An intelligent sectorial delineated and hierarchical control system for controlling flour mills

Abstract

Proposed is a self-optimizing, adaptive processing system and method thereof for grinding and / or crushing cereals and seeds. Measurement data for all relevant physical variables determining the milling process of mills at one or multiple geographical locations are measured and acquired in real time, compared with reference values, and used for autonomous operation adaption, system health check and automated repair of hardware by means of machine learning models. Machine learning output based on system and environmental measurements generates signals to steer steerable devices for maintaining product quality and to monitor and forecast abnormal system behavior in real-time to ensure uninterrupted production plant operation.

Description

[0001] P1480PC00

[0002] An intelligent sectorial delineated and hierarchical control system for controlling flour mills

[0003] Field of the Invention

[0004] The present invention relates to an intelligent sectorial delineated and hierarchical control system, for controlling and self-adaptation of industrial grinding and milling systems, in particular for grinding and / or crushing cereals and seeds and / or for processing, by comminution and homogenization of viscous material, in particular chocolate masses, printing inks and the like. In particular, the present invention comprises the autonomous and adaptive operation of the grinding process given specified characteristics of the input and output material and their relationship to the grinding process, such as the influence of environmental parameters (humidity, temperature, etc.) on the input material (size distribution, humidity, etc.) and optimal grinding parameters (roller speed, roller gap, etc.) for specified output material properties (size distribution, humidity, etc.) . More particular, the invention relates to autonomous, self-adaptive and real-time regulation of the entire milling installation, comprising mills with one or several milling lines, milling plants with several mills at one geographical location and milling plants comprising mills at different geographical locations. The invention further relates to the real-time technical supervision, control, and optimization of the milling installation, including but not limited to roller wear, roller temperature, roller gap, parallelism of roller pairs, roller speed, roller compression force and / or energy consumption of one or several roller drives. Within the present invention the sensor measurements are used to evaluate and update a self-learning machine learning model for optimizing the technical operation of the milling installation for output material consistent with predefined quality requirements, including, for example, milling fineness, residual starch content, water content, ash content (mineral materials) etc. Further to this, the measurement values are used for assessing the medium-to-long- term operability of the plant or installation for predicting imminent replacement of mechanical hardware, such as rollers or roller pairs due to wear.

[0005] Background of the Invention

[0006] Modern milling starts with the selection and quality control of the material to be milled. For uniformly high-quality milled products from organic material, such as grains, coffee, or seeds, a comprehensive physiochemical characterization of the material is required and includes, inter alia, humidity, size, and shape analysis. Technical realization of carrying out the necessary decisions involves sensors that determine physical measurement values which are used to influence the machine units of the mills including, but not limited to, amount and speed of input material, roller speed, or roller gap.

[0007] The milling process involves means for grinding, crushing, pulverizing, or reducing the size of input materials. For the special case of roller-based mills, the rollers are spatially separated by a gap which determines the fineness of the ground product. Subsequent crushing is carried out by continuously decreasing the spacing between the rollers so as to yield the appropriate degree of fineness of the product. Of paramount importance to the milling process is the precise measurement and control of temperature of the rollers and vibrations of the mechanical parts of the mills as these two variable affect the entire process. Temperature has a direct effect on the quality and physiochemical properties of the milling product whereas a change in vibration - in particular its increase - indicates mechanical instability or malfunction of the entire apparatus.

[0008] The term "sectorial delineated sectorial delineated" shall mean the milling system is divided into sectors (sections or functional areas) that are clearly defined and managed separately. In a flour milling system, sectors can be a supply line for the intake, cleaning the input materials, a processing line for milling and sifting the input material and an output line for blending and packing of an end product, for example. The term "Delineated" shall mean the boundaries and responsibilities of each sector are clearly specified, enabling targeted monitoring and control within each area. The term "hierarchical" shall mean that the system is organized in levels of authority and aggregation. Lower levels handle local, real-time control (e.g., machine-level PLCs for a plan sifter or a roller stand), mid-levels coordinate a whole line (e.g., a milling line), and higher levels handle the flour mill or milling system comprising several flour mills, optimization, reporting, and management (e.g., SCADA / MES / ERP integration) . Decisions and data flow up and down these layers. Traditionally, the miller checked the temperature of the rollers at regular intervals to ascertain that they have a uniform temperature by running his hands across the entire length of the rollers to check that the rollers have approximately the same temperature at the center and the outer regions. To ovoid temperatures to increase above thresholds that compromise the quality of the grinding product or lead to excessive material wear optical methods (WO 2014 / 195309) have been proposed. However, such an approach faces the problem of soiling because the sensing elements are located in the region through which the material to be ground flows. On the other hand, temperature sensors located on one or both sides of at least one roller have the disadvantage that the temperature at the center of the roller has already reached the threshold value but the sensors at the rim of the roller still record temperatures below threshold. This may lead to increased mechanical wear or deteriorating product quality before the process can be stopped automatically due to delayed communication of the relevant temperature threshold. Measurement of temperature in a non-contact manner (DE 1022641 1 Al ) is problematic because the temperature of the circumferential surface can deviate significantly from the measured temperature. Furthermore, the necessary calibration can depend on the miller or person operating the mill.

[0009] Monitoring temperature has been the subject of a prior invention (US 2019 / 240672 Al ) . According to the prior art, a device based on machine learning technology which is able to optimally and dynamically monitor and adapt the roller pairs without manual checking or setting various operation conditions, including temperature or roller spacing.

[0010] After milling, the ground material is checked for humidity and quality assessment is performed before the final product is packaged.

[0011] The measurement values and output of sensors at individual stages along the milling process (delivery, rollers, sifters...) constitutes the current state of the art. The present invention relates to the autonomous control, adjustment, optimization, and forecasting in real time of individual mills and assemblies of mills with the possibility to centrally manage and optimized integrated milling systems. This requires uniform and efficient data recording, analysis, management, and storing solutions. The data is recorded and used to train and update machine learning algorithms for verification of operability, product control, tools surveillance, and data-informed, automated, in-time, and forecasting of process plant supervision. Data recorded and prepared in this fashion can be used to trigger partial or full plant shutdown for safety reasons, to inform plant management of impending revision or tool replacement, to gather necessary data for product quality control, and to optimally configure process plant parameters for changing operation conditions. Operation conditions can vary due to environmental factors such as temperature and humidity. Such effects need to be included in sensor and threshold settings in order to guarantee safe operation and quality of the product.

[0012] The main driving force underlying tighter integration of process control, machinery surveillance, product quality assessment and tools logistics within "Industry 4.0" or the Internet of Things (loT) is a) to provide flexible and robust production solutions, b) to have the capability to forecast important abnormal and / or undesirable system events (system and / or machinery failure, required repair of machinery, degradation of starting material, quality decrease of product), and c) to determine the severity and potential impact of ab-normal behavior on the production plant and / or the product.

[0013] For determining the necessary scores to make informed decisions about individual tools, the entire production plant (one mill) or the totality of plants surveilled, numerical and classification data for each element supervised is required. As the data delivered from each component can differ, a first important decision is that of data format to be used. Because ultimately the decision-making progress is based on artificial intelligence and machine learning in particular, numerical data is preferred. For predictive prognosis which is the mode of operation required in the present context, supervised learning is at the forefront.

[0014] In a first step, the totality of numerical data needs to be collected which is referred to as "data fusion". From this point onward it is assumed that all data is numerical. This allows to map the values for each component onto a common interval by data standardization to ensure that the magnitude of a particular physical variable does not skew the final score in undue ways. Such "data preparation" is necessary in order to remove bias in the model and to work with curated data.

[0015] Next, this data is used to train a ML model which is capable to describe the totality of parameters of a single plant or a collection of plants. The preferred solution for predictive prognosis is supervised learning. Supervised learning uses a training set to teach the model to yield the desired output. This training dataset includes inputs and correct outputs, which allow the model to learn over time. The algorithm measures its accuracy through the loss function, adjusting until the error has been sufficiently minimized. Because one of the main reasons of the model is to forecast "abnormal behavior" it is essential that such behavior in all its flavors is part of the training set. Otherwise, there is highly probable that the trained model is unable to associate "abnormal behavior" with a specific machine component or a combination of triggers responsible for undesired behavior.

[0016] One of the challenges are implementation standards. This is particularly true for situations in which heterogeneous hardware is involved such as industrial loT sensors that coexist with legacy equipment.

[0017] Summary of the Invention

[0018] It is an objective of the present invention to provide a system and method for industrial plant operation and process control with machine-based adaptation of control parameters dependent on variable environmental and / or machinery conditions based on machine-learned systems control for optimized, autonomous plant operation using real-time information with the capacity for forecasting and predicting operability of the plant with enhanced operating safety of the plant. It is a further objective of the present invention to centrally and autonomously control and optimize multiple milling lines and / or mills at different geographical locations.

[0019] According to the present invention, these objects are achieved, particularly, by the features of the independent claims. In addition, further advantageous embodiments can be derived from the dependent claims and related descriptions.

[0020] According to the present invention these objectives of a flour milling system with several machine units comprising one or more roller stands and an intelligent sectorial delineated and hierarchical control system for controlling and / or steering the machine units by a control signal, wherein the milling system comprising one or more mills with at least one milling line for the industrial processing of a input material and a central data management unit connected to the one or more mills via a data transmission network, wherein a milling line comprises a supply line for preparing the input material, a processing line for processing the input material and an output line, wherein the supply, processing, and output line each comprises at least one machine unit for processing and / or dosing and / or weighting the input material processed by the milling line, wherein the input material are processed to an end product) by the at the last one milling line, wherein each milling line comprises one or more roller stands for grinding the input material and sensors for measuring sensing data of the corresponding milling line and / or one or more machine controls 23 for steering and / or controlling the operation of the machine units of the corresponding milling line by target parameters, wherein the machine control 23 comprises a control interface (i) for capturing and transmitting an operational data of the one or more machine units of the milling line and (ii) for setting target parameter values based in the control signal generated by the control system for steering and / or controlling the machine units, wherein each roller stand comprises two rollers turning with a corresponding roller speed and forming a roller gap with a roller gap width, a roller feeder for feeding the rollers via the roller gap with a feeding rate for grinding input material and the corresponding machine control for controlling and / or steering the roller speed and / or the roller gap width and / or the feeder speed of the roller feeder of the roller stand by the corresponding machine control based on the target parameter, wherein the target parameter of the one or more roller stands comprises a roller stand target parameter comprising a target roller gap width and / or a target roll temperature and / or a target roller speed for each roll and / or a target roller feeder speed and / or a target feeder gap for steering and / or controlling the roller stand via the control interface, wherein the control system comprises for each mill a mill control unit comprising a data interface and one or more steering units (i) for steering and / or control the machine units by sending the control signal to control interface and / or (ii) for receiving operational parameters of the machine units, and wherein the one or more steering units being connected to the control units via the data interface, in that at least one steering unit comprise a gap control module for steering the roller gap width and / or a feeding control module for steering the feeder roll speed and / or a slip control module for steering the ratio of the roller speed of the roller of the roller stand, in that the mill control unit comprises one or more line controls each with (i) a data acquisition unit for receiving and / or preprocessing milling parameter values and / or operational parameter values of the machine units and / or a line parameter values of the milling line, (ii) a line intelligence for generating machine setpoint parameter values based on the milling parameter and / or operational parameter and / or line parameter values, and a line signal generator for sending the setpoint parameter values to the steering units of the milling line, in that the one or more steering units each comprising a data acquisition unit for receiving setpoint parameter values and / or pre-processing the milling parameters and / or operational parameters of the machine units of the milling line, a machine intelligence unit for generating target parameter values based on the milling parameters and / or operational parameters, an a signal generation unit for generating the control signal based on the target parameter values, in that the signal generation unit generates an operation control signal for controlling and steering the machine units by transmitting the operation control signal via the control interface to the machine control. The invention has, inter alia, the advantage of providing autonomous, self- adaptive operation based on-real time data. It is to be noted, that realizing line intelligence using machine learning to control roller gaps in an industrial mill is a high- impact issue in smart manufacturing. The technical object is to dynamically adjust the roller gaps based on real-time data to maintain product quality, reduce waste, and increase efficiency. The present invention has, inter alia, the advantage to allow to automate and optimize the control of roller gaps to ensure product dimensions (thickness / width) and surface quality. In particular, the invention allows to automatically maintain consistent material thickness, react to material property variations (hardness, temperature), and predict and adjust gaps in real-time to reduce manual intervention. More particularly, the system allows to automatically and intelligently adjust the roller gaps in a flour milling line to optimize flour quality, yield, and energy efficiency, while minimizing operator intervention and material waste. It is to be noted that in flour milling, roller mills are used to grind cleaned and tempered wheat kernels into flour. The roller gap (i.e. the distance between the rollers) is a critical setting: Too tight — ► excessive starch damage, heat, and bran contamination. Too loose — ► poor grinding, low flour extraction rate. Roller gap must be adjusted depending on at least the parameters wheat type (hard / soft), moisture content, roll wear, product specification (fine flour vs semolina, etc.) . In the prior art, many mills rely on manual trial-and-error adjustments or periodic checks. The present invention introduces a machine-learning- based system that dynamically sets and fine-tunes roller gaps based on real-time grain input properties and process feedback. Key input parameters for the present system can e.g. comprise grain properties (e.g. wheat variety, moisture content, hardness index) measured e.g. by NIR sensor, or inline grain analyzer, environmental parameters (e.g. temperature, humidity) measured e.g. by factory sensors, machine data parameters (e.g. roller type, speed, vibration, temperature) measured e.g. by roller sensors, process feedback parameters (e.g. ash content, flour granulation, extraction rate) measured e.g. by inline flour quality sensors, and current settings parameters (e.g. roller gap, feed rate, roll pressure) provided e.g. by PLC / HMI. The control data (output signaling) can at least comprise roller gap adjustments parameter (setpoint changes). For the annotation of the data, e.g. final product defect logs, and / or quality grades, and / or rework reports (optional for supervised learning) can be used. It may be useful to start with historical logs if available and use them to prototype models before deploying real-time learning. To choose the appropriate machine learning approach for the machine intelligence, there are different options for the present inventive system to realize an efficient machine intelligence, as e.g. using a supervised regression model structure, a reinforcement learning (RL) structure, or a hybrid model structure. Selecting a supervised regression model, the optimal roller gap can be precited from input variables including input parameters (Entry thickness, material type, temperature, etc.), target parameter (exit thickness or desired gap) and model selection parameters (Linear regression, Random Forest, Gradient Boosting, or Deep Neural Networks). Selecting a reinforcement learning (RL) structure, allows a controlled learning of control policy over time to optimize for quality. The parameters can e.g. comprise state parameters (Current gap, material parameters etc.), Action parameters (Adjust gap ±x mm), and / or reward parameters (Higher if final thickness is in-spec). This embodiment variant can e.g. be suitable for systems with an associated simulation or digital twin structure. Finally, selecting a hybrid model structure, regression can be used for prediction and rule-based or optimization layer for control. Training and validating the model structure can be achieved by preprocessing the data by (i) normalize / scale the numerical measuring data, (ii) one-hot encode material types, and (iii) handle missing data or outliers. Further by splitting the data in train / validation / test splits, and cross- validation for robustness. And finally, by the appropriate metrics comprising e.g. MAE or RMSE for regression, % of in-spec output, and runtime latency (for real-time suitability) . The present invention provides a real-time integration with PLC / SCADA. Deployment As architecture (i) Edge Device or IPC: Hosts the model and interacts with sensors, (ii) Model Inference API: Takes sensor inputs, returns gap setpoint, and PLC / SCADA: Receives setpoint and applies actuator changes, can e.g. be deployed. It may be important to ensure real-time response (ms to sec), fail-safe mechanisms if model fails, and / or manual override options. The inventive system further provides a feedback loop and model updating allowing log real-time predictions and outcomes, allows to monitor model drift or underperformance, and / or allows to retrain with new data periodically (online learning or scheduled retraining) . Additional enhancements can e.g. comprise a digital twin simulation (Simulate roller behavior to test models), an anomaly detection (Alert if system behaves unexpectedly), and / or an explainability Tools (Use SHAP or LIME to explain model decisions for operators).

[0021] In a use case, a supervised regression model structure can e.g. be realized to predict the optimal roller gap for given input wheat and desired output flour quality. Exemplary input parameters can e.g. comprise moisture = 15.5%, wheat type = Hard Red Winter, target flour fineness = 70 pm. Exemplary output parameters for the appropriate output signaling can then e.g. adjusting gap setting = 250 pm. The supervised regression model structure can e.g. be trained on historical data where wheat properties and final product quality are known. As another real-world example, a milling of 100% Hard Red Spring Wheat can e.g. be taken with target flour type = patent flour. The input parameters can then e.g. be moisture: 15.3%, kernel hardness: 82, target granulation: 60 pm, and environment temp: 28°C. The output signaling would be an automated adjustment of the Bl Roller gap = 280 pm and B2 Roller gap = 180 pm. The system applies it automatically e.g. with inline granulation sensor confirming within tolerance. An operator has only to supervise, no manual tweaks needed.

[0022] Intelligent, sector-delimited and hierarchical control systems for controlling mills ensure improved scalability and modularity, so that additional milling lines or mills can be added sector by sector with minimal re-engineering and standardized interfaces between the levels (PLC-SCADA-MES-ERP) can be used for the control system. A further effect is the improved coordination of mill operation and optimization of the mills (shift planning, recipe management) as well as orchestration at fleet level (cross-plant load distribution, benchmarking) and central visibility with local autonomy for fail-safety. In another embodiment variant the operation parameter comprise a roller operation parameter comprising the roller gap width and / or the roller temperature of the rollers and / or a temperature distribution of the rollers and / or a roller stand status of the roller stand and / or a feeding speed and / or a feeding gap width, so that the flour mill achieves a balance between high-quality flour production and effective, cost effective operation of the flour milling system. The roller gap width determines the fineness of the grind. Furthermore, the proper adjustment of the roller gap width ensures an effective separation of a bran from an endosperm of a wheat kernel of the input material. Furthermore, the roller gap influences the rate of the wear on the rollers. A finer roller gap width can cause significant wear on the rollers due to the increased friction and pressure. The difference in speed of the two rollers of a roller stand influences the separation of the endosperm from bran and impacts the extraction rate and flour purity. The feed rate characterizes the amount of input material fed into the roller gap. An optimal feed rate balances the load of the roller stand and prevents clogging.

[0023] In a further embodiment variant the milling parameters comprise a plant parameter for each of the one or more mills, comprising a mill configuration and / or an product recipe, wherein the mill configuration comprises machine units and / or mass flow parameter in relation to the operational parameter of the operating units and / or status of the machine units and / or mass flow relations between the machine units and wherein product recipe comprises milling product specifications and / or end product specifications and / or initial mill configuration, so that inter alia, during the milling process of an operational process recipe the operational control parameter are monitored continuously in real time by means of the data acquisition and central milling control system, steering units and line control(s), wherein a definable anomalous fluctuation in a parameter value is detected as a defined deviation of the monitored operational parameters and the operational control parameters steering the machine units of the milling line and / or mill are adapted autonomously and In real time by means of the signal generation unit.

[0024] In yet another embodiment variant the central data management unit compromises a repository, wherein the repository comprises a configuration library and / or recipe library and / or end product library, wherein the configuration library comprises at least one mill configuration and wherein the recipe library comprises at least one or more product recipes, so that for the use of the inventive milling system for each product recipe of the recipe library an effective mill configuration is ensured. Furthermore, an advantage of this embodiment is, that each mil and / or milling line is steered and optimized independently of all other mills and / or milling lines, including mills and / or milling lines at different geographical locations.

[0025] In another embodiment variant the control system of at least one of the mills comprises mass flow unit with a mass flow parameter, a mass flow interface for sending the mass flow parameter of the at least one of the mills to the corresponding one or more line controls, and a mass flow module for generating and / or measuring the mass flow parameter based on the product parameter values and / or operational parameter values and / or plant parameter of the at least one of the mills, so that, inter alia, the milling process of the milling lines and / or mills is optimized automatically by the central milling control and / or the line control(s) based on the load across the different machine units of a milling line or a mill. Furthermore, the mass flow parameter values ensure that each machine unit operates at its optimal capacity, and the product specification / quality specification are ensured. The milling control unit monitors based on the mass flow parameters so that, inter alia, unusual mass flow patter are detected and a signal triggered to reduce downtime of a mill and / or milling line and extend the lifespan of the machine units of the mill. Monitoring and / or controlling the milling lines and / or machine units based on the operational mass flow parameter, preferably in real time, ensures greater overall efficiency, quality and profitability as well as compliance with industry standards.

[0026] In a further embodiment variant the data acquisition unit captures and preprocesses measurement data by standardizing, normalizing and filtering / noise- reducing captured data, the output of the acquisition unit being delivered to the line intelligence unit, so that, inter alia, the prediction of machine learning unit are stable and reliable with respect to small variations in the input variable, i.e. are not subject to numerical imprecisions and noise.

[0027] In another embodiment variant the line intelligence unit comprises a machine learning and / or artificial intelligence unit transforming input from the data acquisition unit to setpoints parameter suitable for line signal generator to be communicated to the steering units of the milling lines, so that the control system of the mill providing autonomous, self-adaptive operation based a large volume of complex on-real time data and operational settings, for example operative mill configuration and / or operative product recipe and / or operational parameters of the various machine units. A further advantages of this embodiment is, inter alia, that the controllable and recordable parameters of the machine units comprising but not limited to the roller stand a) the adjustment of roller speed, inspection of input material of the supply line and process line, b) the temperature, pressure, forces, force components in several directions, wear, deformation, vibrational, rotational speed, rotational acceleration, parallelism of the at least two rolls of a roller pair, or the one or multiple rollers, can be measured and autonomously adjusted in real time if needed when compared with predefined tolerance of the optimum operational / trigger values. A further advantage is that forecasting of expected changes in the machine units are possible. For example, time series analysis of observables for the roller stands for example, such as temperature, rotational speed, surface roughness, etc. of the one or two rollers of a roller pair based on autoregressive models, autoregressive integrated moving averages, or temporal fusion transformer models, inter alia, provide information about future expected deterioration that can be anticipated. In such a continuous mode-of-operation, roller replacements due to reduced surface roughness can be anticipated instead of being suddenly and unexpectedly triggered by reaching a threshold value that leads to an abrupt stop of a milling line.

[0028] In a further embodiment variant the machine Intelligence unit comprises a threshold and / or trigger unit configured to trigger the warning / alarm signal unit in case of exceeding of a measuring signal to a defined triggering / threshold value, so that, inter alia, that the machine learning unit can be used to forecast reaching a critical operational / trigger value at time tl when comparing sensor measurement values at time t < tl . This allows to either adjust the parameters controlling the machine units of the milling line and / or mills such as to avoid exceeding trigger values or to avoid instantaneous shutdown, i.e. allow gradual shutdown.

[0029] In yet another embodiment variant the machine learning unit comprises a system-wide loss function unit, the loss function trained with supervised and / or unsupervised learning, so that, inter alia, the milling system comprising one or multiple milling lines and / or mills is surveilled as a whole, in real time, and optimum operational parameters are chosen subject to real-time operational and environmental measurement values to ensure and maintain optimal, predefined output material quality.

[0030] In another embodiment variant wherein supervised machine learning is based on labelled data transmitted from each sensor [2211 ...223N] and / or operational parameter and / or milling parameter for generating the parameters for steering the input part , the processing part, and the output part, in that humidity measurements of the material in the input part dynamically adjust roller pair rotation speed and roller pair separation and / or unsupervised learning based on unlabeled data from each sensor and / or measuring device in that classification of the quality of the output material either leads to recommissioning to the input line or transport to disposal or further processing of the ground product, so that, inter alia, that the physical parameters of the machine units, including but not limited to roller stands including rollers, roller pairs, valves, etc. are continuously monitored in real-time to provide measurement values that can be used to forecast expected parameter values at a future time that are outside the predefined standard range. This information is used by the machine learning unit to predict new system parameter values of machine units, including but not limited to vibrational, rotational speed, rotational acceleration, or parallelism of the at least two rolls of a roller pair, to avoid parameter values outside the predefined standard range and thus issuing of a stop signal due to triggering an alarm as a consequence of a measurement value being outside the allowed range.

[0031] In a further embodiment variant the generated parameters by the machine learning unit comprising monitoring circuitry with a defined monitoring process, at least comprising the steps of: (i) comparing actual parameters with predetermined parameters for each sensor for ahead-of-time alarm generation if actual parameters are predicted to fall outside the interval covered by predetermined parameters, (ii) determining optimal operational parameters including input line supply belt speed, roller pair speed, and roller pair separation by minimizing the loss function of the machine learning model, (iii) predicting servicing times for rollers due to wear and autonomously trigger ordering of replacement rollers and planning repair to minimize or fully eliminate downtime of one or several milling lines, (iv) linking multiple mills to predict local quality parameters including humidity, size, and uniformity of the input material from external parameters including growth and harvesting conditions of the input material, so that, inter alia, that the output material from one or several mills has a uniform quality independent on the geographical location at which the one or several milling lines and / or mills are situated.

[0032] In yet another embodiment variant the steering unit of each mill is connected with the central data management unit through the data transmission network, the central data management unit (5) signaling automated service management procedures and / or alarm devices and / or geographic or typographic surveillance processes, so that, inter alia, that each mill and / or milling line of a milling installation comprised of mills and / or milling lines at different geographical sites is independently steerable depending on local measurement values of environmental sensors transmitting measurement values on, inter alia, local temperature, humidity, or pressure, directly influencing the properties and quality of the input and output material, in particular ensuring uniform output product quality based on the value of the system wide loss function. A further advantage of this embodiment is that collective parameter changes and collective software updates over any number of mills and / or milling lines can be carried out.

[0033] In another embodiment variant the one or more mills and / or milling system comprises environmental sensor [241 ...24N] for measuring environment data and / or contextual parameters. An advantage of this embodiment is, inter alia, the environment-adapted, autonomous, and real-time control and optimization of a milling line and / or mills of a milling system at multiple geographic locations.

[0034] In a further embodiment variant wherein the machine learning unit uses generated parameters for predicting parameters for optimized operation control signal of the milling system given environmental parameters and / or contextual parameters (242) including environmental temperature and / or humidity, and / or humidity of the material to be ground. An advantage of this embodiment is, inter alia, the environment-adapted, autonomous, and real-time control and optimization of a milling line and / or mills of a milling system at multiple geographic locations.

[0035] In an embodiment variant the milling system comprising the generated parameters by the machine learning unit comprising monitoring circuitry to compare actual parameters with predetermined parameters for each sensor for ahead-of-time alarm generation if actual parameters are predicted to fall outside the interval covered by predetermined parameters, to determine optimal operational parameters including input line supply belt speed, roller pair speed, and roller pair separation by minimizing the loss function of the machine learning model, to predict servicing times for rollers due to wear and autonomously trigger ordering of replacement rollers and planning repair to minimize or fully eliminate downtime of one or several milling lines, linking multiple mills to predict local quality parameters including humidity, size and uniformity of the input material from external parameters including growth and harvesting conditions of the input material. An advantage of this embodiment is, inter alia, that the output material from one or several mills has a uniform quality independent on the geographical location at which the one or several milling lines and / or mills are situated.

[0036] Brief Description of the Drawings

[0037] The present invention will be explained in more detail by way of example in reference to the drawings in which:

[0038] Figure 1 shows a diagram illustrating schematically an embodiment variant of the invention in which a mill 1 comprising a production line 3 and a milling controlling system 4 is used for milling material by an intelligent, data-driven process.

[0039] Figure 2 and 3 shows a block diagram illustrating schematically an exemplary autonomous adaptation of the milling process by means of the steering units 43 comprising the data acquisition unit 412, the line intelligence 414 and the signaling unit 416 controlling the mill 2 comprising machine units 45

[0040] Figures 4 shows a block diagram illustrating schematically an exemplary data acquisition unit 412 comprising sensors [221 ...22N], connected to a PLC 4121 and followed by a Control Client / Interpreter 232. Data is preprocessed comprising at least filtering and / or standardization and / or homogenization (data format) and sending the data to the line intelligence 414 through a network interface 4124.

[0041] Figure 5 shows a block diagram illustrating schematically an exemplary milling control system 4.

[0042] Figure 6 shows a flow chart of an exemplary milling line 3 comprising machine units 45, 451 a, 451 b, 451 c, 451 d, 452a, 452b, 453a, 453b, 453c and 453d and a mass flow Bl , Bl-1 , Bl-2, SB1 , SBI Br, SB1C1, SB1 F1 , Cl , Cl-l ,Cl-2, SCI , SCI Br, SC1 F2, SCI Fl , FF1 , F2 and Br generated by the machine units 45, 451 a, 451 b, 451 c, 451 d, 452a, 452b, 453a, 453b, 453c and 453d. Figure 7a, 7b and 7c shows a separation matric ASeP, a collection matric Acoi and a target vector d, and a detailed mass flow x of the milling line 3 showed in Figure 6.

[0043] Figures 8 and 9 show an exemplary realization of an embodiment variant of the inventive grinding gap optimizer system per mills 2 and milling line 3, respectively, (see figure 9), as well as per machine unit 45 (figure 8). As such, figure 8 defines also a machine unit 45 part of figure 9.

[0044] Figure 10 shows exemplarily an operational level 1 of the inventive system's operation and operation providing operational transparency by monitoring, collecting, structuring and presenting data visually is the first level of the inventive system. Connectivity and data readiness are one of the technical bases of the inventive system 1. Once measuring and production data is digitalized, the inventive system 1 provides technical transparency. Further, it is also enabled to provide a historic record of production parameters that can, inter alia, be fed into an Internet of Things gateway for further data processing and analysis to help optimize operation and production of the mill 2. The inventive connectivity and data readiness is the foundation of the inventive system 1 , unlocking exponential benefits in further operation levels 2, 3, and 4, as described below.

[0045] Figure 1 1 shows exemplarily an operational level 2 of the inventive system's operation and automation, the operational level 2 involving applying algorithms and signal generation structures to analyze and monitor the collected data. The technical ability to provide actionable production and steering signaling with intelligent steering based on empirical structures and measuring data is the fundament of the inventive automation, rather than instinct and experience required of human operator in prior art systems. The inventive system gathers and assembles data in a comprehensive monitoring structure and analyses and as more historical production data is stored digitally different yield and quality outcomes could be compared depending on production parameters. This allows to automatedly select the parameters that need to be adjusted to best optimize efficiency and quality, depending on variables such as a raw material characteristics or differing energy efficiencies. The operational level 2 also enables the monitoring of each machine's performance throughout the production process by the system 1 and the controlling how machine parameters are interrelated. By automatedly detecting key performance indicators for each machine and process, the inventive data processing can automatically analyze performance trends and then autonomously optimize energy usage, maintenance scheduling and how best to optimize machine performance to achieve the highest quality and most efficient endproduct. For example, an error and downtime analysis is a data processing module with appropriate data processing structures that interprets and records machine incidents that cause production losses and recognizes patterns and trends. The energy management system monitors thereby energy consumption with every part of the process monitored and inefficiencies flagged up.

[0046] Figure 12 shows exemplarily a possible human interface realization indicating the drastic reduction of the level of human intervention through the inventive individual (i.e. machine related) real-time, self-optimizing and autonomous production processes. As such, human intervention is, on an operational level 3 and 4, either reduced (assist level of the system's operation) or completely overcome autonomous, self-optimizing level of the system's operation. In particular, on operational level 3 human intervention is reduced due to the system's ability to self-optimize some of its key production processes, while on the operational level 4, the system requires only minimal human intervention reduced to and apart from setting the product characteristics required. The milling system 1 autonomously and automatically optimizes quality while running at the greatest possible efficiency, which is technically a multi-dimensional, non-trivial optimization problem. Through intelligent learning, the milling system 1 continuously and dynamically adapt its used production parameter values based on measured variables like raw material characteristics or weather conditions etc.

[0047] Detailed Description of the Invention

[0048] Figure 1 illustrates, schematically, an architecture for a possible implementation of an embodiment of the inventive flour milling system 1 with several machine units 45, 451 , 451 a, 451 b, 451 c, 451 d, 452, 452a, 452b, 453, 453a, 453b, 453c, 453d comprising one or more roller stands 451 , 451 a, 451 b, 451 c, 451 d and an intelligent sectorial delineated and hierarchical control system 4 for controlling and / or steering the machine units 45, 451 , 451 a, 451 b, 451 c, 451 d, 452, 452a, 452b, 453, 453a, 453b, 453c, 453d by a control signal 434. The milling system 1 comprising one or more mills 2, [201 20N] with at least one milling line 3 for the industrial processing of an input material 31 1 and a central data management unit (CDMU) 5 connected to the one or more mills 2, [201 20N] via a data transmission network 47,471 . The mills 2, [201 20N] of the milling system 1 can be situated at a milling plant at a geographical location 70A and / or at different geographical locations 70A to 70N. A geographical location can e.g. be characterized by their geographical latitude and longitude. A geographical location 70A..70N can be characterized in that environmental data 241 comprising temperature and / or humidity and / or air pressure and / or production conditions comprising stability of power supply and / or availability of spare parts and / or availability of trained personnel for maintenance and repair are identical.

[0049] The one or more mills 2, [201 ...20N], as used herein, comprises the machine units 45, 451 , 451 a, 451 b, 451 c, 451 d, 452, 452a, 452b, 453, 453a, 453b, 453c, 453d for cleaning, conditioning, sifting, grinding, crushing, pulverizing or reducing the size of input materials 31 1 to an end product 331 with defined properties. The mill 2, [201 ..20N] comprises one or more milling lines 3 and one or more mill controls 49.

[0050] The mill 2, [201 ..20N] comprises sensors 22 including sensors for the milling product 221 1 ..221 N and / or input material 31 1 , sensors for the machine units [2221 ..222N], sensors for the final product [2231 ..223N] and sensors for the environmental data [2251 ..225N] for sensing environmental data 241 and / or contextual parameters. Furthermore, the mill 2, [201 ...20N] comprises milling parameters 21 with product parameters 21 1 and mill parameters 212. The product parameter 21 1 captures measured properties of the end products 331 and / or input material 31 1 properties including final product properties such as starch damage, ash content, moisture content and protein content.

[0051] A milling line 3 comprises a supply line 31 for preparing the input material 31 1 , a processing line 32 for processing the input material 31 1 and an output line 33, wherein the supply, processing, and output line 31 , 32, 33 each comprises at least one of the machine unit 45, 451 , 451 a, 451 b, 451 c, 451 d, 452, 452a, 452b, 453, 453a, 453b, 453c, 453d for processing and / or dosing and / or weighting the input material 31 1 processed by the milling line 3, wherein the input material 31 1 is processed to an end product 331 and a waste 332 and a return supply 333 by the at the last one milling line 3, wherein the return supply 333 is feed back to the supply line 31 . Each of the milling lines 3 comprises one or more roller stands 451 , 451 a, 451 b, 451 c, 451 d for grinding the input material 31 1 and sensors 22, [221 1 ..221 N], [2221 ..222N], [2231 ..223N], [2251 ..225N] for measuring sensing data 224 of the corresponding milling line 3 and / or one or more machine controls 23 for steering and / or controlling the operation of the machine units 45, 451 , 451 a, 451 b, 451 c, 451 d, 452, 452a, 452b, 453, 453a, 453b, 453c, 453d of the corresponding milling line 3 by target parameters 233.

[0052] Each of the processing line 32 of a milling line 3 comprises one or more passages wherein a "passage" refers to the flow of the input material 31 1 through different stages of the processing line 32. Each passage typically corresponds to a specific stage in the processing line 32, encompassing various processes such as grinding, sifting, and purifying. The one or more passages comprise a break passages for breaking the input material 31 1 into smaller pieces while separating kernels of the input material 31 1 in bran, germ, and endosperm including roller stands and / or a reduction passages for reducing the particle size by grinding using a roller stand 451 , 451 a, 451 b, 451 c, 451 d for achieving finer flour while ensuring that unwanted byproducts (like bran and germ) are separated out by using roller stands 451 , 451 a, 451 b, 451 c, 451 d and sifters 452a, 452b and / or one or more sifting passages for sifting separate different flour grades or fractions based on particle size including multiple stages of sifting to ensure that the flour meets the desired quality and consistency and / or one or more purification passages for an additional separation and removal of the bran and any remaining coarse particles from the flour. The structuring of the processing line 32 in one more passage ensures the optimization of the processing line 32 and achieving the desired end product 331 specifications. Each passage is determining the quality, texture, and functionality of the end product 331 . Each passage is identified by a passage information in the present invention.

[0053] The central data management unit 5 (CDMU) comprises a repository 51 including a configuration library 51 1 with one or more mill configurations and / or a recipe library 512 with one or more product recipes and / or a machine unit library 514 with one or more specifications of the machine units 45, 451 , 451 a, 451 b, 451 c, 451 d, 452, 452a, 452b, 453, 453a, 453b, 453c, 453d and / or an end product library 513 with one or more end products specifications. The mill configuration capturing machine units 45, 451 a, 451 b, 451 c, 451 d, 452a, 452b, 453a, 453b, 453c, 453d of the mill 2 and / or initial setpoint parameter values 415 of the setpoints parameter 415 and / or a starting procedure for all the machine units 45, 451 a, 451 b, 451 c, 451 d, 452a, 452b, 453a, 453b, 453c, 453d and / or one or more machine learning structure and / or at least one mass flow configuration matric ASeP, , Acoi for defining the mass flows between the machine units 45, 451 a, 451 b, 451 c, 451 d, 452a, 452b, 453a, 453b, 453c, 453d and / or a weight of the input material 31 1 and / or a weight of the end product 331 and / or weight of the waste 332.

[0054] For the present invention, the mill control 49 comprises one or more line controls 41 each with (i) a data acquisition unit 412 for receiving and / or pre-processing milling parameter values 21 and / or operational parameter values 232 of the machine units 45, 451 , 451 a, 451 b, 451 c, 451 d, 452, 452a, 452b, 453, 453a, 453b, 453c, 453d and / or a line parameter values 413 of the milling line 3, (ii) a line intelligence 414 for generating the machine setpoint parameter values 415 based on the milling parameter 21 and / or the operational parameter 232 and / or the line parameter values 413, and a line signal generator 416 for sending the setpoint parameter values 415 to steering units 43 of the milling line 3 and / or a data transmission network 472 for connecting the line control 41 with the steering units 43 for data signaling for example. Furthermore, the mill control 49 comprises a mass flow unit 48 for generating a detailed mass flow vector x for the mill 2.

[0055] In a further embodiment of the present invention the line intelligence 414 comprises a machine learning and / or artificial intelligence unit 4141 transforming input from the data acquisition unit 412 to the setpoints parameter 415 suitable for a line signal generator 416 to be communicated to the steering units 43 of the milling lines 3. The machine Intelligence unit 4141 comprises a threshold and / or trigger unit 417 configured to trigger the warning / alarm signal unit 46 in case of exceeding of a measuring signal to a defined triggering / threshold value as illustrated in Figure 3.

[0056] In the embodiment of the present invention the data acquisition unit 412 comprises a data interface 4124 for receiving the sensing data 224 measured by the sensors 22, [221 1 ..221 N], [2221 ..222N], [2231 ..223N], [2251..225N] and / or operational parameter of the machine unites 45, 451 a, 451 b, 451 c, 451 d, 452a, 452b, 453a, 453b, 453c, 453d, a PLC 4121 , a data preprocessing 4122 and a data homogenization 4123 for processing data received by the data interface 4124 to output data 4125 processable for the mill control unit 49 and / or the mass flow unit 48 and in particular for the line intelligence 414. In the inventive method the line control unit 41 can be e.g. connected to the machine control 23 and / or sensors 22 [221 ..22N], [221 1 ..221 N], [2221 ..222N], [2231 ..223N], [2251 ..225N] via a data interface 4124. The data acquisition unit 412 can e.g. be connected to the sensors 22 [221 ...22N], [221 1 ...221 N], [2221...222N], [2231 ...223N], [2251 ...225N] via PLC 4121 as illustrated in Fig. 4. Each measured sensing data 224 of the sensors 22 [221 ...22N], [221 1 ..221 N], [2221 ..222N], [2231 ..223N], [2251..225N] can be preprocessed by the data preprocessing 4122 before data homogenization 424 and transmitted to the line intelligence 414 and / or the mass flow unit 48. Data preprocessing can include, e.g., assessment of data quality, threshold detection, filtering, data format conversion, or unit conversion. After data preprocessing, data to be transmitted to the line intelligence 414 can be homogenized and standardized for a use in the machine learning unit 4141 .

[0057] The product recipe captures an end product specification and / or a specification of the input material 31 1 and / or one of the one or more mill configuration for producing the end product 331 specified in the end product specification and / or an evaluation function for defining the objectives. The evaluation function comprises one or more success criteria and a weighting value for each success criteria. The evaluation function is used in the field of machine learning for reinforcement learning for example. Success criteria comprise yield and / or energy consumption per produced end product 331 and / or down time of the mill 2 and / or down time of the milling line 3.

[0058] The machine unit specification capturing the following information of the respective machine unit: type of the machine unit 45, 451 a, 451 b, 451 c, 451 d, 452a, 452b, 453a, 453b, 453c, 453d and / or a passage information and / or a mass flow function.

[0059] Furthermore, the mill control 49 comprises an operation configuration parameter for defining the initial configuration of the machine units 45, 451 a, 451 b, 451 c, 451 d, 452a, 452b, 453a, 453b, 453c, 453d and / or an operation recipe for specifying the input material 31 1 and / or the product material 331 and / or an initial mass flow matric. The operative configuration parameter capturing a type of the machine units 45, 451 a, 451 b, 451 c, 451 d, 452a, 452b, 453a, 453b, 453c, 453d and / or passage information of the operative machine units 45, 451 a, 451 b, 451 c, 451 d, 452a, 452b, 453a, 453b, 453c, 453d and / or mass flow logic of the operative machine units 45, 451 a, 451 b, 451 c, 451 d, 452a, 452b, 453a, 453b, 453c, 453d and / or energy consumption.

[0060] In an another embodiment of the present invention the mill control 49 comprises mill optimizer for controlling and optimizing the use of the milling lines 3 based on the milling parameter values 21 and / or operational parameter values 232 of the machine units 45, 451 a, 451 b, 451 c, 451 d, 452a, 452b, 453a, 453b, 453c, 453d and / or a line parameter values 413 of the milling lines 3.

[0061] Within the present invention, a "control system" can e.g. comprise a distributed control system (DCS) which can be realized as a digital process control system (PCS) for a process or plant, wherein controller functions and field connection modules are distributed through-out the system. As the number of control loops grows, DCS can be more cost effective than discrete controllers. Additionally, a DCS provides supervisory viewing and management over large industrial processes. In a DCS, a hierarchy of controllers can be connected by communication networks, allowing centralized control rooms and local on-plant monitoring and control which can communicate using industry-standard protocols. Networking allows the use of local or remote SCADA operator interfaces and enables the cascading and inter-locking of controllers. However, as the number of control loops increases for a system design there can be a point where the use of a programmable logic controller (PLC) or distributed control system (DCS) can be more manageable or cost-effective.

[0062] "Sectorial delineated and hierarchical" within the present invention shall be understood as the hierarchical structure defined by the mill system 1 , the one or more mill 2, the milling

[0063] "Intelligent" within the present invention shall be understood to include but not be limited to employ technology-driven processes to monitor, control, adapt, modify, and optimize production processes, characterized by high levels of adaptability, rapid design changes, and the use of digital information technology. In other words, "smart" manufacturing can e.g. comprise technology-driven procedures based on a large number of sensor measurement data suitable but not limited to machine-driven adaptation going beyond rule- and / or trigger-based approaches to monitor industrial plants. "Smart" technology not only monitors and surveils industrial plants but can, e.g., also optimize, control in real time, devices and adapt processing conditions according to environmental conditions, such as humidify input mate-rial on the supply line 31 depending on measurements of environmental humidity and / or temperature. Adaptability as used herein comprises machine-induced adjustments of industrial processes to variable external conditions (weather, road conditions, etc.) and variable processing conditions comprising the input material to be processed, humidity (which determines optimal milling conditions to be used, including roller speed, roller gap, etc.) at run-time. Key technologies of "smart manufacturing" comprise big data processing capabilities, industrial connectivity devices and services, autonomous plant operation, autonomous supervision of manufacturing tolls, or autonomous management of supply chains.

[0064] The input material 31 1 can be or comprise bulk material or a mass. Bulk material comprises powdery, granular or pellet-shaped products which are used in the bulk goods processing industry including grains, milled grain products and grain end products, in particular soft wheat, durum, rye, corn, and / or barley or husks and / or milling of soy, buckwheat, spelt, pseudo-cereals or legumes, the production of feed for livestock and pets, fish and crustaceans, the processing of oil seeds, the processing of biomass and production of energy pellets, industrial malting and crushing systems, the processing of beans (cacao, coffee), nuts, and the production of fertilizers used in the pharmaceutical industry or in the fine chemical industry, together with stones, coal, or ore. Milling product can be powdery, granular, pellet-formed, or solid-state material as used in the bulk good processing industry, i.e. in the processing of cereal end products (e.g. wheat, durum wheat, rye, maize, and / or barley) or special milling industry (milling of soy, buckwheat, barley, pseudo-cereals and / or vegetables), the processing of oil seeds, fish, crustaceans, milling of cocoa beans, nuts, or coffee beans, the manufacturing of fertilizers, the pharmaceutical industry or the chemistry of solids, of the processing of biomass and manufacturing of energy pellet, of solid comprising stone or coal. By "processing of a product" in the sense of the present invention particularly the following is understood herein: (i) milling, crushing, shredding and / or flaking of bulk goods, particularly cereals, cereal milling products and cereal end products of the milling industry or special milling industry with pairs of milling rolls or flaking rolls; (ii) refining of masses, particularly food masses such as chocolate or sugar masses using, e.g., pairs of fine rolls; (iii) wet milling, particularly of printing inks, coatings, electronic materials, chemicals in particular fine chemicals; (iv) milling of solid fuels such as for coal / biomass power plants to reduce particle size to improve combustion efficiency and stability; (v) milling and / or crushing of stone, such as limestone, to exact particle sizes.

[0065] "Mass flow" refers to the movement of input material 31 1 / bulk materials, tons per hour for example, through the mill 2, 201 ..20N, encompassing the mass of flour and other components as they pass through various passages of the milling line 3. This concept is crucial for optimizing the milling process, as it ensures that the input material 31 1 , such as wheat, are efficiently transformed into the end product 331 , flour for example. The flour milling process involves several passages, including cleaning, conditioning, grinding, and sifting, each contributing to the overall mass flow. Maintaining a consistent mass flow ensures in achieving uniform quality of the end product 331 , minimizing waste 332, and ensuring operational efficiency. In the mill 2, the measurement of mass flow can be done using various methods such as weigh conveyors or flow meters. These machine units 453, 453a, 453b help monitoring and controlling the amount of input 31 1 material entering and leaving each passage of the milling line 3. Analyzing mass flow data ensures effective target parameters 233, roller parameter for example. Furthermore, effective mass flow management allows for control and / or steering of the machine units 45, 451 a, 451 b, 451 c, 451 d, 452a, 452b, 453a, 453b, 453c, 453d, reducing the risk of stockouts or overproduction. Overall, understanding mass flow data is vital for a flour mill's productivity, cost-effectiveness, and quality assurance.

[0066] In the present embodiment of the invention the supply line 31 supplies a flow of material to be ground, such that autonomous quality assessment of the material to be ground is possible. The sensors [221 1 ...221 N] measure, e.g., machine-specific (local humidity, local temperature) and / or grain-specific (color, roughness) characteristics of the material to be ground. As an example, controlling the moisture content of grain and products of their industrial processing under industrial conditions can be based on electrical systems, comprising a dielectric method for measuring the humidity of various classes of products, and the construction of moisture control devices on its basis, and also analyzes the measuring transducer, as well as the measuring schemes of the humidity control devices under consideration. The supply line 31 supplies input material towards the processing unit 32, in particular, e.g., the doser / feeder, sifters 452, 452a, 452b. Supply lines can comprise bucket elevators, chain conveyors, Auger systems, belt conveyors, pneumatic conveyors, or tubular conveyors in particular for specialty grain. Initial product quality assessment can take place along the supply line, for example by using optical sorting techniques which can be used for color, size and mass sorting of rice kernels which are important quality features. Sorting and elimination of impurities, e.g. stones or metal parts for grain mills on the supply line can e.g. be accomplished through optical or electromagnetic methods.

[0067] The technical realization of the processing line 32 can e.g. depend on the type of milling line 3 and its specific technical realization to which the data acquisition 414 and the milling control system 4 can be connected. The processing line 32 comprises one or multiple dosers / feeders 45 and one or multiple roller stands 451 a, 451 b, 451 c, 451 d comprising one or multiple roller mills, ball mills, hammer mills, or other mechanical devices for reducing the size of input material 31 1 . The feeder 45 guides the bulk or mass input material 31 1 towards the roller stand 451 , 451 a, 451 b, 451 c, 451 d for grinding, crushing, pulverizing, or reducing the size of input material depending on the configuration of the roller stand 451 , 451 a, 451 b, 451 c, 451 d.

[0068] The one or multiple output lines 33 comprise at least one set of sensors [2231 ...223N] for screening the end product 331 . Through a milling line 3 the end product 331 with specific milling product parameters is produced. These parameters can e.g. depend specifically on the end product 331 . By "end product" in the sense of the invention in particular bulk goods or a mass of material is understood. By "bulk goods" in the sense of the invention a powdery, granular, pellet-formed, or solid state material is understood which is used in the bulk good processing industry, i.e. in the processing of cereal end products (e.g. wheat, durum wheat, rye, maize, and / or barley) or special milling industry (milling of soy, buck-wheat, barley, pseudo-cereals and / or vegetables), the processing of oil seeds, fish, crustaceans, milling of cocoa beans, nuts, or coffee beans, the manufacturing of fertilizers, the pharmaceutical industry or the chemistry of solids, of the processing of biomass and manufacturing of energy pellet, of solid comprising stone or coal. By "processing of a product" in the sense of the present invention particularly the following is understood: (i) milling, crushing, shred-ding and / or flaking of bulk goods, particularly cereals, cereal milling products and cereal end products of the milling industry or special milling industry for which pairs of milling rolls or flaking rolls may be used; (ii) refining of masses, particularly food masses such as chocolate or sugar masses, for which, for example, pairs of fine rolls may be used; (iii) wet milling, particularly of printing inks, coatings, electronic materials, chemicals in particular fine chemicals; (iv) milling of solid fuels such as for coal / biomass power plants to reduce particle size to improve combustion efficiency and stability; (v) milling and / or crushing of stone, such as limestone, to exact particle sizes.

[0069] Within the processing unit 32 the doser 45 / feeder 45 / scale device 453, 453a, 453b, 453c, 453d can e.g. collect and streamline the input material 31 1 towards the roller stand 451 , 451 a, 451 b, 451 c, 451 d. For hammer mills in particular even distribution of the input material 31 1 over the entire width of the hammer mill has significant influence on the grinding process for fully utilizing the screen surface of a hammer mill 45 and to reduce wear due to more even use of screens 45 and beaters 45 and for energy-efficient use of the mill 2. Specifically for input material 31 1 mixtures homogeneous mixing of the two or multiple components needs to be ensured to avoid individual layers in the premix. The input material 31 1 throughput in the material feeder 45 needs to be adjusted so that the mill 2 is always operated at its optimum operating point. For this the input material 31 1 load and the load on the roller stand motor need to be monitored and controlled for optimum load-dependent dosing. The feeder 45 also supplies the aspiration air required for the grinding process. Optimal aspiration depends, e.g., on the material to be ground, its humidity, and the environmental humidity which are monitored through sensors 2251 ..225N and sensor measurements 241 can be used to optimize and adjust airflow. Control over input material feeding is also required to avoid clogging due to, e.g., elevated moisture / humidity (grains, powdery material), rapid input material 31 1 flow, or input of unshaped material (coal, stone) .

[0070] The machine units 45 within the processing unit 32 comprises at least one grinder for grinding, crushing, pulverizing, or reducing the size of input materials 31 1 to an end product 331 with defined properties. The grinder can comprise roller stands 451 , 451 a, 451 b, 451 c, 451 d, ball mills, hammer mills, etc. Roller mills comprise at least one roll, in particular two rolls of a milling roll pair and can e.g. comprise smooth rolls or fluted rolls. Smooth rolls may be cylindrical or dished. Fluted rolls can e.g. comprise various fluted geometries, e.g. roof-shaped or trapezoidal fluted geometries. A ball mill may e.g. comprise a hollow cylindrical shell rotating about its axis, partially filled with balls as the grinding medium. The balls can be made of steel (chrome steel), stainless steel, ceramic, rubber, or other hard and temperature-resistant materials. The inner surface of the cylindrical shell can be lined with an abrasion-resistant material such as manganese steel or rubber lining. Hammer mills can be based on a steel drum containing a vertical or horizontal rotating shaft or drum on which hammers are mounted, with the hammers free to swing or fixed to the central rotor. Sensors in hammer mills. The steering units 43 generates an operation control signal 434 for controlling and steering the machine units 45, 451 a, 451 b, 451 c, 451 d, 452a, 452b, 453a, 453b, 453c, 453d by transmitting the control signal 434 via the control interface 231 to the machine control 23.

[0071] In the inventive control system 4 each milling line 3 comprises one or more roller stands 451 , 451 a, 451 b, 451 c, 451 d for grinding the input material 31 1 and sensors 22 [221 1 ..221 N] [2221 ..222N] [2231 ..223N] [2251 ..225N] for measuring sensing data 224 of the corresponding milling line 3 and / or the one or more machine controls 23 for steering and / or controlling the operation of the machine units 45, 451 a, 451 b, 451 c, 451 d, 452a, 452b, 453a, 453b, 453c, 453d of the corresponding milling line 3 by the target parameters 233, wherein the machine control 23 comprises a control interface 231 (i) for capturing and transmitting an operational data 232 of the one or more machine units 45, 451 a, 451 b, 451 c, 451 d, 452a, 452b, 453a, 453b, 453c, 453d of the milling line 3 and (ii) for setting the target parameter values 233 based in the control signal 434 generated by the control system 4 for steering and / or controlling the machine units 45, 451 , 451 a, 451 b, 451 c, 451 d, 452a, 452b, 453a, 453b, 453c, 453d .

[0072] Each roller stand 451 , 451 a, 451 b, 451 c, 451 d comprises two rollers turning with a corresponding roller speed and forming a roller gap with a roller gap width, a roller feeder for feeding the rollers via the roller gap with a feeding rate for grinding input material 31 1 and the corresponding machine control 23 for controlling and / or steering the roller speed of each roll and / or the roller gap width and / or the feeder roll speed of the roller feeder of the roller stand 451 , 451 a, 451 b, 451 c, 451 d by the corresponding machine control 23 based on the target parameter 233. The target parameter 233 define for the corresponding machine unit 45, 451 , 451 a, 451 b, 451 c, 451 d, 452a, 452b, 453a, 453b, 453c, 453d the target values for the steerable devices of the machine unit 45. The target parameter 233 for a sifter 452, 452a, 452b comprises sifting speed for adjusting a speed of the sifter affecting the efficiency of the input material 31 1 separation of the processing line 32 and / or an amplitude of vibration for optimizing the movement of flour through the sifter 452, 452a, 452b and / or an airflow rate for adjusting the airflow generated by an central aspiration system 45 to remove fine particles of the input material 31 1 within the sifter 452, 452a, 452b and / or a sifter feed rate for controlling the rate at which input material 31 1 to be sifted is fed into the sifter 452, 452a, 452b for maintaining optimal operating conditions for example. The target parameter 233 of the one or more roller stands 451 , 451 a, 451 b, 451 c, 451 d comprises a roller stand target parameter capturing a target roller gap width and / or a target roll temperature and / or a target roller speed for each roll and / or a target roller feeder speed and / or a target feeder gap for steering and / or controlling the roller stand 451 , 451 a, 451 b, 451 c, 451 d by the machine 23 via the control interface 231.

[0073] In the embodiment of the present invention the at least one steering unit 43 comprise a gap control module 435 for steering the roller gap width and / or a feeding control module 436 for steering the feeder roll speed and / or a slip control module 437 for steering a ratio of the roller 4512a, 4512b for controlling and steering at least one of the one or more roller stands 451 , 451 a, 451 b, 451 c, 451 d.

[0074] The one or more steering units 43 each comprising (i) a steering input unit 431 for receiving the setpoint parameter values 415, a machine intelligence unit 432 for generating target parameter values 233 based on the milling parameters 21 and / or operational parameters 232, an electronic signal generator 44 for generating the electronic control signal 434 based on the target parameter values 433. The signal generator 44 generates the control signal 434 for controlling and steering the machine units 45, 451 , 451 a, 451 b, 451 c, 451 d, 452, 452a, 452b, 453, 453a, 453b, 453c, 453d by transmitting the control signal 434 via the control interface 231 to the machine control 23.

[0075] The mill control unit 49 and / or the one or more steering units 43 controlling and / or steering the machine units 45, 451 , 451 a, 451 b, 451 c, 451 d, 452, 452a, 452b, 453, 19

[0076] 453a, 453b, 453c, 453d and its machine units 45, 451 , 451 a, 451 b, 451 c, 451 d, 452, 452a, 452b, 453, 453a, 453b, 453c, 453d can e.g. comprise web technology, i.e. decentralized network technologies, which enables collective parameter changes and collective software updates. Collective parameter changes and collective software updates over any number of mills 2 and / or milling lines 3, and the interconnectivity of the machine units 45, 451 , 451 a, 451 b, 451 c, 451 d, 452, 452a, 452b, 453, 453a, 453b, 453c, 453d are advantages of a centralized control system 5 according to the invention.

[0077] The mill control 49 comprises a mill interface 4 1 for sending setting point parameter and / or target mass flow parameter and / or operational recipe parameter and / or passage information and receiving setpoint instruction parameter and / or recipe parameter, wherein human machine interfaces, such as mobile devices, displays, are connectable to the mill control 49 via a data transmission network.

[0078] In a possible embodiment the machine intelligence unit 432 can e.g. be connected by an interface with the worldwide backbone network, i.e. the internet and / or an intranet. The machine intelligence unit 432 can e.g. be connected by means of the network interface via a data transmission network and the network interface to the control data management unit (milling plant control system) 5. A web server application provides the display and input / output front end and / or control / monitoring information for remote client or browser wherein the remote client can e.g. be connected to the network by means of the network interface.

[0079] In one possible embodiment a mobile application provides access to the webserver application of the central data management unit 5. The usage can, e.g., be limited to a local area network (LAN) which can e.g. be connected to the installation 1 . The machine control connection can e.g. comprise a software-based engine. In one possible realization commercial engines can e.g. comprise the Phoenix Engine which provides real time, graphics-based and interactive web-based applications to remotely monitor, control and steer individual mills 2 or multiple connected mills 2 at one or multiple geographical locations 70A..70N. Together with the machine intelligence unit 432 such a configuration of the one or multiple mills 2 and / or milling lines 3 provides the technical realization and infrastructure to remotely surveil the operability and functionality of the machine units 45, 451 , 451 a, 451 b, 451 c, 451 d, 452,. 452a, 452b, 453, 453a, 453b, 453c, 453d, the quality of the end product 331 and to causally relate the quality of the end product 331 with the sensing data 224 of the sensors 22, [221 1 ...221 N] [2221...222N] [2231 ...223N] [2251 ...225N.

[0080] In a further embodiment of the present invention the setpoint parameters 415 generated by the line control 41 are send via the mill interface 491 to one or more mill operators. The one or more operator reviews the setpoint parameter values 315 received from the line control 41 and sends an instruction to send the setpoint parameters values 315 to the one or more steering units 43 and / or manually changes the target parameter values 233 of the one or more machine units 45, 451 , 451 a, 451 b, 451 c, 451 d, 452, 452a, 452b, 453, 453a, 453b, 453c, 453d or ignores the received setpoint parameter values 415. In this embodiment of the present invention the mill control 49 is semi-automatically controlling and / or steering the machine units 45, 451 , 451 a, 451 b, 451 c, 451 d, 452, 452a, 452b, 453, 453a, 453b, 453c, 453d of the mill 2, so that the operator controls the setpoint parameter values 315 of the mill 2. An advantage of this embodiment of the invention is that new product recipes and / or new configuration of the mill 2 can be tested under the control of the operator.

[0081] In a further embodiment of the present invention as illustrated in the figure 5 the milling control system 4 of at least one of the mills 2 comprising the mass flow unit 48 with a mass flow parameter 482, a mass flow interface 483 for sending the mass flow parameter 482 of the at least one of the mills 2 to the corresponding one or more line controls 41 , and the mass flow module 48 for generating and / or measuring the mass flow parameter 482 based on the plant parameter 212 of the at least one of the mills 2. In another embodiment of the present invention, not illustrated in Figure 5, the mass flow module 48 for generating and / or measuring the mass flow parameter 482 based on the product parameter values 21 1 and / or operational parameter values 232 and / or plant parameter 212 of the at least one of the mills 2.

[0082] The mass flow module 481 comprises a data analyzer and a linear leastsquares solver (LLSS), wherein the data analyzer generates based on the plant parameter 212 and the target parameter 433 a separation matric ASeP, a collection matric Acoi and a target vector d, and the LLSS generates based on the ASeP„ Acoi and target vector d the detailed mass flow x of the milling line 3.

[0083] In a flowsheet the mass flow of the input material 31 1 is separated in the several passages separated and collected resulting in the ‘detail mass flows' (x) . The mass flow matrices comprise a separation matric ASePand a collection matric Acoi and can be described and fed as constrained into the solver to describe the topology:

[0084] Figure 7a illustrates the matric Asep and Figure 7b the matric Acoiof this the simple mill illustrated in Figure 1 . Figure 7c illustrates the vector x with all mass flow

[0085] In the target vector (d) scale values, indicative values or estimates can be added for each or a combination of a detailed mass flow x as illustrated by an example in the figures 6, 7a, 7b and 7c. The detailed mass flows x is generated by LLSS by minimizing the difference to the target vector d. In a further embodiment the present invention the measured / generated values of the target vector d are weighted according to their accuracy. The target vector values d comprises a measured scale value of machine units 45, 453a, 453b, 453c, 453d, a generated target value and a configuration design point, wherein:

[0086] C is the identity matrix with its size matching to length of x x is the vector of all detail mass flows d is the target vector of the solver, consisting of the plant configuration points or indicative measured values or measured scale values.

[0087] In another embodiment of the present invention the mass flow unit 48 comprising a training module configured to optimize the weights of the target vector d to improve the accuracy based on sensing data 224 and / or simulated data.

[0088] Figure 1 illustrates an example of the one or more mills 2 comprising a weighting device 453a for weighting the input material 31 1 with a mass flow Bl , two roller stands 451 a, 451 b each grinding a mass flow Bl -1 and Bl -2 separated out of the mass flow Bl , a sifter 452a sifting a mass flow SB1 based on the combination of the mass flow Bl -1 and Bl -2 wherein the sifter 452a is separating the mass flow SB1 into three mass flows SB1 Br, SB1 C1 and SB1 Fl based on the particle size of the material products of mass flow SB1 , two roller stands 451 c, 451 d each grinding a separated mass flow Cl -1 and Cl-2 out of SB1 C1 resulting in a combined mass flow SCI , a sifter 452b for separating the mass flow SCI into three mass flows SCI Br, SCI Fl , SCI F2 based on the particle size of the product materials of the mass flow SCI and three scale device 453b, 453c and 453d for weighting the mass flows SCI Br, SCI Fl and SCI F2, wherein the mass flow Fl is the end product 331 and a combination of the mass flows SBF1 and SCI Fl , wherein the mass flow F2 / SC1 F2 is the end product 331 and wherein the mass flow Br is waste 332 and a collection of the mass flows SCI Br and SB1 Br.

[0089] Based on the mass flow SCI , SCI Br, SCI Fl and SCI F2 and / or SB1 , SB1 Br, SB1 Fl and SB1 F2 of the sifters 452a, 452b provides an indication of a granulation distribution of the sifters 452a, 452b processed material.

[0090] The gap control module 435 comprising a gap control interface for receiving operational parameters 232, including the roller operation parameters and / or the setpoints parameter values 415 and / or operational recipe parameter of the corresponding roller stand 451 , 451 a, 451 b, 451 c, 451 d and / or sensors 22 for the product [2231 ...223N] and / or sensors for the machine units [2221 ..222N] and for setting the gap setpoint for corresponding steering unit 43, a gap control intelligence for generating the gap setpoint based on the roller operation parameter values and / or the setpoint parameter values 415 and / or operational recipe parameter of the corresponding roller stand 451 , 451 a, 451 b, 451 c, 451 d. In a further embodiment the present invention the recipe comprises a steering curve parameter with roll temperature and corresponding gap width and / or passage information. The gap control module 435 comprises the following process steps:

[0091] Measuring roll 4512a, 4512b temperature

[0092] Receiving passage information of the roller stand 451 , 451 a, 451 b, 451 c, 451 d

[0093] Reading gap control status of the roller stand 451 , 451 a, 451 b, 451 c, 451 d

[0094] Look-up target gap width based on roll temperature and passage information of the roller stand 451 , 451 a, 451 b, 451 c, 451 d

[0095] Setting target gap width for adjusting gap width of the roller stand 451 , 451 a, 451 b, 451 c, 451 d.

[0096] In a further embodiment of the invention the line intelligence 414 can e.g. be used for forecasting of expected changes in the machine units 45, 451 , 451 a, 451 b, 451c, 451 d, 452, 452a, 452b, 453, 453a, 453b, 453c, 453d. For example, time series analysis of observables such as temperature, rotational speed, surface roughness, etc. of the one or two rollers 4512a, 4512b of a roller pair based on autoregressive models, autoregressive integrated moving averages, or temporal fusion transformer models, inter alia, provide information about future expected deterioration that can be anticipated. In such a continuous mode-of-operation, roller re-placements due to increased surface roughness can be anticipated instead of being suddenly and unexpectedly triggered by reaching a threshold value that leads to an abrupt stop of a milling line 3. In another embodiment, the line intelligence 414 can e.g. be trained to causally relate deteriorating output material quality, such as color variations or increased spread in the particle size distributions, with the recipe and to predict realtime changes in the recipe to reach desired target values in color or particle size.

[0097] The line intelligence 414 can e.g. be trained to provide a plurality of machine-learned models (ensemble method) with the continuously measured sensor values as input to evaluate the ensemble averaged system loss function as the target quantity. Each machine-learned model comprises at least one neural network structure. The at least one neural network structure can e.g. comprise at least one Convolutional Neural Network (CNN) . Each machine-learned model can also comprise other machine learning structures as deep learning or deep structured learning or any other applicable machine learning method producing appropriate information, wherein the deep learning structures are based on artificial neural networks with representation learning. The learning in the learning mode can e.g. be supervised, semisupervised or unsupervised. However, supervised learning modes may be a preferred embodiment variant, in particular in relation to active learning data structures of the present invention. Deep learning structures can e.g. comprise deep neural network structure and / or deep belief network structures and / or recurrent neural network structures and / or CNNs etc. In the case of supervised learning the proposed machine learning structure can be trained by an initial dataset annotated by a human expert, which typically outperforms classification results by human raters and / or rule-based decisions and allows continuous evaluation of the system-wide loss function instead of simple on / off decisions based on thresholds. For supervised learning a regression model can e.g. be trained to predict output from measurement values provided by sensors [SAI ...SDN], By providing the model with new observations at runtime, the machine learning model can continuously learn and refine the relationship between output material quality and sensing data values 224 of the sensors [221 1 ..221 N] [2221 ..222N] [2231..223 N] [2251 ..225 N].

[0098] The remote control and regulation of the machine units 45, 451 , 451 a, 451 b, 451 c, 451 d, 452, 452a, 452b, 453, 453a, 453b, 453c, 453d can e.g. be locally carried out by means of the milling control system 4, in particular, the line control 41 , through real time data transfer. In particular, the line control 41 and / or the mill interface 431 of the milling control system 4 may e.g. comprise a network interface. In the present embodiment of the invention the data acquisition unit 412 comprises the network interface. Via the network interface 4126 access can be gained to the milling control system 4 of the mill 2 and / or milling line 3 with the PLCs 4121 and the network interfaces of the milling control system 4. The line control 41 connected to the signaling unit 44 for generating the control signal 434 to be transmitted to passage-specific sensors [2221 ...222N] and environmental sensors [2251 ...225N], In so doing, according to the invention, by means of at least one of the control signal 434 for the at least one of the passage-specific sensors [2221 ...222N] and environmental sensors [2251 ...225N], one or several mills 2 and / or milling lines 3 with machine units 45, 451 , 451 a, 451 b, 451 c, 451 d, 452, 452a, 452b, 453, 453a, 453b, 453c, 453d are centrally and autonomously optimized and / or individually controlled by the milling control system 4. The environmental sensors [2251 ...225N] may e.g. comprise at least sensors for measuring humidity and / or air pressure and / or ambient temperature. The passage-specific sensors [2221 ...222N] may e.g. comprise at least sensors measuring local operating parameters of the steerable de-vices 45 such as, for example, temperature of the input material, color of the input material, humidity of the input material, temperature of the at least one or all rollers of a roller pair, rotational velocity of one or all rollers of a roller pair, milling fineness of the output material, color of the output material, etc.

[0099] Operation of the invention requires electrical power on site which can, e.g., be provided by power lines, power generators, photovoltaic or other means to produce electricity, or by emergency power systems. Power grid stability can, e.g., be controlled via a library of control algorithms comprised by the repository 51 for active and / or reactive power certification. Communicating of the controller can, e.g., be managed as a switch via ethernet ports, or separately over two isolated networks to communicate directly from the controller with the grid operator through a VPN tunnel with secure end-to-end encryption, via standard protocols like IEC 60870 -104 / 101 and Modbus or ripple control receivers. This allows transfer of the active power, the reactive power, or the power factor, among others.

[0100] Monitoring of transportation means, such as roads or rails, by truck and / or train to as-certain 24 / 7 operation can be accomplished through continuous, autonomous monitoring of transportation means. Autonomous detection of road pavement defects, including cracks, potholes, or related faults that compromise or prevent industrial access can be accomplished, e.g., by imaging methods or by using unmanned aerial vehicles equipped with hyperspectral imaging equipment. Plant condition monitoring and predictive maintenance ensures uninterrupted plant availability. By measuring environmental parameters, such as temperature, pressure, or weather parameters (precipitation), the machine learning unit 432 determines a score characterizing the probability for operability P (op) of the entire plant. If P (op) is within a specified margin (or threshold) of a trigger value, the plant is deemed operable. For lower probability P (op), but above a critical lower value the likelihood for impediments in operability is sufficiently large to warrant detailed analysis for specific, preventive maintenance and for P (op) below the critical lower value, operability of the plant is not given, and emergency shutdown is executed.

[0101] The at least one or more sensors [2251 ...225N] measuring environmental and / or contextual parameters including, for example, temperature, humidity, or air pressure and / or contextual parameters such as stability of the power grid used to operate the mill 2 or milling line 3, or road driving conditions for trucks transporting goods to / from the milling plant.

[0102] Sensors [2251 ...225N] measuring environmental parameters can be connected to the data acquisition unit 412 through a PLC 4121 . The measurement parameters are fed to the line unit 43 and the machine intelligence unit 432, optimizing target parameters 233 and / or target parameter 433 (such as roller speed or materials flow, depending on environmental conditions) by minimizing the total loss function of the milling system 1 .

[0103] In an embodiment of the invention multiple milling lines 3 can be operated together to form a milling installation. A milling installation shall be understood to comprise all technological devices and processes for producing the grainy and / or powdery (flour-like) or only husked or squashed end products 331 from coarse, solid milling materials, which are used for processing the milling material. In particular, the steering unit of each mill of a milling installation can be connected with or linked to a central data management unit 5 through a data transmission network, the central data management unit signaling automated management procedures and / or alarm devices and / or geographic or typographic surveillance processes.

[0104] Reference list

[0105] 1 Milling system

[0106] 2 Operation device, mill(s)

[0107] 201 ..20N Several mills

[0108] 21 Milling parameters

[0109] 21 1 Product parameter

[0110] 212 Mill parameter / Plant parameter

[0111] 22 Sensors

[0112] 221 1 ...221 N Sensors for milling product

[0113] 2221 ...222N Sensors for machine units

[0114] 2231 ...223N Sensors for final product

[0115] 224 Sensing data

[0116] 2251 ...225N Sensors for environmental and / or contextual parameters

[0117] 23 Machine control

[0118] 231 Control interface

[0119] 232 Operational parameter

[0120] 233 Target parameter

[0121] 24 Environmental sensor

[0122] 241 Environmental data

[0123] 3 Milling line / passage

[0124] 31 Supply line

[0125] 31 1 Input material

[0126] 32 Processing line

[0127] 33 Output line

[0128] 331 End product

[0129] 332 Waste

[0130] 333 Return to Supply

[0131] 4 Milling control system

[0132] 401 40N Several milling control systems

[0133] 41 Line Control

[0134] 412 Data acquisition unit

[0135] 4121 PLC

[0136] 4122 Data preprocessing

[0137] 4123 Data homogenization 4124 Data interface

[0138] 4125 Output data

[0139] 413 Line parameter

[0140] 414 Line intelligence

[0141] 4141 Machine learning / artificial intelligence

[0142] 415 Setpoint parameters

[0143] 416 Line signal generator

[0144] 417 Reference / trigger values Steering Unit

[0145] 431 Steering Input Unit

[0146] 432 Machine intelligence unit

[0147] 433 Target parameter

[0148] 434 Control signal

[0149] 435 Gap controller

[0150] 436 Feeding controller

[0151] 437 Slip controller Signal Generation Unit Machine units

[0152] 451 Roller Stand

[0153] 451 a, 451 b, 451 c, 451 d Roller stands

[0154] 452 Sifter

[0155] 452a, 452b Sifters

[0156] 453 Scale device

[0157] 453a, 453b, 453c, 453d Scale units Alarm / Shutdown Data links

[0158] 471 Data transmission network

[0159] 472 Data transmission network Mass flow unit

[0160] 481 Mass flow module

[0161] 482 Mass Flow Parameter

[0162] 483 Mass flow interface

[0163] Asep Separation mass flow matric

[0164] Acoi Collection mas flow matric X Mass flow vector

[0165] 49 Mill control

[0166] 491 Mill Interface Central data management unit (CDMU) 51 Repository

[0167] 51 1 Configuration library

[0168] 512 Recipe library

[0169] 513 End product library

[0170] 514 Machine unit library A ...70N Geographical location

Claims

Claims l .A flour milling system (1 ) comprising one or more mills (2) with at least one milling line (3) with several machine units (45, 451 , 451 a, 451 b, 451 c, 451 d, 452, 452a, 452b, 453, 453a, 453b, 453c, 453d) for the industrial processing of an input material (31 1 ) to an end product (331 , 332, 333) and a machine-learning based milling control system (4) comprising for each mill (2) an electronic mill control unit (49) with one or more steering units (43) for steering and / or controlling associated machine units (45, 451 , 451 a, 451 b, 451 c, 451 d, 452, 452a, 452b, 453, 453a, 453b, 453c, 453d) by generated control signaling (434), the machine units (45) comprising one or more roller stands (451 , 451 a, 451 b, 451 c, 451 d) for grinding the input material (31 1 ) with sensors (22) [2211 ...221 N] [2221...222N] [2231 ...223N] measuring sensing data (224) of the corresponding milling line (3) and machine controls (23) for steering and / or controlling the operation of the machine units (45, 451 , 451 a, 451 b, 451 c, 451 d, 452,. 452a, 452b, 453, 453a, 453b, 453c, 453d) of the corresponding milling line (3) by target parameters (233), wherein the machine controls (23) comprise control interfaces (231 ) capturing and transmitting operational data (232) of the one or more machine units (45) and setting target parameter values (233) to the one or more machine units (45) based in the control signaling (434) generated by the control system (4), wherein each roller stand (451 , 451 a, 451 b, 451 c, 451 d) comprises at least two rollers turning with a corresponding roller speed and forming a roller gap with a roller gap width, a roller feeder feeding the rollers via the roller gap with input material (31 1 ) at a feeding rate and a corresponding machine control (23) for controlling and / or steering the roller speed and / or the roller gap width and / or the feeder speed of the roller feeder of the roller stand (451 , 451 a, 451 b, 451 c, 451 d) by the corresponding machine control (23) based on the target parameter (233), wherein the target parameter (233) of the one or more roller stands (451 , 451 a, 451 b, 451 c, 451 d) comprises roller stand target parameters capturing a target roller gap width and / or a target roll temperature and / or a target roller speed for each roll and / or a target roller feeder speed and / or a target feeder gap for steering and / orcontrolling the roller stand (451 , 451 a, 451 b, 451 c, 451 d) via the control interface (231 ), characterized in that the mill control unit (49) comprises one or more steering units (43), each associated with a mill (2), comprises a gap controller (435) steering the roller gap width and / or a feed controller (436) steering the feeder roll speed and / or a slip controller (437) steering the ratio of the roller speed of the rollers (451 , 451 a, 451 b, 451 c, 451 d), in that the mill control unit (49) comprises one or more line controls (41 ) with a data acquisition unit (412) for receiving and / or pre-processing milling parameter values (21 ) and / or operational parameter values (232) of the machine units (45) and / or a line parameter values (413) of the milling line (3), with a line intelligence (414) for generating machine setpoint parameter values (415) based on the milling parameter (21 ) and / or operational parameter (232) and / or line parameter values (413), and with a line signal generator (416) for sending the setpoint parameter values (415) to the steering units (43) of the milling line (3), in that the one or more steering units (43) comprise a steering input unit (431 ) for receiving setpoint parameter values (415), a machine intelligence unit (432) for generating target parameter values (433) based on the milling parameters (21 , 21 1 , 212) and / or operational parameters (232), an a signal generator (44) for generating the control signal (434) based on the target parameter values (433), the signal generator (44) generating an operation control signal (434) for controlling and steering the machine units (45) by transmitting the operation control signal (434) via the control interface (231 ) to the machine control (23) .

2. The milling system ( 1 ) according to claim 1 , characterized in that the operation parameter (232) comprise a roller operation parameter capturing the roller gap width and / or the roller temperature of the rollers and / or a temperature distribution of the rollers and / or a roller stand status of the roller stand (322) and / or a feeding rpm and / or a feeding gap width.

3. The milling system ( 1 ) according to according to claims 1 to 2, characterized in that the milling parameters (21 ) comprise a plant parameter (212) foreach of the one or more mills (2) comprising a mill configuration and / or an product recipe, wherein the mill configuration comprises machine units (45, 451 , 451 a, 451 b, 451 c, 451 d, 452,. 452a, 452b, 453, 453a, 453b, 453c, 453d) and / or mass flow parameter in relation to the operational parameter (232) of the machine units (45, 451 , 451 a, 451 b, 451 c, 451 d, 452, 452a, 452b, 453, 453a, 453b, 453c, 453d) and / or status of the machine units (45, 451 , 451 a, 451 b, 451 c, 451 d, 452, 452a, 452b, 453, 453a, 453b, 453c, 453d) and / or mass flow relations between the machine units (45, 451 , 451 a, 451 b, 451 c, 451 d, 452, 452a, 452b, 453, 453a, 453b, 453c, 453d) and wherein product recipe comprises milling product specifications and / or end product specifications and / or initial mill configuration.

4. The milling system (1 ) according to ) according to claims 1 to 3, characterized in that the flour milling system (1 ) comprising a central data management unit (5) with a repository (51 ), wherein the repository (51 ) comprises a configuration library 51 1 and / or recipe library (512) and / or end product library (513), wherein the configuration library captures at least one mill configuration and wherein the recipe library comprises at least one or more product recipes.

5. The milling system ( 1 ) according to ) according to claims 3 to 4, characterized in that the control system (4) of at least one of the mills (2) comprises mass flow unit (48) with a mass flow parameter (482), a mass flow interface (483) for sending the mass flow parameter (482) of the at least one of the mills (2) to the corresponding one or more line controls (41 ), and a mass flow module (48) for generating and / or measuring the mass flow parameter (482) based on the product parameter values (21 1 ) and / or the operational parameter values (232) and / or the plant parameter (212) of the at least one of the mills (2) .

6. The milling system (1 ) according to claims 1 to 5 wherein the data acquisition unit (412) captures and preprocesses measurement data by standardizing, normalizing and filtering / noise-reducing captured data, the output of the acquisition unit (412) being delivered to the line intelligence unit (414).

7. The milling system ( 1 ) according to claims 1 to 6, wherein the line intelligence unit (414) comprises a machine learning (4141 ) and / or artificial intelligence unit (4141 ) transforming input from the data acquisition unit (412) to setpointsparameter (415) suitable for a line signal generator (416) to be communicated to the steering units (43) of the milling lines (3) .

8. The milling system ( 1 ) according to claim 1 to 7 wherein the machine Intelligence unit (4141 ) comprises a threshold and / or trigger unit (417) configured to trigger the warning / alarm signal unit (46) in case of exceeding of a measuring signal to a defined triggering / threshold value.

9. The milling system (1 ) according to claim 7 wherein the machine learning unit (432) comprises a system-wide loss function unit, the loss function trained with supervised and / or unsupervised learning.

10. The milling system (1 ) according to claim 7 wherein supervised machine learning (414) is based on labelled data transmitted from each sensor [221 1...223N] and / or operational parameter (232) and / or milling parameter (21 , 21 1 , 212) for generating the parameters for steering the input part, the processing part, and the output part in that humidity measurements of the material in the input part dynamically adjust roller pair rotation speed and roller pair separation and / or unsupervised learning based on unlabeled data from each sensor and / or measuring device in that classification of the quality of the output material either leads to recommissioning to the input line or transport to disposal or further processing of the ground product.11 . The milling system ( 1 ) according to claim 7 wherein the generated parameters by the machine learning unit (432) comprising monitoring circuitry with a defined monitoring process, at least comprising the steps of comparing actual parameters with predetermined parameters for each sensor for ahead-of-time alarm generation if actual parameters are predicted to fall outside the interval covered by predetermined parameters,determining optimal operational parameters including input line supply belt speed, roller pair speed, and roller pair separation by minimizing the loss function of the machine learning model, predicting servicing times for rollers due to wear and autonomously trigger ordering of replacement rollers and planning repair to minimize or fully eliminate downtime of one or several milling lines, linking multiple mills to predict local quality parameters including humidity, size, and uniformity of the input material from external parameters including growth and harvesting conditions of the input material.

12. The milling system (1) according to claim 1 to 1 1 characterized in that the steering unit (41) of each mill (2) is connected with the central data management unit (5) through the data transmission network (473), the central data management unit (5) signaling automated service management procedures and / or alarm devices and / or geographic or typographic surveillance processes.

13. The milling system (1) according to claim 1 to 12 characterized in that the one or more mills (2) and / or milling system (1) comprises environmental sensor [2251 ...225N] for measuring environment data (241) and / or contextual parameters.

14. The milling system (1) according to claim 13 wherein the machine learning unit (414) uses generated parameters for predicting parameters for optimized operation control signal (424) of the milling system (1) given environmental data (241) and / or contextual parameters including environmental (24) temperature and / or humidity, and / or humidity of the material to be ground.

Images