TR202006947A2 - System and method for ai controlling mill operations to lower electricity consumption - Google Patents

Abstract

The present invention relates to a method for automatically reducing electricity/energy consumption and controlling processes occurring in a mill by means of mill parameters that result in higher efficiency in industrial production, preferably in a cement manufacturing process. The method of the invention is an AI (artificial intelligence) based method and includes a sensor motion prediction and optimization system. The AI method includes an end-to-end automated process for sensor data analysis, machine learning model training optimized for vertical/horizontal mills, and operationalization of model results with an optimization system for parameter management. The invention also relates to a system for controlling processes occurring in a mill by means of an AI (artificial intelligence) based method.

Description

DESCRIPTION AI CONTROLLED TO REDUCE ELECTRICITY CONSUMPTION SYSTEM AND METHOD FOR MILLING OPERATIONS Technical Field of the Invention This invention is used in industrial production, preferably in a cement manufacturing process. for control operations occurring in a mill through its parameters. which automatically leads to reduced electricity/energy consumption and higher efficiency. relates to the method. The method of the invention is an AI (artificial intelligence) based method and the sensor includes motion prediction and optimization system. AI method, sensor data end-to-end automated process for analysis, vertical/horizontal mills machine learning model training and model results optimized for It includes making it operational with an optimization system for management. Meet also, through an AI (artificial intelligence) based method, It relates to a system for controlling transactions.

Operation of a mill with lower electricity consumption and higher efficiency The inventive system for milling is done with a minimal need for human intervention. It is set up to automatically generate and instruct.

Background of the Invention Cement production has a standard process that has not changed over the years. Cement production is responsible for 12-15%* of industrial energy use in the world. Each ton of cement produced can use 80 to 100 kWh of electrical energy.

The cost of electricity consumption is more than 21%* of the cement industry cost. is doing. Electric mills mostly used in mixing & grinding processes consumed by In particular, 4.0 solutions are industry-specific, massive energy consumption, increased cost challenges, and overall process complexity. can manage.

In conventional cement plants, vertical and/or horizontal mills, 3-shift production are instructed by the operators during These instructions, operating parameter are determined based on the set values (SP), which are It is determined according to the experience and decision of the operator during the operation. Operators continually sensor values to decide the setting parameter values periodically, It displays the production and the status of the machine.

In a typical automated control system, each machine connected to a mill The data from the measuring instrument is sent to the PLCS (programmable logic controller element). is being transmitted. The measuring tool includes all the sensors connected to the different parts of the mill and It contains the operating parameter values set by the operators. The data obtained It includes all sensor values and set values instructed by the operators.

Data obtained from PLCS's input/output cards (respectively, analog/digital (A/D) card, digital/analog (D/A) board) has a small capacity and execution built inside the PLCS It is processed through a central processing unit (CPU) with speed and PID (proportional integral differential) and ON/OFF controlled.

The computer simply reads the data given as input and output from PLCS, visually identifiable, viewable, recordable and controllable Function of converting a physical quantity (physical quantity data) to its data is conducting.

The application of this method without an automatic control system will cause sensor movements, of various set values depending on machine condition, production status and energy consumption. difficulties in controlling simultaneously in a time window brings. It is mainly the operator's responsibility to change each SP over time. to control by applying intuition, to different experience level of each operator and analysis of multidimensional sensor motion in a time window. It is difficult due to the real-time human attention capacity. In production, non-optimal setpoint instructions, interaction between setpoints and sensors a machine to a fault or near-failure condition observed to be possible. Moreover, it increases both efficiency and electricity consumption. manual operation of all SPs together as a method for increasing the efficiency of reducing production. It is impossible to optimize each SPS sequentially or to control each SPS sequentially. Additional Additionally, in the absence of the operator, existing efficiency is difficult to maintain or increase.

The aforementioned inadequacies are also accompanied by grinding during data & process analyses. Therefore, the operation during an industrial production, preferably cement manufacturing, of a more efficient and stable manufacturing process that anticipates changes in for the machine in the process in order to minimize the energy consumption and It is necessary to build a system that provides the necessary adjustment parameters.

For this purpose, the inventors of the present invention are responsible for the processes occurring in a mill. provides a method for checking Trained machine learning uses models and makes optimal use of the results of the models. It includes an optimization algorithm for selecting parameters.

Brief Description of the Invention An object of the invention is in industrial production, including cement and mining. employee, regardless of machine make and type, for a vertical or horizontal mill. overcome the disadvantages of the known systems and methods described above. is to come.

According to one embodiment of the invention, the crushing or A method for controlling grinding operations is provided. Methods contains: from the set values instructed during the elapsed time window, and to estimate the next period electricity consumption of a mill from sensor movement. a method of training an artificial intelligence and principles of machine and process operation for operating the mill with maximum efficiency and minimum electricity consumption according to on the calculation of optimal setpoint values for the next period. optimization method.

According to another embodiment of the invention, crushing in vertical/horizontal mills or a system for controlling grinding operations is provided.

The method for controlling the processes occurring in the mill of the invention, It provides an improvement in the whole process that eliminates human error or inadequate operation.

The method of the present invention, historical data of sensors from mill operations, setpoint and monitoring of energy consumption and instantaneous data captured by the sensors associated with the mill. taking into account the cement manufacturing process, in particular the milling manages its operation. The method of the present invention is independent of operator intervention. stable and efficient grinding / mixing process that takes place inside the mills. enables it to be carried out.

Artificial intelligence models used in the method of the invention do not require any operator intervention. the most effective way in terms of time and energy expenditure in the manufacturing process without management of the mill in cement manufacture by the critical setting parameters that provide it provides.

Other objects, features and advantages of the present invention, together with the accompanying figures When considered, it is clear from the detailed description of preferred applications below will become.

Brief Description of Figures Figure 1 is a cement production process flow used in an application of the present invention. is the diagram.

Figure 2 is a vertical or horizontal mill used in an application of the present invention. is a process flow diagram showing the process.

Figure 3 shows the processes occurring in a mill according to an embodiment of the present invention. is a flowchart showing the method for controlling

Figure 4 shows an application and training of the method in an industrial production process according to the invention. It is a flow diagram showing the methods.

Figure 5 is a flow showing the working of the inventive method in a system according to the invention. is the diagram.

References 1: mill 2: PLC 3: IOT gateway 4: IOT Data Collector : Data storage unit 6: Time Series Data Parser 7: Model Manager Detailed Description of Preferred Applications The invention is the optimal setting for milling an AI in a Model Manager (7). It is applied to determine the point values of the It relates to a method for controlling transactions. The method in question is It contains the steps: 1. Mill containing manual setting parameter data using last time window measuring and collecting historical data on 2. Mill containing manual setting parameter data using last time window pre-processing of historical data about 3. The next time window is preset to predict sensor movement metrics. the processed manual tuning parameter data to a trained sensor motion model. nutrition, 4. Sensor movement for estimation of next time window electricity consumption feeding the output of the model to a trained electricity consumption model, . Optimal next time window setpoint values window setpoint value setting and each possible option's electrical with an optimization algorithm that evaluates consumption and efficiency estimates. calculation, 6. Instructing the mill with optimal setpoint values, The method of the invention is firstly based on the manual parameters of the sensor movement. estimation of electricity consumption based on the predicted sensor movement afterward. provides the forecast.

Flow diagram of a cement production process according to an embodiment of the present invention It is shown in Figure 1. Cement production process, cement raw material limestone supplied by removing the stones and cutting them into small pieces (101) and transporting limestone and other raw materials to warehouses and their processing (102) It starts with homogenization by mixing first. afterward, before homogenized materials, to obtain finer raw material powder they are fed to the raw mill where they are crushed and ground (103). Process, raw powder flour taking it to the warehouse and homogenizing the new production batch with the previous batches (104) and raw flour to clinker by a chemical reaction initiated by calcination combustion. It continues with its feeding (105) into the kilns where it is transformed - a by-product of cement.

Afterwards, the newly produced hot clinker is cooled and taken to the warehouses (106).

Clinker and different additives are used to obtain various types of cement final product. They are fed into cement mills, where they are mixed and ground (107). produced cement is taken to the warehouses according to its type (108).

The method of the present invention involves controlling a mill belonging to such a process. covers.

The sensor movement model and electricity consumption model used in the method of the invention, next time window sensor movement metrics and next time window electrical for consecutive operation/operation for estimating the consumption of is being trained.

The sensor movement pattern and the electricity consumption pattern are based on what happens in a mill (1). It is trained and executed in a Model Manager (7) that controls the processes.

The method of the present invention is a vertical or horizontal mill (1) operating according to Figure 2. can be applied for. Raw material is transported from the warehouse via production lines and continuously fed to the machine (201). Vertical or horizontal mill(s) (1) different using mechanical processes (202), it simply mixes the fed materials, grinds and crushes. The powdered end product is determined as an output. while the other materials remain in the mill until they have sufficient fineness (203).

According to the invention, the mill (1) is a raw or cement mill used in industrial production. vertical or horizontal used for crushing, grinding or mixing can be a mill.

A cement mill produces hard, nodular clinker from the cement kiln and specific additives. It is the equipment used to turn the materials into fine gray powder, which is cement. Most cement is more effective in ball mills today and also more effective than ball mills. It is milled in vertical roller mills.

A raw mill, raw materials pulverized during cement manufacture The cement is fed into the kiln, where it is converted into clinker and then cement. It is ground in the cement mill with specific additives to produce it. Raw & The grinding and mixing processes occurring in cement mills are largely chemistry and quality (and therefore physical properties) of the cement end product defines a major impact on the efficiency and energy efficiency of the entire manufacturing process. has an effect.

The method of the present invention is used for industrial applications including cement and mining. for all mills used in the grinding and mixing processes of production applicable.

In a preferred embodiment of the invention, the mill is used in cement production. is used.

The study of a production process using the method according to the invention is shown in Figure 3. is explained. Depending on the process in question, the machine can be operated automatically or mechanically. is being launched (301). Pre-coded conditions are replaced automatically or manually. When it is brought, the program of the machine on which the artificial intelligence model is applied starts. (302). The program is started with the relevant systems (303). Sensor data and setting parameter data is measured with measuring devices (304). Data is collected from measuring instruments (305) and the collected data are used to control data quality and rendering properties from raw data. It is preprocessed by means of (306). preprocessed data, one after the other to trained machine learning models (sensor movement model & electricity consumption model) is given (307). Efficiency outputs of trained machine learning models is given to the optimization algorithm (308). Optimal setting values in the interface optimization displayed and instructed to mill through the PLC (2) system It is calculated by the algorithm (309). Model & algorithm results and each The raw data in the sample is stored in the Data storage unit (5) (310) for further analysis.

The program automatically or manually when the pre-coded conditions are met. ends (311). The program terminates after the termination mechanism works (312). Steps 305 to 310 continue each predetermined time until the program ends. is repeated continuously for the time window. Preprocessing means that milling is obtained by means of measuring instruments of historical data; so that each data point represents the value in the same time window. This includes synchronizing the data and preparing the data for modelling.

Preprocessing of historical data about the mill includes the following steps is: 1. Obtaining historical data of the mill by means of measuring instruments; 2. The obtained data represent the value of each data point in the same time window. data to synchronize and prepare for the model preprocessing; 3. Live data collection, data quality and problems caused by data anomaly implementation of rules and warnings for its elimination; Preprocessing to prepare data for feeding into the model, data quality control of the past stable production periods to determine the model training set. definition and time window synchronization of data with different frequencies. covers. Historical data about the mill collected by the measuring devices (1), manual adjustment parameters, actual sensor values and electricity consumption values.

Measuring devices that collect historical data about the mill (1) one or more of the following: contains more: I. sensor data that measures the actual sensor values, linked to different parts of the mill collection tool, ii. mill operating parameters manually set by the operators. programmable logic controller element that enters manual setting parameters (PLC) (2) system, iii. incremental amount of energy consumed connected to different parts/units of the mill electricity meter.

In the method according to the present invention, the actual sensor values measured from a mill, temperature, pressure, working speed, vibration, current, flow and position of part of the mill and their contain parameters that include, but are not limited to, any combination of is doing.

which allows the measurement of any of the above-mentioned parameters. The sensor data collection tool can be connected to parts of the mill.

Setting parameter collection tool, manually set by operators it saves the manual setting parameters, which are the mill operating parameters. Promise These parameters need to be adjusted periodically.

Manual setting parameters, working speed, material feed amount, to be applied pressure and the position of a specific part of the mill and any of these contains the combination.

of the mill position that must be set manually by the operators. The specific part includes valves such as water and circulation.

Electricity meter connected to different parts/units of the mill, consumed energy measure the amount incrementally. mill connected to an electricity meter parts/units containing motor, fan, separator and any combination thereof selected from the group. The final time window of the cement manufacturing process, the optimal setting of the algorithm predetermined in minutes before each time it calculates the parameter It refers to the said data collected during the optimum period. The next time window of the cement manufacturing process, the optimal setting of the algorithm the optimum in minutes after each time it calculates its parameters. Optimal next time window setpoint values are in the method according to the invention. depending on the sensor values used and the output of the trained machine learning models. determined by the optimization algorithm. Optimization algorithm, while minimizing energy expenditure, maximizing efficiency and mechanical and operational setting values for the next time window that vary in restrictions. calculates.

Time window, predictive power of models, operational constraints and energy consumption as any period in minutes depending on the trend and sensor movement can be adjusted.

The method of the present invention includes an optimization algorithm.

Optimization algorithm first, final set values, operational and mechanical possible set values for the time window after restrictions finds. Operational and mechanical constraints, the mill used in cement manufacture is variable due to its type.

Operational and mechanical restrictions must be determined by technically qualified persons in the relevant technical field. known and may vary depending on the type of mill.

Operational constraints in particular, chemical modifications and quality of the process may include setting limits determined depending on the requirements in terms of

Adjustment of the machine due to mechanical restrictions, default settings It may contain restrictions for the minimum and maximum values of the parameters. Promise These restrictions also apply to the rate of change of parameters in each setting. may be related to limitations, because the maintenance of production requires adjustment within a single period. can be prevented due to frequent changes in parameters.

Therefore, optimal set values are related to this process and machine working principles.

The optimization algorithm results in the highest possible efficiency within operating limits. It works towards the possible range of adjustment values for the calculation of the adjustment value.

The optimization algorithm uses an optimization criterion so that the highest depending on the possible efficiency and the lowest possible electricity consumption setpoint range can be reached.

The operation of the optimization algorithm used in the method according to the invention is as follows: contains 1. The latest manual setting parameters, process & machine working principles creation of possible next time window set value options; 2. An optimization aimed at higher efficiency and lower electricity consumption evaluating the electricity consumption estimates for each possible option with the running the optimization algorithm; The best possible set value can be determined among the possible set value options.

In a preferred embodiment of the invention, the possible next window of time is set first. value options are determined. Next, the electricity consumption for each option is estimated and the optimal next time window set value is matched to the optimization criterion. is selected accordingly.

The optimization algorithm used in the method of the present invention consists of permutations. brute-force, gradient descent (gradient-descent) or a similar algorithm.

The method for training a sensor motion model according to the present invention is as follows. It contains the steps: 1. Raw manual setting parameter passed through the corresponding IOT data collector (4) system taking values and raw sensor values, 2. Model features by means of feature generation methods from collected raw data creation and input of the generated features into the sensor motion prediction models. given as, 3. Training and testing different machine learning algorithms for sensor motion prediction to be made, 4. Selecting the best model according to the success criteria, the model's specific mill (1) termination as a sensor motion prediction model from manual tuning parameters.

The inputs are the past manual setting parameter data, the sensor to be tested to generate features for each of the motion metric prediction model. is used. A feature is one of the individual units on which the analysis or prediction will take place. An attribute or attribute shared by all. For any attribute, model It can be a feature as long as it's useful. Feature generation is combined with the use of domain knowledge through data mining techniques. It is the process of creating a new feature from one or more existing features. This The process adds new information that will be available during model construction. created features are used to improve the performance of machine learning algorithms. can be used.

Machine learning algorithms are tested for the sensor motion prediction model and are trained and the said algorithms are based on artificial neural networks and stochastic gradient descent. can be selected.

Success criteria for sensor motion model in algorithm selection, prediction error, estimation speed and ease of application.

Best model, sensor movement from the manual setting parameters of the specific mill terminated as a prediction model.

From another perspective, the invention helps to train an electricity consumption pattern recognition model. relates to a method. This method includes the following steps: 1. Raw sensor values and raw sensor values that have passed through the relevant IOT data collector (4) systems obtaining the electricity consumption values, 2. Estimation period by means of feature generation methods from collected raw data creating model features that reflect sensor movement during created giving the features as input to the electricity consumption forecasting model, 3. Different machine learning pattern recognition algorithms for electricity consumption estimation training and testing, 4. Selection of the best model according to success criteria and similar metrics, specific electricity consumption pattern recognition model from sensor motion features for mill termination as.

Electricity consumption values, electricity consumed by the mill in a period refers to the amount. The electricity consumption of a mill, during this period It is affected by the state of the machine described by the value of its sensors.

Electricity consumption pattern recognition from sensor data for model specific mill (1) model is terminated.

Success criteria for electricity consumption pattern recognition model in algorithm selection, estimation error, estimation speed and model generalizability.

Pattern recognition used in training electricity consumption pattern recognition model machine learning models, artificial neural networks, random forest, decision trees, gradient the increment classification method and ensemble/assembly models and models can be selected from the group containing the community.

According to the present invention, it is aimed to control the processes occurring in a mill. The method for the training also includes the following steps of the trained sensor motion model. It is characterized by being trained by the method: 1. Raw manual setting parameter passed through the corresponding IOT data collector (4) system taking values and raw sensor values, 2. Model features by means of feature generation methods from collected raw data creation and input of the generated features into the sensor motion prediction models. given as, 3. Training and testing different machine learning algorithms for sensor motion prediction to be made, 4. Selection of the best model according to estimation error, estimation speed and ease of application, sensor motion prediction model from manual adjustment parameters of specific milling (1) termination of the model. and the trained electricity consumption model by means of the method including the following steps being trained: 1. Raw sensor values and raw sensor values that have passed through the relevant IOT data collector (4) systems obtaining the electricity consumption values, 2. Estimation period by means of feature generation methods from collected raw data creating model features that reflect sensor movement during created giving the features as input to the electricity consumption forecasting model, 3. Different machine learning pattern recognition algorithms for electricity consumption estimation training and testing, 4. Estimation error, estimation speed, and model generalizability and the most similar metrics Choosing the good model, electrical from the sensor movement characteristics for the specific mill Termination of the model as the consumption pattern recognition model.

According to the present invention, it is aimed to control the processes occurring in a mill (1). method for the trained sensor movement model and the trained electricity consumption characterized by the model being trained to run/operate sequentially. is being done.

On the other hand, the invention is for an AI to mill in a Model Manager (7). in a mill where it was applied to determine the optimal setpoint values. It relates to a system for controlling transactions that occur. According to the invention The system includes the following: being instructed by a Model Manager (7) operating an artificial intelligence model a mill configured to (1), measuring devices for measuring data related to the mill (1), collection of sensor data for the operational management of the mill (1) and existing with optimal set values for the next time window of setting parameter values a PLC (2) system to override, For receiving raw sensor data in real time from PLCS and collected sensor data at periodic time intervals defined by the system administration. an IOT gateway (3) for delivery to the centralized IOT data collector (4), of predefined file extension type defined by system administration Storing the raw data collected from each mill in the data storage unit (5) an IOT data collector (4) for IOT Data for last N minutes defined by system administration In order to receive raw sensor data of each mill (1) from its totalizer (4), for data preprocessing and for prediction & pattern recognition models a time series data parser (6) for generating features, the last N minutes of received sensor data, sensor motion prediction models and electricity consumption prediction model, prediction & pattern recognition machine learning Receiving the forecast result for training and executing algorithms a model to be sent to the relevant machine learning model or models for manager (7), A data storage unit for storing final settings for operational needs Programmable logic controller (PLC) (2) or programmable designed for the control of manufacturing processes such as controller, assembly lines, and adapted from an industrial digital computer, or robotic devices, or high reliability, any activity that requires ease of programming and process error diagnosis.

An IOT (Internet of Things) Gateway (3) (IOT GW) for various types of data collection of sensor data placed on the mills (1) PLC (2) to receive new data values for Responsible for setting the new Setting Sensor data on. IOT GW (3) factory network in a central Data Center that can be placed inside or depending on the factory network infrastructure. can be positioned. from the collection and exposure of newly calculated Tuning Sensor data to the corresponding IOT GWS (3). It is a kind of Data Station that is responsible for releasing it. IOT Data Collector (4) a can manage a factory or more. Data connection, TCP (transmission control via protocol).

Data storage unit (5), raw sensor data from mills (1) and calculated It is the main repository for sensor data. System logs, sensor data, and application logs All types are stored in the Data storage unit (5). Data, txt, csv, XlS, parquet, json, etc. It can be stored in different formats such as

Time Series Data Parser (6), Machine Learning in Solution model(s) raw sensor data to be used as input data on the incoming mill responsible for its preparation.

Model Manager (7), Machine Learning Model or Models to manage and result It is the main editor for the preparation of the setting.

Machine Learning Models show that the sensor movement for the next time window and mill specific for a specific period for estimating electricity consumption is trained by the model manager (7). The optimization algorithm also model manager for calculating optimal setting values for the next time window It is operated by (7).

System for controlling the processes occurring in a mill (1), here the system in question is configured with: 1. real-time work through work on real-time data, and real time output for the next time window; 2. The time window after the given setting parameters is the sensor movement and electrical forecasting its consumption; 3. Finding the optimal set values as output for the next time window.

System for controlling the processes occurring in a mill (1), may contain more than one mill (1).

Figure 4 shows the working of the method according to an application of the invention and how the models are represents being trained.

In one embodiment of the AI-controlled system according to the inventive method, a one-time model There are training steps. When models are trained, their performance deteriorates. works in place. Steps 401 through 407 are model training steps. 409 and Steps 412 are real-time model work steps.

AI control begins when training begins (401). Historical sensor data, setting parameter and electricity consumption data are obtained (402). historical data, data quality by applying controls and generating features from raw data. is processed (403). The preprocessed data is used to train the sensor motion prediction model. is given as input (404). The preprocessed data is used for the electricity consumption forecasting model. It is given as input for training (405). Optimization algorithm target and constraint functions are determined (406). Machine learning models and optimization algorithm for them to run consecutively until satisfactory results are achieved. being trained (407).

Trained models for real-time remote automatic control system recorded (408). In each time window, new data is collected (409). Each In the time window, the collected data is preprocessed, input into the recorded models. and models produce optimal tuning parameters (410). Each in the time window, the instructed setting parameters and properties are recorded (41 1).

Figure 5 is a flow showing the working of the inventive method in a system according to the invention. is the diagram.

In a system for controlling the processes occurring in a mill (1), Firstly, the Sensor data, the operational management of the mill (1) by the mill operator and It is collected by the PLC (2) unit for the notification of the purposes (, It receives real-time or near-real-time raw sensor data from PLCS (2) (502). It sends it to the centralized IOT Data Collector (4) at time intervals. (503). from each mill (1) in volume Data Storage (5) of type file extension it stores the collected raw data (504). Time Series Data Parser (6), system collected from the IOT Data Collector (4) for the last N minutes defined by management It takes the raw sensor data of each mill (1) and uses the Machine Learning (ML) model. needed as part of ML feature extraction to be used as input making data preparations (505). Last N minutes processed for each mill (1) sensor data is sent to the relevant Model Manager (7) (506). Model Manager (7), corresponding ML model to get the prediction result for each Setting sensor value, or It sends the last N minutes of processed sensor data received to the models (507). Model When the Manager (7) gets all the predicted Tune Sensor value, the result setting will be prepared (508). Result Setting, centralized Data for operational needs is stored in the storage unit (5). Result setting to be used by PLC (2). To IOT Data Collector (4) for Estimated Tuning Sensor Values being sent ( is being sent Method to return the Estimated Adjustment Sensor values to PLCS for recovery from the Operator. (2) sends (512).

The method gives a different perspective and time series to the holistic machine operating system. It offers a different approach to the estimation problem. In the found method, machine operation system, external control factors, internal state, and in this case electricity consumption divided into three subsystems. From this perspective, manual adjustment parameters refer to external control factors, principal sensor data refer to internal state and electricity consumption is the result of the internal state. The approach is between the three subsystems in question. sensor movement pattern and electricity consumption depending on interaction and cause-effect relationship It uses two training methods as its model. In the first, external factors (manual setting parameters) affect the internal state (main sensor data), and this time series relationship, the sensor is included in the motion pattern. In the second, the internal state (main sensor data) is electrical consumption and this time series relationship is included in the electricity consumption model. is happening. The perspective and approach in question is the interaction between the three subsystems. taking into account delays.

AI application system with different number of sensors and types of sensors It is designed for flexible and hybrid operation to be applied in mills.

In addition, real-time application data preprocessing steps, sensor data anomalies, sensor corruption, or data flow problems. It is designed to prevent interruptions.

REMOVING & CRUSHING LIME STONE STORAGE AND PRE-HOMOGENIZATION OF RAW MATERIAL RAW MILL CRUSHING & GRINDING PROCESSES RAW FLOUR HOMOGENIZATION & FILTERING CALSINATION - BURNING OF RAW FLOUR TO CLINKER COOLING & STORING CLINKER MIXING CLINKER WITH ADDITIVES IN CEMENT MILLS STORAGE CEMENT RAW MATERIAL FEED .1.Hand GRINDING & MIXING WITH DIFFERENT KINDS OF PRESSURE OUTPUTTING THE POWDERED PRODUCT 3134 .

STARTING MECHANISM (AUTO OR MANUAL) PROGRAM BEGINNING STARTING ARTICLE ABOUT THE MILL SENSORS & SETTING DATA COLLECTION PARAMEIRES' (FLEXIBLE) TERMINATION MECHANISM (AUTOMATIC OR MANUAL) DATA PREPROCESSING / ' MACHINE LEARNING MODELS WORK EFFICIENCY OPTIMIZATION IS WORKING MODEL INSTRUCTIONS TO PLC SYSTEM RETURNING WORKING END OF PROGRAM - AI CONTROL RECEIVING HISTORICAL DATA PRE-PROCESSING OF HISTORICAL DATA SENSOR BY AI ESTIMATING MOVEMENT MODEL TRAINING ELECTRICITY BY AI ESTIMATED CONSUMPTION MODEL OPTIMIZATION TARGET & RESTRAINT ADJUSTMENT TO SATISFACTORY RESULTS UNTIL MODELS & END TO END OF OPTIMIZATION TRAINING THROUGH A REMOTE AUTOMATIC CONTROL SYSTEM MAKING MODELS FUNCTIONAL IN REAL TIME COLLECTING NEW DATA IN ALWAYS WINDOW GENERATION OF OPTIMAL SETTING PARAMETERS IN THE ALWAYS WINDOW 11119 INPUT DATA AND INSTRUCTED SETTING PARAMETERS FOR THE ALWAYS WINDOW

Claims (3)

It is a method for controlling the operations occurring in a mill (1), in which an artificial intelligence model is applied in the Model Manager (7) to determine the optimal setpoint values for the mill (1), and the method includes the following: , where the sensor movement model and electricity consumption model, Measurement and collection of historical data about mill (1) containing manual tuning parameter data using the latest time window, Historical data about mill (1) containing manual tuning parameter data using the latest time window Preprocessing, Feeding manual tuning parameter data to a trained sensor movement model for estimating next time window sensor movement metrics Feeding the outputs of the sensor movement model to a trained electricity consumption model for estimating next time window electricity consumption Feeding the optimal next time window setpoint values , calculating with an optimization algorithm that creates the possible next time window setpoint value setting and evaluates the electricity consumption and efficiency estimates of each possible option, instructing the Mill (1) with the optimal setpoint values, it is trained to run/operate sequentially. The method according to claim 1, wherein preprocessing includes: Milling (1) obtaining historical data by means of measuring devices; Preprocessing the data to synchronize the obtained data so that each data point represents the value in the same time window and prepare it for the model; Enforcement of rules and warnings for live data collection, data quality and elimination of problems caused by data anomaly. The method according to claim 1, wherein the historical data about the mill (1) collected by the measuring devices (1) includes manual adjustment parameters, actual sensor values and electricity consumption value. The method according to claim 1, wherein the measuring devices (1) collecting historical data about the mill (1) contain one or more of the following: i. sensor data collection instrument connected to different parts of the mill (1), measuring the actual sensor values, ii. The programmable logic controller element (PLC) (2) system, which enters the manual setting parameters -mill operating parameters- which are set manually by the operators, iii. electricity meter connected to different parts/units of the mill (1), which progressively measures the amount of energy consumed. The method according to claim 4, wherein the actual sensor values include, but are not limited to, temperature, pressure, operating speed, vibration, current, flow and position of a portion of the mill (1) and any combination thereof. The method according to claim 4, wherein the manual adjustment parameters include the working speed, the material feed amount, the pressure to be applied and the position of a specific part of the mill (1) and any combination thereof. The method according to claim 4, wherein the parts/units of the mill (1) connected to an electricity meter are selected from the group consisting of motor, fan, separator and any combination thereof. It is a method according to claim, where the operation of the optimization algorithm includes the following: 1. Generating the possible next time window setting value options in terms of the last manual setting parameters, process & machine operating principles; 2. Execution of the optimization algorithm, which evaluates the electricity consumption estimates for each possible option, with an optimization criterion aimed at higher efficiency and lower electricity consumption; The method of claim 9, wherein said optimization algorithm is brute-force, gradient-descent, or a similar algorithm based on generation, testing, and selection of permutations. It is a method for training a sensor motion model and includes the following: 1. Retrieval of raw manual tuning parameter values and raw sensor values passed through the respective IOT data collector (4) system, 2. Generating model features from the collected raw data by means of feature generation methods and generated giving the features as input to the sensor motion prediction models, 3. Training and testing different machine learning algorithms for sensor motion prediction, 4. Selecting the best model according to the success criterion, finalizing the model as a sensor motion prediction model from the manual tuning parameters of the specific milling (1). The method for training the sensor motion prediction model according to claim 11, wherein the machine learning algorithms of step 33 can be selected from the group consisting of artificial neural networks and stochastic gradient descent. It is a method for training an electricity consumption forecasting model and it includes the following: 1. Retrieval of raw sensor values and raw electricity consumption values from the respective IOT data collector (4) systems, 2. The sensor movement during the forecast period by means of feature generation methods from the collected raw data. generating reflective model features; Inputting the generated features as input to the electricity consumption prediction model, 3. Training and testing different machine learning pattern recognition algorithms for electricity consumption prediction, 4. Selecting the best model according to success criteria and similar metrics, electrical sensor motion features for the specific mill (1) Termination of the model as the consumption pattern recognition model. The method for training an electricity consumption prediction model according to claim 13, wherein the machine learning pattern recognition algorithms of step 33 can be selected from the group consisting of artificial neural networks, random forest, decision trees, gradient increment classification method, and ensemble/assembly models. The method for controlling processes occurring in a mill (1) according to claims 1 to 10, wherein the trained sensor movement model is trained according to the method in claim 11 and the trained electricity consumption model is trained according to the method in claim 133. The method according to claim 15, wherein the trained sensor movement pattern and the trained electricity consumption pattern are trained to operate/operate sequentially. It is a system for controlling the processes occurring in a mill, in which an artificial intelligence model is applied to determine the optimal setpoint values for the mill (1) in a Model Manager (7), the system includes the following: Model Manager (7) running an artificial intelligence model a mill (1) configured to be instructed by the mill (1), measuring devices for measuring data about the mill (1) to collect sensor data for operational management of the mill (1) and override existing setting parameter values with optimal set values for the next time window a PLC (2) system, an IOT gateway (3) for receiving raw sensor data in real time from the PLCS and sending the collected sensor data to the centralized IOT data collector (4) at periodic time intervals defined by the system management, predefined file defined by the system administration extension, an IOT data collector (4) for storing the raw data collected from each mill (1) in the Data storage unit (5), the IOT Data Collector (4) for the last N minutes defined by the system administration, the raw data collected (1) of each mill a time series data parser (6) for retrieving sensor data, preprocessing the data and generating features for prediction & pattern recognition models, sensor motion prediction models and electricity consumption prediction model, prediction & pattern recognition a model manager (7) for training and executing machine learning algorithms and sending the prediction result to the relevant machine learning model or models for receiving, Viii. in a data storage unit (5) for storing final settings for operational needs. 18. The system of claim 173, wherein said system is configured with: 1. real-time operation through working on real-time data and real-time output for the next time window; 2. estimating the time window sensor movement and electricity consumption after the given setting parameters; 3. Finding the optimal set values as output for the next time window.