sorting of fine-grained coal slime. However, the process has high requirements for feed concentration and feed particle size, and the system has a large processing capacity. The regulation of a single process parameter can often have a great impact on the sorting effect. Fine regulation is one of the conditions to ensure the sorting effect of the flotation column process. At the same time, the rapidly changing policies and technologies put forward higher requirements for the intelligent control of the flotation column process. At present, the domestic coal slime flotation column process generally lacks the technology of intelligent regulation. The main problems are: insufficient detection accuracy of process variables, poor stability of control strategies, and no participation of product ash content in system regulation.

[0003] With the increasing popularity of artificial intelligence technology in industrial applications, the intelligent control of flotation column processes has been supported by new technologies. Taking coal slime flotation as an example, the process variables mainly include flotation feed properties, flotation column equipment and product variables. These parameter variables show a strong nonlinear relationship. The current traditional control methods are mostly supplemented by manual experience. This method has low accuracy, and clean coal products and reagents are frequently lost during the production process. In addition, the online detection technology of tailings ash has matured. Adding flotation tailings ash as a basis for feedback control can effectively prevent the entrainment of flotation clean coal in coal slime.

[0004] In summary, in the current related technologies, the production method of the flotation column process is backward, and the existing manually-dominated control method has poor stability due to the influence of subjective factors, resulting in frequent occurrence of low flotation clean coal recovery rate and high reagent loss during the production process. Summary of the invention

[0005] In view of the above analysis, the embodiments of the present invention aim to provide a method and system for intelligent control of a flotation column process based on online feedback of tailings ash content, so as to solve the problem of poor stability of the flotation column process in the prior art.

[0006] The purpose of the present invention is mainly achieved through the following technical solutions:

[0007] On the one hand, an embodiment of the present invention provides a flotation column process intelligent control method based on online feedback of tailings ash content, comprising the following steps:

[0008] Real-time collection of operating data related to tailings ash content during flotation column production;

[0009] Based on the state variable data that meets the working condition requirements in the collected working condition data, the predicted control variable data is obtained by using the trained predictive control model;

[0010] In response to the tailings ash value read in real time being qualified, using an actuator to perform regulation according to the predicted control variable data;

[0011] In response to the tailings ash value being unqualified, using the predicted control variable data and the tailings ash value, based on a quadratic programming optimization algorithm, control variable tuning is performed to obtain an optimal control sequence; and using an actuator to perform control according to the optimal control sequence;

[0012] During the production process, the predictive control model and optimization algorithm are used to perform cycle tuning.

[0013] Further, the predictive control model is trained based on a multi-input multi-output long short-term memory model, including:

[0014] Collect historical control data in the flotation column production process to obtain a flotation column process production data set; wherein the state variable data in the production data set is used as the input data set, and the corresponding control variable data is used as the output data set;

[0015] Based on a multi-input and multi-output long short-term memory model, training is performed using the loss function calculated by the mean square error of the control variables to obtain a trained predictive control model.

[0016] Further, obtaining the optimal regulatory sequence includes:

[0017] Obtaining a cost function based on the tailings ash value and the control variable data;

[0018] Optimize with the goal of minimizing the cost function to obtain the optimal control variable output;

[0019] The input features are extracted based on the state variables inputted by the predictive control model, the contribution values ​​of the input features are calculated, and the corresponding optimal control variables are sorted according to the contribution values ​​to obtain the optimal regulation sequence.

[0020] Furthermore, the cost function is expressed as:

[0021]

[0022] Among them, Q and F are both adjustment matrices; Y is the sample matrix to be optimized, including the tailings ash value and the control variables; k is the time or serial number of the current optimized sample.

[0023] Further, according to the local interpretable model-independent interpretation technology, the contribution value is calculated based on the following formula:

[0024] F(x i ) = argminL(x i )+Ω(g),

[0025] Among them, x i is the input feature; L is the loss function; Ω is the complexity of the model.

[0026] Furthermore, the loss function is:

[0027]

[0028] L=Loss(t)+Loss(t+1),

[0029] Among them, L is the loss function; t is the current data moment; n is the total number of control variables; y is the true value of the control variable; y p is the predicted value of the control variable; w is the weight coefficient.

[0030] Furthermore, the state variables include feed flow, feed concentration, dilution water flow, circulation flow, circulation pressure, foam layer thickness and tailings ash content; the control variables include foaming agent addition amount, collector addition amount, feed valve opening, circulation valve opening, water addition valve opening and tailings gate motor.

[0031] Furthermore, when the collected operating data does not meet the operating requirements, the control variables output by the flotation column production process control feedback are sent to the corresponding actuator to complete the control until the operating requirements are met.

[0032] On the other hand, an embodiment of the present invention provides an intelligent control system for a flotation column process based on online feedback of tailings ash content, comprising:

[0033] Production parameter detection module, used to collect working condition data in the flotation column production process through detection devices;

[0034] The prediction control module is used for training the prediction control model; it is also used for obtaining predicted control variable data based on the real-time data of the state variables that meet the working condition requirements in the collected working condition data, using the trained prediction control model, and in response to the qualified tailings ash value read in real time, using the actuator to perform regulation according to the predicted control variable data; in response to the unqualified tailings ash value, using the predicted control variable data and the tailings ash value, based on the quadratic programming optimization algorithm to perform control variable tuning to obtain the optimal control sequence; using the actuator to perform regulation according to the optimal control sequence; in the production process, using the prediction control model and the optimization algorithm to perform cyclic tuning;

[0035] A control module, used for sending the optimal control sequence to a corresponding actuator in the form of a control instruction through a PLC controller;

[0036] The actuator is used to execute the control instructions to complete intelligent regulation.

[0037] Further, obtaining the optimal regulatory sequence includes:

[0038] calculating a cost function based on the tailings ash value and the predicted output of the control variable;

[0039] Optimize with the goal of minimizing the cost function to obtain the optimal control variable output;

[0040] The input features are extracted based on the state variables input by the predictive control model, the contribution values ​​of the input features are calculated, and the corresponding optimal control variables are sorted according to the contribution values ​​to obtain the optimal regulation sequence.

[0041] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects:

[0042] 1. The present invention proposes to select real-time tailings ash content as an online feedback condition, collect the production process variables of the flotation column process, establish an intelligent control system for the flotation column process based on prediction and optimization control, realize the intelligent start and stop and operation of the flotation column process, and achieve the stability of the control strategy;

[0043] 2. Based on the multi-input multi-output MIMO model, a predictive control technology analysis and an optimization control module for predicting the control sequence and action amplitude of the target variables were established, and an intelligent control system for the flotation column process was established, which improved the sorting efficiency and intelligence of the flotation column process system;

[0044] 3. The system is applied to the coal slime flotation process system. Through the online feedback of tailings ash content, it can effectively prevent the flotation clean coal from being carried in the coal slime, improve the clean coal recovery rate of the coal slime flotation process system, and increase the economic benefits for the target enterprise.

[0045] In the present invention, the above-mentioned technical solutions can also be combined with each other to achieve more preferred combination solutions. Other features and advantages of the present invention will be described in the subsequent description, and some advantages can become obvious from the description, or can be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained through the contents particularly pointed out in the description and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The drawings are only for the purpose of illustrating particular embodiments and are not to be considered limiting of the present invention. Like reference symbols denote like components throughout the drawings.

[0047] Figure 1 It is a flow chart of the intelligent control method of the flotation column process based on online feedback of tailings ash content according to an embodiment of the present invention;

[0048] Figure 2 It is a structural schematic diagram of an intelligent control system for a flotation column process with online feedback of tailings ash content according to an embodiment of the present invention;

[0049] Figure 3 It is a schematic diagram of the fluctuation of feed flow rate after exponential smoothing of feed flow rate data in an embodiment of the present invention;

[0050] Figure 4 A control flow chart of a flotation column process system based on MIMO model prediction according to an embodiment of the present invention;

[0051] Figure 5 It is a schematic diagram of the structure of an intelligent control system for a flotation column process based on online feedback of tailings ash content according to an embodiment of the present invention.

[0052] Reference numerals:

[0053] 1-collector metering pump, 2-frother metering pump, 3-slurry preprocessor, 4-feed flowmeter, 5-feed concentration meter, 6-dilution water valve, 7-dilution water flowmeter, 8-feed valve, 9-pressure gauge, 10-foam layer thickness meter, 11-bubble generator, 12-flotation concentrate discharge port, 13-tailings gate motor, 14-tailings ash meter, 15-tailings discharge port, 16-flotation column, 17-circulation pipeline, 18-circulation volume flowmeter, 19-circulation volume valve. DETAILED DESCRIPTION

[0054] The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not used to limit the scope of the present invention.

[0055] Example 1

[0056] A specific embodiment of the present invention discloses an intelligent control method for flotation column process based on online feedback of tailings ash content, such as Figure 1 As shown, the following steps are included:

[0057] Step S1, real-time collection of operating data related to tailings ash content during flotation column production;

[0058] Step S2, based on the state variable data that meets the working condition requirements in the collected working condition data, using the trained predictive control model to obtain predicted control variable data;

[0059] Step S3, in response to the tailings ash value read in real time being qualified, using the actuator to perform regulation according to the obtained predicted control variable data;

[0060] Step S4, in response to the tailings ash value being unqualified, using the predicted control variable data and the tailings ash value, based on a quadratic programming optimization algorithm, to perform control variable tuning to obtain an optimal control sequence; using an actuator to perform control according to the optimal control sequence;

[0061] Step S5: During the production process, the predictive control model and the optimization algorithm are used to perform cycle tuning.

[0062] Through the above method, on the basis of the existing coal slime flotation column process, by combining the control variables and tailings ash index in the flotation column process production process, and using the predictive control model and optimization algorithm, the flotation column intelligent control with online feedback of tailings ash is realized, the sorting efficiency of the flotation column process system is improved, and the stability of the control strategy is achieved.

[0063] Specifically, in step S1, the coal slime flotation column production process includes a flotation feed section, a flotation column equipment section, a flotation reagent section, and a flotation product section. The operating data in the flotation column production process include feed flow, feed concentration, dilution water flow, circulation pressure, foam layer thickness, collector flow, frother flow, tailings ash content, and circulation flow.

[0064] Specifically, in step S2, state variables closely related to the tailings ash index are extracted from the operating data, including feed flow rate, feed concentration, circulation volume (including circulation flow rate and circulation pressure), foam layer thickness and tailings ash content as variables to be monitored for intelligent control.

[0065] At the same time, the early warning boundary of the historical control data is determined based on expert knowledge, and the steady-state interval of the corresponding state variable is set. In the non-early warning range, the coal slime flotation column process system is intelligently controlled by monitoring the state variables, including using the feedback mechanism to control the state variables that do not meet the working conditions to the stable interval; for the state variables that meet the working conditions, the tailings ash value is combined for cyclic optimization. Among them, the specific steps for setting the steady-state interval of the state variable are as follows:

[0066] S211, collect historical control data of flotation column production process in sections, including detection variables and control variables;

[0067] S212. Abnormal data is eliminated by combining expert knowledge, historical control data is sorted out according to time series, and existing data is normalized based on the following formula:

[0068]

[0069] S213. For each single historical control data, perform exponential smoothing to obtain a smooth parameter change trajectory and remove errors caused by noise and instrument fluctuations. The exponential smoothing calculation formula for historical control data is as follows:

[0070] p n =α*x n +(1-α)*(α*x n-1 +(1-α)*p n-2 ),

[0071] Among them, p is the predicted value after exponential smoothing; α is the attenuation weight; x is the original historical control data; and n is the number of exponential times.

[0072] S214. Arrange the historical control data of the flotation column process production after pretreatment, and determine the warning boundary of the historical control detection variable data and the steady-state interval of the corresponding state variable based on expert knowledge.

[0073] Exemplarily, in combination with the equipment performance of the flotation column, the parameter fluctuation range of the flotation clean coal ash content is set to be less than 0.5%, as shown in Table 1, which is the set steady-state operating range (conventional range):

[0074] Table 1

[0075] variable Stable operating range Feed flow 250-350m3 / h Feed concentration 50-80g / L Foam layer thickness 35-60cm Circulation flow 250-350m3 / h Cyclic Pressure 0.10-0.20MPa

[0076] Within the above-mentioned interval, if the corresponding state variable is outside the stable operating range, the intelligent control feedback mechanism of the flotation column production process is utilized to send the control variable output by the control feedback to the corresponding actuator through a control module such as a PLC controller to control the above-mentioned flotation column state variable within the steady-state operating range. If it cannot be controlled within the steady-state operating range, a system alarm is triggered.

[0077] It should be noted that among the state variables, the feed concentration mainly comes from the regulation of the dilution water valve opening, the circulation flow is controlled by the circulation valve opening, and the foam layer thickness and slurry level are controlled by the tailings gate motor. Therefore, the control variables that affect the state variable values ​​are determined as the control objects in the control data, and the control variables include the collector metering pump, the foaming agent metering pump, the dilution water valve opening, the feed valve opening, the tailings gate motor and the circulation valve opening.

[0078] Specifically, for state variables within the stable operating range, based on the state variable data collected in real time, the trained predictive control model is used to obtain the corresponding predicted output of the control variable data;

[0079] Exemplarily, a predictive control model is trained based on a multi-input multi-output long short-term memory model (MIMO), including:

[0080] Step S221: The flotation column production process historical control data collected in segments are preprocessed to obtain a flotation column process production data set, and the flotation column process production data set is matched one-to-one with the input variables and the output variables, and the screened data set is randomly divided into training data and test data; wherein, the state variable data in the production data set is used as the input data set, and the corresponding control variable data is used as the output data set;

[0081] Exemplarily, all detection instruments and actuators are connected to the PLC controller, and all flotation column process production control data are saved on the local server with the help of SQL database. It took one month to collect 35,000 data from the flotation column control system of the coal preparation plant, and invited industry flotation experts and professionals to eliminate historical data, and then normalize the existing data; after data normalization, perform exponential smoothing on the historical data of each detection variable to eliminate the error caused by the large fluctuation amplitude of the dependent variable. Select the smoothing index as 2, set the attenuation weight α to 0.9, and the smoothing calculation process formula is:

[0082] p2=α*x2+(1-α)*(α*x1+(1-α)*p0),

[0083] Among them, Figure 3 As shown, the ordinate is the flow value, the unit is Kg / m3; at the initial moment, p0=x0, the effect diagram of exponential smoothing of feed flow data.

[0084] Among the data collected above, 24826 valid data were extracted and randomly divided into training data and test data in a ratio of 8:2; among them, 19860 training data were used as the training set and 4966 test data were used as the test set.

[0085] Step S222: Based on the output variables of the MIMO training flotation column production process, i.e., the control variables, after the flotation column process production data set is input into the model, the calculation formula of the first input gate is as follows:

[0086] C = tanh(W c k (t-1) +U c x t +b),

[0087] Among them, W and U are linear relationship coefficients respectively; b is bias; tanh is activation function; k is hidden state variable; t is time series; x is flotation production feature vector.

[0088] Step S223: After obtaining the output of the input gate, the backbone parameters of the MIMO model are updated and passed into an output gate. The output calculation formula of the output gate O is as follows:

[0089] O=δ(W*h (t-1) +U*x (t) +b)⊙tanh(C (t) ),

[0090] Among them, δ is the sigmoid activation function; W and U are the updated linear relationship coefficients; b is the updated bias; C (t) is the hidden state output at time t; x is the output data of the previous stage.

[0091] Step S224: Continuously execute step S222 and step S223, and forward propagate the flotation column process status data input by the model to calculate the model output. The model forward propagation prediction output calculation formula is as follows:

[0092] P (t) =δ(V*O (t) +c),

[0093] Among them, O (t) is the output gate output at time t; V is a fixed coefficient; c is the output unit state at time t.

[0094] Step S225: Based on step S224, the model parameters are back-propagated, and the loss value of the model prediction result is calculated based on the mean square error. The loss function of back-propagation is divided into two parts, one is the loss at the calculation time, and the other is the loss after the calculation time; the loss function calculation formula is as follows:

[0095]

[0096] L=Loss(t)+Loss(t+1),

[0097] Where L is the loss function; t is the current data moment; t+1 is the next data moment; n is the total number of samples of the control variable; y is the true value of the control variable; y p is the predicted value of the control variable; w is the weight coefficient.

[0098] Step S226: Train and optimize model parameters until the model loss value is stable and reaches the optimal value, export and save the flotation column process history control model, and save the data for later model update and analysis.

[0099] Exemplarily, a multi-input multi-output (MIMO) temporal memory network model is established based on the Pytorch structure, as shown in Table 2, which is the main structure design of the model.

[0100] Table 2

[0101] Layer Type (Name) Output Dimensions Parameter scale First layer LS_1 (None, 9, 9) 8848 Dropout (None, 9, 9) 0 Second layer LS_1 (None, 15, 256) 78842 Dropout (None, 15, 256) 0 The third layer LS_1 (None, 15, 256) 1357624 Dropout (None, 15, 256) 0 Fourth layer LS_1 (None, 15, 256) 1357642 Dropout (None, 15, 256) 0 The fifth layer is Timedistri (None, 15, 6) 512

[0102] The hyperparameters in the model training process are defined as: TIME_STEPS = 15, BATCH_SIZE = 64, INPUT_SIZE = 32, OUTPUT_SIZE = 6, PRED_SIZE = 32, CELL_SIZE = 256, LR = 0.0001. Under this hyperparameter combination, after multiple forward propagation and back propagation combinations. Finally, after 800 epochs of repeated training, the output error MSE of the model on the test set is 0.12.

[0103] Specifically, in step S3, the tailings ash content y Ad It is one of the main target variables and reference variables of the flotation column control system. If the tailings ash content is qualified, it is considered that there is no flotation clean coal mixed in the tailings. If the tailings ash content is low, it is considered that the flotation column separation condition is poor. By reading the tailings ash meter, the real-time value of the tailings ash content is obtained to check whether the tailings ash content value is qualified. If it is qualified, the actuator is used to perform regulation according to the obtained predicted control variable data.

[0104] Specifically, in step S4, if the tailings ash value is unqualified, the flotation column process is optimized and controlled by using the predictive control model combined with the optimization mechanism. The cost function is defined based on the control variable data output in step S2 and the tailings ash value to characterize the control deviation between the predicted variable and the expected variable, and then the optimal control sequence is found within the given range of Table 1 based on the quadratic programming (QP) algorithm to minimize the cost function; the specific steps include:

[0105] Step S41, obtaining a cost function based on the tailings ash value and the control variable data; the cost function is expressed as:

[0106]

[0107] Among them, Q and F are both adjustment matrices, such as positive definite / semi-positive definite diagonal matrices; k is the current optimization sample time or sequence number; Y is the sample matrix to be optimized, including control variables and tailings ash values.

[0108] For example, the dimension of the control variable matrix is ​​set to 7, and in addition to the tailings ash value, it also includes 6 control variables: frother given, collector given, feed valve opening, circulation valve opening, water addition valve opening, tailings liquid level motor opening. Set the cost function Y T The FY expansion formula is as follows:

[0109]

[0110] Step S42, continuously optimizing the cost function in step S41 until the minimum value of the cost function is found to obtain the optimal control variable;

[0111] Step S43, extracting input features based on the input state variables of the MIMO model, calculating the contribution values ​​of the input features, sorting the corresponding optimal control variables according to the contribution values, and obtaining the optimal control variable sequence of the control system, that is, the optimal regulation sequence.

[0112] Exemplarily, according to the Local Interpretable Model-agnostic Explanations (LIME) technology, the contribution value of the input feature in the MIMO model is calculated based on the following formula:

[0113] F(x i ) = argminL(x i )+Ω(g),

[0114] Among them, x i is the corresponding input feature; L is the loss function; Ω is the model complexity. Each input feature corresponds to its own control variable. The tuning order of the control variables is determined according to the contribution value, and the system is controlled in this order.

[0115] Specifically, in step S5, the flotation column process system control flow of the tailings ash online feedback based on the predictive control model and optimization algorithm is as follows: Figure 4 As shown, Fc is the output set by the system (the output vector based on the control variable), fr is the flotation column control variable parameter in the allowable fluctuation range, y is the detection input variable, ys is the actual output value of the control variable, ya is the control variable output predicted by the MIMO model, and fa is the tailings ash content under system control. The cycle tuning strategy includes:

[0116] First, read the current state variable. If the state variable is not within the normal range (as shown in Table 1), based on the control variable output by the control feedback of the flotation column production process, use the corresponding actuator to actuate the corresponding valve to adjust the state variable to the normal range.

[0117] Then, the current state variables that meet the conventional interval are input into the MIMO model for control variable prediction to obtain ya;

[0118] Finally, check whether the tailings ash content value read in real time by the tailings ash meter meets the requirements of the beneficiation plant / customer. If qualified, use the actuator to perform regulation according to the obtained ya.

[0119] If it is unqualified, the cost function J is calculated according to the QP optimization method to obtain the optimal control sequence, and the control is performed based on the optimal control sequence and the actuator; if the tailings ash content after control still does not meet the production setting range, repeat the above steps, read the real-time state variables again, and obtain the control variable prediction output based on the predictive control model, and re-judge whether the real-time value of the tailings ash content is qualified, until the tailings ash content reaches the production setting range, stop the update of this stage.

[0120] During the production process, the predictive control model and optimization algorithm are used for cyclic tuning to stabilize the state variables within the normal operating range while ensuring that the tailings ash content index is qualified, thereby ensuring the stability of the control strategy.

[0121] For example, when the tailings ash content is less than 65%, it is determined that the flotation column separation process system does not meet the operating requirements. When the MIMO model works, as shown in Table 3, it is an example of prediction on an industrial flotation column:

[0122] Table 3

[0123] Input variables Output variables Feed flow rate: 296.15m3 / h Collector flow rate: 26.15L / h Feed concentration: 55.26 g / L Foaming agent flow rate 3.17L / h Dilution water flow rate 0m3 / h Feed valve opening 59.16% Foam layer thickness 48.17cm Circulation valve opening 78.38% Circulation flow rate 325.46m3 / h Dilution water valve opening 0% Circulation pressure 0.16MPa Tailings gate motor range 76.24% Tailings ash content 68.59% /

[0124] In this example, all data is saved by the local PLC controller in the local Windows server. The PyQt-based HMI host computer is responsible for embedding the control model and sending the action instructions to the PLC controller, which is responsible for driving the control mechanism. In addition, the local control system data in the example is uniformly connected to the industrial ring network of the coal preparation plant to realize remote sharing of system data. Among them, all data is saved for later model update and analysis.

[0125] Compared with the prior art, the present embodiment provides an intelligent control method for flotation column process based on online feedback of tailings ash content. By monitoring the relevant state variables that affect the tailings ash content index, the state variable data is used to predict the corresponding control variable data using a predictive control model, and the flotation column process production intelligent control with online feedback of tailings ash content is established in combination with an optimization mechanism, and finally the stability of the control strategy is achieved. Through online feedback of tailings ash content, the entrainment of flotation clean coal in the coal slime is effectively prevented, and the clean coal recovery rate of the coal slime flotation process system is improved.

[0126] Example 2

[0127] Another specific embodiment of the present invention discloses an intelligent control system for flotation column process based on online feedback of tailings ash content, such as Figure 5 As shown, including:

[0128] Production parameter detection module, used to collect working condition data in the flotation column production process through detection devices;

[0129] A data set acquisition module is used to store the collected historical operating condition data, and analyze the historical operating condition data based on expert knowledge after preprocessing to obtain the operating condition range requirements that the state variables must meet;

[0130] The prediction control module is used for training the prediction control model; it is also used for obtaining predicted control variable data based on the real-time data of the state variables that meet the working condition requirements in the collected working condition data, using the trained prediction control model, and in response to the qualified tailings ash value read in real time, using the actuator to perform regulation according to the predicted control variable data; in response to the unqualified tailings ash value, using the predicted control variable data and the tailings ash value, based on the quadratic programming optimization algorithm to perform control variable tuning to obtain the optimal control sequence; using the actuator to perform regulation according to the optimal control sequence; in the production process, using the prediction control model and the optimization algorithm to perform cyclic tuning;

[0131] A control module, used for sending the optimal control sequence to a corresponding actuator in the form of a control instruction through a PLC controller;

[0132] The actuator is used to execute the control instructions to complete intelligent regulation.

[0133] Further, obtaining the optimal regulatory sequence includes:

[0134] calculating a cost function based on the tailings ash value and the predicted output of the control variable;

[0135] Optimize with the goal of minimizing the cost function to obtain the optimal control variable output;

[0136] The input features are extracted based on the state variables input by the predictive control model, the contribution values ​​of the input features are calculated, and the corresponding optimal control variables are sorted according to the contribution values ​​to obtain the optimal regulation sequence.

[0137] The system adds detection devices for feed flow rate, feed concentration and supplementary water flow rate in the flotation feed section of the flotation column process; Figure 2 As shown, the actuator adds feed valve opening and water supply valve opening control devices in the flotation feed section; adds circulation volume pressure, circulation volume flow, foam layer thickness, tailings liquid level height and other detection devices in the flotation column equipment section, and adds circulation volume valve opening, tailings lifting motor and other control devices; adds collector flow, frother flow, collector liquid level, frother liquid level detection devices in the flotation reagent section, and adds collector metering pump and frother metering pump control devices; adds tailings ash meter and set time label in the flotation product section.

[0138] It should be noted that the system can be implemented according to the method described in any one of the solutions in Example 1. The relevant parts are referenced from each other and are not described repeatedly in this embodiment.

[0139] Compared with the prior art, the present embodiment provides an intelligent control system for a flotation column process based on online feedback of tailings ash content. By adding corresponding detection and control devices in the flotation column process production process, and based on real-time detection and intelligent regulation of tailings ash content indicators and related variable parameters, the industrial flotation column process can be operated more efficiently, the clean coal recovery rate can be effectively improved, and the consumption of flotation agents can be reduced, thereby achieving the purpose of increasing the efficiency of mineral processing enterprises and improving the intelligence level of mineral processing plants.

[0140] Those skilled in the art will appreciate that all or part of the processes of the above-mentioned embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, wherein the computer-readable storage medium is a disk, an optical disk, a read-only storage memory, or a random access memory, etc.

[0141] 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 changes or substitutions that can be easily conceived by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. An intelligent control method for flotation column process based on online feedback of tailings ash content, characterized in that: The steps include: Real-time collection of operating data related to tailings ash content during flotation column production; Based on the state variable data that meets the working condition requirements in the collected working condition data, the predicted control variable data is obtained by using the trained predictive control model; In response to the tailings ash value read in real time being qualified, using an actuator to perform regulation according to the predicted control variable data; In response to the tailings ash value being unqualified, using the predicted control variable data and the tailings ash value, based on a quadratic programming optimization algorithm, control variable tuning is performed to obtain an optimal control sequence; Using an actuator to perform regulation according to the optimal regulation sequence; During the production process, the predictive control model and optimization algorithm are used to perform cycle tuning.

2. The intelligent control method of flotation column process based on online feedback of tailings ash content according to claim 1 is characterized in that: The predictive control model is trained based on a multi-input multi-output long short-term memory model, comprising: Collect historical control data in the flotation column production process to obtain a flotation column process production data set; wherein the state variable data in the production data set is used as the input data set, and the corresponding control variable data is used as the output data set; Based on a multi-input and multi-output long short-term memory model, training is performed using the loss function calculated by the mean square error of the control variables to obtain a trained predictive control model.

3. The intelligent control method of flotation column process based on online feedback of tailings ash content according to claim 2 is characterized in that: Obtaining the optimal regulatory sequence comprises: Obtaining a cost function based on the tailings ash value and the control variable data; Optimize with the goal of minimizing the cost function to obtain the optimal control variable output; The input features are extracted based on the state variables inputted by the predictive control model, the contribution values ​​of the input features are calculated, and the corresponding optimal control variables are sorted according to the contribution values ​​to obtain the optimal regulation sequence.

4. The intelligent control method of flotation column process based on online feedback of tailings ash content according to claim 3 is characterized in that: The cost function is expressed as: Among them, Q and F are both adjustment matrices; Y is the sample matrix to be optimized, including the tailings ash value and the control variables; k is the time or serial number of the current optimized sample.

5. The intelligent control method of flotation column process based on online feedback of tailings ash content according to claim 3 is characterized in that: According to the local interpretable model-independent explanation technique, the contribution value is calculated based on the following formula: F(x i )=argminL(x i )+Ω(g), Among them, x i is the input feature; L is the loss function; Ω is the complexity of the model.

6. The intelligent control method of flotation column process based on online feedback of tailings ash content according to claim 2 is characterized in that: The loss function is: L=Loss(t)+Loss(t+1), Among them, L is the loss function; t is the current data moment; n is the total number of control variables; y is the true value of the control variable; y p is the predicted value of the control variable; w is the weight coefficient.

7. The intelligent control method of flotation column process based on online feedback of tailings ash content according to claim 2 is characterized in that: The state variables include feed flow, feed concentration, dilution water flow, circulation flow, circulation pressure, foam layer thickness and tailings ash content; the control variables include foaming agent addition amount, collector addition amount, feed valve opening, circulation valve opening, water addition valve opening and tailings gate motor.

8. The intelligent control method of flotation column process based on online feedback of tailings ash content according to any one of claims 1 to 7, characterized in that: When the collected operating data does not meet the operating requirements, the control variables output by the flotation column production process control feedback are sent to the corresponding actuator to complete the control until the operating requirements are met.

9. An intelligent control system for flotation column process based on online feedback of tailings ash content, characterized in that: include: Production parameter detection module, used to collect working condition data in the flotation column production process through detection devices; Predictive control module, used for training predictive control models; It is also used to obtain predicted control variable data based on the real-time data of state variables that meet the working condition requirements in the collected working condition data, using the trained predictive control model, and in response to the qualified tailings ash value read in real time, use the actuator to perform regulation according to the predicted control variable data; in response to the unqualified tailings ash value, use the predicted control variable data and the tailings ash value to tune the control variables based on the quadratic programming optimization algorithm to obtain the optimal regulation sequence; Using an actuator to perform regulation according to the optimal regulation sequence; During the production process, the predictive control model and optimization algorithm are used to perform cycle tuning; A control module, used for sending the optimal control sequence to a corresponding actuator in the form of a control instruction through a PLC controller; The actuator is used to execute the control instructions to complete intelligent regulation.

10. The intelligent control system for flotation column process based on online feedback of tailings ash content according to claim 9, characterized in that: Obtaining the optimal regulatory sequence comprises: calculating a cost function based on the tailings ash value and the predicted output of the control variable; Optimize with the goal of minimizing the cost function to obtain the optimal control variable output; The input features are extracted based on the state variables input by the predictive control model, the contribution values ​​of the input features are calculated, and the corresponding optimal control variables are sorted according to the contribution values ​​to obtain the optimal regulation sequence.

Images