EP3513263A2 - Verwendung umfassender künstlicher intelligenz bei anlagen der grundstoffindustrie - Google Patents
Verwendung umfassender künstlicher intelligenz bei anlagen der grundstoffindustrieInfo
- Publication number
- EP3513263A2 EP3513263A2 EP17762085.3A EP17762085A EP3513263A2 EP 3513263 A2 EP3513263 A2 EP 3513263A2 EP 17762085 A EP17762085 A EP 17762085A EP 3513263 A2 EP3513263 A2 EP 3513263A2
- Authority
- EP
- European Patent Office
- Prior art keywords
- data
- artificial intelligence
- automation system
- machine interface
- anl
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Ceased
Links
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B23/00—Testing or monitoring of control systems or parts thereof
- G05B23/02—Electric testing or monitoring
- G05B23/0205—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults
- G05B23/0218—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults
- G05B23/0243—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults model based detection method, e.g. first-principles knowledge model
- G05B23/0254—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults model based detection method, e.g. first-principles knowledge model based on a quantitative model, e.g. mathematical relationships between inputs and outputs; functions: observer, Kalman filter, residual calculation, Neural Networks
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
- G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
- G05B13/0265—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric the criterion being a learning criterion
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B21—MECHANICAL METAL-WORKING WITHOUT ESSENTIALLY REMOVING MATERIAL; PUNCHING METAL
- B21B—ROLLING OF METAL
- B21B37/00—Control devices or methods specially adapted for metal-rolling mills or the work produced thereby
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B19/00—Program-control systems
- G05B19/02—Program-control systems electric
- G05B19/418—Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS] or computer integrated manufacturing [CIM]
- G05B19/4184—Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS] or computer integrated manufacturing [CIM] characterised by fault tolerance, reliability of production system
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B23/00—Testing or monitoring of control systems or parts thereof
- G05B23/02—Electric testing or monitoring
- G05B23/0205—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults
- G05B23/0259—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterized by the response to fault detection
- G05B23/0286—Modifications to the monitored process, e.g. stopping operation or adapting control
- G05B23/0289—Reconfiguration to prevent failure, e.g. usually as a reaction to incipient failure detection
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B2219/00—Program-control systems
- G05B2219/30—Nc systems
- G05B2219/31—From computer integrated manufacturing till monitoring
- G05B2219/31368—MAP manufacturing automation protocol
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B2219/00—Program-control systems
- G05B2219/30—Nc systems
- G05B2219/32—Operator till task planning
- G05B2219/32128—Gui graphical user interface
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B2219/00—Program-control systems
- G05B2219/30—Nc systems
- G05B2219/33—Director till display
- G05B2219/33002—Artificial intelligence AI, expert, knowledge, rule based system KBS
Definitions
- the present invention is based on an operating method for a plant of the basic industry
- an automation system determines control data and outputs them to controlled elements of the installation and thereby controls and / or monitors the installation
- the automation system at least partially receives the measurement data and takes into account the received measurement data as part of the determination of the control data
- a human-machine interface outputs measurement data of the sensor devices and / or data of the automation system and / or planning data of a production planning system to a person and / or receives control commands from the person and forwards them to the automation system
- the artificial intelligence utilizing the received data and commands, determines evaluation results and makes the evaluation results available to the person via the man-machine interface and / or an independent output device and / or makes them available to the production planning system for the plant of the basic industry; or directly or through the Man-machine interface pretends to the automation system in the form of control commands,
- the data received from artificial intelligence are at least partially dimensional data.
- Dimensional data in the sense of the present invention are similar data which follow one another sequentially in at least one dimension.
- the dimension can be a local dimension.
- the dimensional data form, for example, a local distribution, for example a temperature distribution.
- the distribution can be one-dimensional, two-dimensional or three-dimensional.
- a Fourier transform of the place is to be understood as a dimension.
- the dimension may be a temporal dimension.
- the dimensional data form, for example, a time course, for example a temperature profile.
- Even a Fourier transform of time is to be understood as a dimension.
- An example is a frequency spectrum.
- a combination (at least) of a local dimension with a time dimension is possible, so for example, the time course of a local distribution.
- the temperature is only an example. Instead of the temperature also other sizes can be utilized, for example, a pressure or train course and / or a corresponding distribution. Also, a crowning over the bandwidth or the length of the roll bale of a roll of a rolling stand - possibly in addition as a time course - is possible. Likewise, other sizes can be used.
- the present invention is further based on a computer program comprising machine code which can be processed by a computer, wherein the processing of the machine code by the computer causes the computer to implement an artificial intelligence, which Measured data acquired by sensor devices of a plant of the basic industry in the operation of the basic industry plant, control data and / or internal data determined by an automation system controlling and / or monitoring the plant of the basic industry and in the case of control data for controlling the plant Basic industries are issued to controlled elements of the basic industry plant, and of a
- Man-machine interface of the basic industry plant at least partially accepts data output to a person
- the data received from artificial intelligence are at least partially dimensional data.
- the present invention is further based on a computing device, wherein the computing device is programmed with such a computer program and connected for the transmission of information at least with an automation system, sensor devices and a man-machine interface of a plant of the basic industries.
- the present invention is further based on a plant of the basic industry,
- the system has an automation system, which determines control data and outputs to controlled elements of the system and thereby controls the system and / or monitored, - wherein the system comprises sensor devices which detect measurement data of the system,
- the automation system at least partially receives the measurement data and takes into account the received measurement data as part of the determination of the control data
- the system has a man-machine interface that outputs measurement data of the sensor devices and / or data of the automation system to a person and / or receives control commands from the person and forwards them to the automation system,
- the system has such a computing device.
- the primary industry may be an installation of the steel industry or a plant of the metal-producing industry.
- examples of such systems are an iron production plant such as a blast furnace or a direct reduction plant, such as a Finex, Corex or Midrex type direct rotary kiln or rotary hearth with or without Submerged Are Furnace.
- Further plants are, for example, downstream facilities in which steel is produced from pig iron. Examples of such systems are an electric arc furnace, a converter and plants in which Pfannenvone done such as a vacuum treatment plant.
- Further plants are the steelmaking subordinate institutions in which a primary shaping of the metal and a transformation of the urgeformten metal take place.
- the rolling mills may, for example, be a rolling mill for rolling a flat rolled stock such as a roughing train, a finishing train, a Steckel mill and others. Furthermore, the rolling mills may be a rolling mill for rolling a different cross section act, for example, a billet cross-section.
- the rolling mill may alternatively be a metal rolling mill, a cold rolling mill, or a combined rolling mill where the metal is first hot and then cold rolled. Even a cooling section - possibly in combination with a rolling mill - can be regarded as an investment in steelmaking.
- Other plants upstream or downstream of a rolling mill are also suitable, for example a glow or a pickle.
- Process automation usually involves several levels. Level 0 is formed by the sensors and the actuators. Level 1 is the so-called basic automation, which implements the control loops. Level 2 contains the technological automation, which includes the process models and determines the setpoints for the control loops. There are also other levels known, for example, concern the production planning.
- the object of the present invention is to provide opportunities to automate the operation of the plant of the basic industry not only in normal operating conditions, but also in exceptional operating conditions automisie ren. Ideally, any control intervention should be avoided. At least the number of control interventions should be reduced.
- an operating method of the type mentioned in the introduction is configured in that the dimensional data comprise an image output to the person via the human-machine interface, a temporal sequence of such images, a part of such an image or a temporal sequence of such a part.
- the picture can be an op- be a table or an infrared image.
- the images may be in the form of pixel arrays, 2D surfaces, or 3D volumes. Also, a specification of local or temporal curves is possible.
- the dimensional data may include an acoustic oscillation or a spectrum of such oscillation.
- the dimensional data extend at least partially in at least two dimensions-in particular in at least two local dimensions-and that the artificial intelligence determines curves of the same value and / or gradient for the determination of the evaluation results for the dimensional data in the case of temperatures, for example, isotherms or temperature gradients.
- the dimensional data from the automation system may include computed data based on a model. Examples of such data are a calculated temperature profile or a calculated temperature distribution of a metal strip in a finishing train or in a cooling section. Optionally, in addition, the consideration of the phase of the metal strip can take place. Other calculated data can be used, for example, not directly measurable, but calculated by means of a so-called soft sensor data.
- Artificial intelligence is, generally speaking, a unit implemented by a computing device that acts on outsiders like a human intelligence. Artificial intelligence includes in particular machine learning (ma- chine learning), machine vision and object recognition (computer vision), speech processing and robotics.
- Learning Machines is an Artificial Intelligence that can learn how to behave in order to get the highest possible score based on a variety of situations or processes, together with the associated evaluation (or a criterion for determining the rating).
- the evaluation can either be given to Artificial Intelligence or independently determined by Artificial Intelligence on the basis of a valuation criterion known to Artificial Intelligence.
- the Artificial Intelligence accumulates and thus contains the knowledge of a suitable, optimal operation of the plant of basic material theory.
- Machine vision and recognition of objects is in particular the extraction of objects in images and their assignment to a general category.
- Objects in this context are two-dimensional elements that can be moved without influencing other elements.
- Speech processing in particular includes the independent recognition of words and in the context of the present invention also generally the analysis of acoustic vibrations.
- the present invention shows its full strength in the case where the artificial intelligence, in parallel with the determination of the evaluation results, determines an evaluation of the evaluation results and retrains itself based on the evaluation. As a result, an increasingly better evaluation of the data and commands received by artificial intelligence takes place.
- Data sources for learning can be, in particular, the automation system, the sensor devices and the human-machine interface.
- the data includes - at least the dimensional data - both for learning and in later operation.
- singular data - that is, individual values (as opposed to sequential values) - can also be exploited.
- the artificial intelligence used can be determined as needed and according to the application.
- the nature of artificial intelligence can be different.
- artificial intelligence may be referred to as an artificial neural network, a support vector machine, a decision tree, Bayesian belief network, as so-called k-nearest neighbors, as a self-organizing map, case-based reasoning, case-based learning, or as a so-called hidden Markov model. Combinations of such embodiments are possible.
- CNN deepening neural network
- an artificial intelligence in particular dimensionala le data such as image data or acoustic vibrations can be well processed.
- the artificial intelligence with several neural networks using different network types, for example at least one normal neural network with only one or two hidden layers of neurons and additionally at least one neural network of the type DNN or CNN.
- the latter neural networks are also suitable for traj ectorial control.
- a neural network comprises an input layer, an output layer and at least one hidden layer of neurons. Via the input layer, data can be transferred to the neural network; the output layer can be used to retrieve the answers and results of the neural network.
- Knowledge storage takes place in the hidden layers. Learning takes place through changes in the weights with which the neurons interact. Such learning is often referred to as training.
- a deep neural network is a neural network in which there are at least three hidden layers of neurons. This is in contrast to a normal or flat neural network with a maximum of two hidden layers. Often, even a single hidden layer is present in a normal neural network.
- a deep neural network allows the storage of complex relationships and thus the analysis of complex tasks.
- a folding neural network is another very common form of neural network. It is particularly suitable for processing dimensional data.
- Such a neural network consists of an input layer, several alternating convolutional and subsampling layers (hidden layers) and an output layer.
- the network is divided into the areas Input, Feature Extraction and Classification.
- the input data is convolved by means of a convolutional layer.
- the information from neighboring data points is bundled and passed on to the next layer.
- the corresponding data points of the input matrix are compared with the coefficients of a convolution kernel multiplied. From the sum of a new data point formed in the next layer. This results in several parallel convolutions that cause generalization. This results in several independent feature maps.
- a subsequent subsampling layer can further compress the feature maps, for example halve them.
- Each subsequent convolutional layer can then contain further feature maps.
- the last convolution or subsampling layer is fully connected to the neurons of a classification layer, ie each neuron is connected to each feature map.
- the output layer provides the result of the analysis.
- a folding neural network in contrast to a normal neural network, instead of the weights of the connections between the neurons, the coefficients of the convolution kernels are determined.
- a known backpropagation learning algorithm can be used.
- a folding neural network is a special case of a deep neural network. In particular, it makes use of the circumstance that relevant relationships and thus connections often only exist between locally and / or temporally closely adjacent quantities. Such structures are particularly advantageous for constructing increasingly abstract, localized representations of an image. In particular for the processing of dimensional input data - as explained above - such neural networks are advantageous.
- a description of a folding neural network can be found, for example, in the paper "Deep Learning” by Yann LeCun et al., Published in Nature, Vol. 521, 28.05.2015, pages 436 to 444.
- neural networks are the reduced time required for the learning process.
- very large neural networks which have, for example, 5, 8 or 10 hidden layers (or even more hidden layers) and have a total of 10 4 or 10 5 neurons, with a reasonable time in practice (usually only a few hours) trainable.
- RNN recurrent neural network.
- Such a neural network is suitable, in particular, for data which is available as a time sequence.
- the data received by the artificial intelligence and / or its reference in the dimension are unitless or are converted by the artificial intelligence into such data.
- dimensions in the strip width direction of a metal strip-preferably starting from the middle of the metal strip- can be normalized to the width of the metal strip. Similar approaches are possible for other dimensions of the metal strip.
- Bale center - be standardized to the bale length of the respective roller.
- locations within an image may be normalized to the overall dimensions of the particular image.
- the reference reference is preferably the center of the image.
- a gray level of a pixel value may be normalized to a value between 0 and 1 (or alternatively between -0.5 and +0.5).
- a setting range of an actuator can also be normalized.
- Other standards are possible. It is also possible to combine different quantities to determine unitless sizes. kidney. Examples are a standardization of chemical analyzes, temperatures, heat transfer and flows.
- This configuration makes it possible in a simple manner to transfer the knowledge stored in the artificial intelligence for a particular plant of the basic industry to another, similar plant of the basic industry
- the data received from the artificial intelligence to comprise at least one image by means of which the lateral position of a metal strip between two rolling stands of a multi-stand rolling train can be determined, and a difference manipulated variable of the upstream of the two rolling stands, and evaluation results determined by the Artificial Intelligence include a tape travel control strategy.
- the control of the lateral position is often difficult in the prior art and is often still done manually.
- the present invention remedy can be created by the present invention remedy.
- the data received from artificial intelligence can include dimensional data describing wear of a component of the plant of the basic industry and for the evaluation results determined by the artificial intelligence to include a prediction of a remaining lifetime of the component of the plant of the basic industry.
- the component can be, for example, a work roll of a rolling stand.
- the data taken by the artificial intelligence includes a roll bale contour as a function of the roll travel of the work roll and the stitch plan data of the roll stand. Based on these data, artificial intelligence can predict when to change the respective work roll (usually together with the other work roll of the respective stand, ie as a set of work rolls). This prediction can For example, be used for the adjustment and optimization of production planning.
- a quantity and / or a composition of an exiting the converter vessel exhaust gas as a function of time at least one stam from the area of the converter vessel ⁇ Mende acoustic vibration and / or its spectrum, at least one the area of a Konverterhutes included image (visible spectrum or infra-red),
- the evaluation results determined by the artificial intelligence provide a prediction of a probability, a time, a degree of expected slopping from the converter vessel, an expected final level of oxygen and / or carbon of the metal in the converter vessel, a a Abstichzeityak predicted temperature of the vessel located in Konverterge metal, a result of taking place in the converter vessel Entphosphorungsreaes and / or metallurgi see sizes during a blowing process within the converter vessel include.
- a computer program of the type mentioned in the introduction comprises dimensional data of an image output to the person via the human-machine interface, a temporal sequence of such images, a part of such an image or a temporal sequence of such a part.
- the image may be an optical or an infrared image.
- the images may be in the form of pixel arrays, 2D surfaces, or 3D volumes.
- a specification of local or temporal curves is possible.
- the dimensional data may include an acoustic vibration or a spectrum of such vibration.
- the processing of the machine code by the computing device causes the artificial intelligence implemented by the computing device to implement the above-explained advantageous embodiments of the operating method.
- the object is further achieved by a computing device having the features of claim 14.
- a computing device of the type mentioned above is programmed with a computer program according to the invention.
- the computing device may alternatively be a single
- FPGAs field programmable gate array
- An (alternative or additional) use of a TPU ( tensor processing unit) or several TPUs is also possible.
- the computing device will be in the area of the plant of the primary industry and will be permanently assigned to the plant. However, it is also conceivable to arrange the computing device completely or partially at a remote location, for example at a manufacturer of the basic industry plant or "somewhere" in a cloud.
- the object is further achieved by a plant of the basic industry with the features of claim 15.
- the system has a computing device according to the invention.
- FIG. 11 shows an integration of an artificial intelligence into a comprehensive data processing concept.
- a system ANL of the basic industry which is in principle arbitrary, has an automation system 1, sensor devices 2 and a human-machine interface 3.
- PDA Process Data Acquisition
- a PDA system saves recorded process data in the sense of a history, so that later on even earlier incurred process data can be accessed. As a result, in particular, for example, in the context of troubleshooting, a determination of the underlying cause can be made. Furthermore, all the process variables as well as the previous time histories are available.
- the sensor devices 2 acquire measured data M of the system ANL.
- the measurement data M may be individual values such as a singular temperature or a singular velocity or a singular force. Alternatively, it may be dimensional data such as a temporal temperature profile or a local temperature distribution. Other sizes are also possible.
- the measured data M acquired by means of the sensor devices 2 are at least partially supplied to the automation system 1, which receives the measured data M supplied to it.
- the measured data M acquired by means of the sensor devices 2 are furthermore transmitted at least partially to the human-machine interface 3.
- the human-machine interface 3 outputs the measured data M transmitted to it to a person 4.
- the person 4 may be, for example, an operator of the plant ANL of the basic industry, a fitter, an in-company or another person or belong to the maintenance personnel.
- the human-machine interface 3 furthermore receives data D from the automation system 1 and outputs it to the person 4. Furthermore, the human-machine interface 3 receives control commands S from the person 4. The received control commands S forwards the human-machine interface 3 to the automation system 1.
- the automation system 1 determines based on the information available to him control data S '.
- the control data S ' outputs the automation system 1 to controlled elements 5 of the system ANL.
- the system ANL is controlled and / or monitored by the automation system 1.
- the automation system 1 takes into account both the received measurement data M and the control commands S forwarded by the human-machine interface 3.
- the automation system 1 often takes into account internal data I as part of the determination of the control data S'. ie data that is available only within the automation system 1. This internal data 1 are (within the control process as such) received by the automation system 1 neither from the outside nor delivered to the outside. It may be, for example, data that is model-based determined by the automation system 1.
- the information by means of which the automation system 1 determines the control data S ' may include, for example, planning data P, which are transmitted to the automation system by a production planning system 11.
- the system ANL furthermore has a computing device 6.
- the computing device 6 is for the transmission of information to the automation system 1, the sensor devices
- the computing device 6 is programmed with a computer program 7.
- the computer program 7 comprises machine code 8, which can be processed by the computer 6.
- the processing of the machine code 8 by the computing device 6 causes the computing device 6 to implement an artificial intelligence 9.
- Artificial Intelligence 9 is designed as a neural network.
- Artificial Intelligence could also be designed as a support vector machine, as a decision tree, as a Bayesian reliability network, as a self-organizing map, as a case-oriented consideration, as a case-oriented learning, as a so-called hidden Markov model or as a so-called k-nearest neighbor.
- the Artificial Intelligence 9 works as follows:
- the artificial intelligence 9 receives - at least partially - against:
- the artificial intelligence 9 continues to receive from the man-machine interface 3 or the automation system 1, the control commands S. Utilizing the received data M, S ', I and
- the evaluation results A outputs the artificial intelligence 9.
- the artificial intelligence 9 can provide the evaluation results A via the human-machine interface 3 and / or an independent output device 10 of the person 4.
- the independent output device 10 may be designed, for example, as a smartphone or the like.
- the artificial intelligence 9 can make the evaluation results A available to the production planning system 11.
- the evaluation results A of the artificial intelligence 9 can be the reproduction of digitally stored knowledge. Digitally stored knowledge is taken to mean specialist knowledge that operators 4 have built up over many years, but which has not been previously documented, ie is (so far) only in the minds of operators 4.
- control commands S takes place directly.
- the specification of the control commands S via the man-machine Interface 3 done.
- the control commands S to be first offered to the person 4 and then to be validly connected or supplied to the automation system 1 when the person 4 activates the control commands S" by a confirmation command C.
- the artificial intelligence 9 outputs the evaluation results A only on specific request by the person 4.
- the artificial intelligence 9 does not normally output evaluation results A, but continuously and automatically checks whether it recognizes or predicts a suboptimal or even critical state of the plant based on the data supplied to it.
- the artificial intelligence 9 can output the evaluation results A together with an alarm or notification message to the person 4 in the event of detection of such a plant condition.
- the evaluation results A may also include a suggestion as to how the detected sub-optimal or even critical state of the system can be counteracted.
- the Artificial Intelligence 9 can thus - depending on its design and integration - implement an "artificial helmsman.”
- the "artificial helmsman” can act on the automation system 1 directly or indirectly via the human-machine interface 3.
- Artificial Intelligence 9 can act as a "digital assistant” to the (human) helmsman who points the helmsman to suboptimal / critical equipment conditions and / or makes suggestions to the helmsman for control commands that the helmsman can only implement, or possibly even only must confirm.
- the Artificial Intelligence 9 can act as a "digital expert" of the (human) helmsman, who provides the helmsman on request his knowledge.
- the data M, I received by the artificial intelligence 9 are at least partially dimensional data. This will be explained in more detail in conjunction with FIG 2 using several examples.
- the dimensional data may include, for example, a time profile of a measured value or of an internal value.
- the corresponding data is one-dimensional and the dimension is time.
- An example is a temperature profile.
- the temperature profile can be measured or calculated.
- Artificial Intelligence 9 the appropriate time course as a pure sequence of values.
- Artificial Intelligence 9 needs to know the time base.
- predefine the corresponding temporal course of the artificial intelligence 9 as a sequence of pairs of values, wherein for each value pair of one of the two values the respective time and the other value is the respective measured value M. It is even possible to specify the corresponding time course of the artificial intelligence 9 as an image, on the basis of which the artificial intelligence 9 determines the time course. This will become apparent from the following explanations.
- the dimensional data may also extend in at least two dimensions.
- this is shown in the illustration for a data field that extends in two local dimensions.
- An example of such a data field is the surface temperature of a metal strip, spatially resolved in the bandwidth direction and band length direction.
- the artificial intelligence 9 in the context of determining the evaluation results A for such data in particular curves equal value 12 (in the case of a temperature field so isotherms) or gradient 13 determine. It is also possible, as can be seen from the representation in FIG. 2 within the artificial intelligence 9, that to determine both the curves of the same value 12 and the gradients 13.
- the dimensional data can be an acoustic oscillation (ie the temporal
- the dimensional data may comprise a spectrum of such an acoustic oscillation.
- the dimensional data may also include a local profile of a measured value or of an internal value.
- the corresponding data is one-dimensional, but the dimension is the location.
- An example is a contour of a roll gap. The contour curve will usually be determined by calculation.
- the dimensional data may, for example, comprise an image captured by a sensor device 2 (or a part of such an image). As can be seen from the multiple representation of the image, the dimensional data may also comprise a temporal sequence of such an image or of such a part of an image.
- An example of such an image is an image showing a metal strip between two stands of a multi-stand rolling mill.
- an infrared image is advantageous here. It can also be an image in the visible spectrum.
- the dimensional data may comprise, for example, an image (or a part of such an image) output to the person 4 via the man-machine interface 3. Again, the evaluation of a sequence of such image or such a part of an image is possible again.
- artificial intelligence 9 can also be supplied with further data, for example data from a chemical analysis.
- the chemical analysis may include, for example, feedstocks (which chemical composition, for example, does the steel supplied to a converter?), Production results (which chemical composition, for example, does the steel have after a ladle process?), Or by-products (for example, which chemical composition has an effluent gas? ) be.
- the dimensional data may include not only measurement data M but also data calculated by the automation system 1 on the basis of a model.
- An example of such data is a temperature profile across the thickness inside a metal strip.
- the dimensional data comprise at least one image (be it an image acquired by a sensor device 2 or an image output via the human-machine interface 3), a part of such an image, a temporal sequence of such images, or a temporal sequence of a Part of such images or an acoustic oscillation or a spectrum of an acoustic oscillation.
- image be it an image acquired by a sensor device 2 or an image output via the human-machine interface 3
- a part of such an image a temporal sequence of such images, or a temporal sequence of a Part of such images or an acoustic oscillation or a spectrum of an acoustic oscillation.
- artificial intelligence 9 are fed exclusively such dimensional data.
- artificial intelligence 9 is also supplied with other, non-dimensional data.
- this data will also be evaluated by Artificial Intelligence 9.
- the following statements always refer to situations which either exclusively concern the dimensional data. or, in addition to non-dimensional data, also affect the dimensional data.
- the artificial intelligence 9 is preferably capable of performing a machine learning.
- the artificial intelligence 9 is thus preferably a (suitably programmed) computing device 6, which only sufficiently often a respective fact and a rating B of the respective facts must be specified and can independently determine rules from the large number of predetermined circumstances together with the associated rating B, such the best possible rating can be achieved.
- Machine learning is thus a method to learn from sample data. This approach is often referred to in the art as "programming by input-output examples rather than by coding.”
- Such artificial intelligence 9 is well known to those skilled in the art, and once such artificial intelligence 9 has learned its expertise, it can continue to learn.
- the artificial intelligence 9 determines the corresponding evaluation B of the evaluation results A in parallel with the determination of the evaluation results A.
- the artificial intelligence 9 can train itself on the basis of the evaluation B determined by it. In principle, however, the artificial intelligence 9 can also be designed differently.
- Artificial Intelligence 9 For the implementation of Artificial Intelligence 9 different possibilities are given. At the present time, as shown in FIGS. 3 and 4, it is preferred to construct Artificial Intelligence 9 as an artificial neural network.
- the artificial neural network may in particular, according to the illustration in FIG. 3, be a deep neural network or, as shown in FIG. 4, it may even be a folding neural network.
- a neural network as shown in FIGS. 3 and 4, has a number of input neurons e and a number of output neurons a. Zwisehen the entrance neuro- At least one layer s of hidden neurons n is located at the e and output neurons a.
- the neurons e, n, a are linked together in a manner known per se. Of the neurons e, n, a are only a few provided with their reference numerals in Figures 3 and 4.
- a deep neural network as shown in FIG. 3, is a neural network having a plurality of layers s (layers) of hidden neurons n.
- the number of layers s with hidden neurons n is at least three. It can also be larger, for example 5, 8 or 10. As a rule, 15 layers are completely sufficient.
- a convolutional neural network is a deep neural network having an input layer (input), a plurality of convolutional and subsampling layers (hidden layers) and an output layer (output or classification).
- the convolutional and subsampling layers represent the actual intelligence in the narrower sense, which extracts the relevant facts from the information provided via the input layer.
- the output layer provides the result of the analysis.
- the complexity of the neural network compared to a deep neural network can be significantly reduced without reducing the performance of the neural network related to the respective application.
- the concrete representation in FIG. 4 is used to identify, based on an image, which digit is shown in the image. However, the principle of FIG. 4 can also be applied to other circumstances if the folding neural network is designed accordingly.
- Unit-related quantities are quantities to which a physical unit belongs. ordered, for example, the unit meters, the unit meters per second, Newton, Newton / mm 2 and more.
- the data received by the artificial intelligence 9 and / or its reference in the dimension are unitless or are converted by the artificial intelligence 9 into such data. This process will be explained in more detail with reference to FIG 5.
- FIG. 5 shows a locally one-dimensional variable, namely a contour of a work roll of a rolling mill over the bale length L.
- the length of the bale L is-in principle arbitrarily-assumed to be exactly 2.00 m.
- the location on the roll bale is plotted, upwards the crown, i. the deviation 5R from a base radius R.
- This type of representation is not uniform.
- the crowning 5R must be related to the basic radius R and the location on the roll barrel to the bale length L.
- the data accepted by the artificial intelligence 9 include, inter alia a profile of a flat rolled stock before execution of a plurality of rolling passes, with which a flat rolling stock is rolled successively,
- the rolling stock may alternatively be a heavy plate or a strip.
- the number of rolling passes is generally between 3 and 20, for example between 4 and 7, in particular 5, 6 or 7.
- the stitches are carried out in a multistage finishing mill.
- the rule - each rolling pass is executed by its own rolling stand.
- the reversing stand or in some cases a pair of reversing stands) performs the stitches. In this case, therefore, several rolling passes are carried out by a rolling stand
- the artificial intelligence 9 is given the contour of the roll gap and / or the roll bales of the work rolls of the roll stand which carries out the respective roll pass in a spatially resolved manner over the rolling stock width.
- the spatial resolution is such that viewed over the Walzgutbreite the roll gap or the respective radius (or diameter) of the two work rolls over the Walzgutbreite for at least 5 places is specified.
- a default is made for at least 10 digits. In the simplest case, the default is static. However, it can be defined as a function of time. In the last-mentioned case, it is possible, in particular, to associate the respective contour or the respective roll bales via a path tracking of the rolling stock to specific points of the finished rolled rolled stock.
- the stitch plan data in particular comprise the respective nominal roll gap, the respective rolling force, possibly the respective return bending force, possibly a respective wedge setting and a respective work roll displacement.
- the stitch plan data include the static and dynamic data of the rolling stock.
- the static data of the rolling stock comprise at least its chemical composition, possibly also its width.
- the dynamic data comprise (in up to 3 dimensions spatially resolved and / or temporally resolved) the temperature of the rolling stock during the execution of the respective rolling pass as well as the thickness of the rolling stock before and after the respective rolling pass and possibly additionally the width of the rolling stock before and after the respective rolling pass.
- the artificial intelligence 9, the initial profile of the rolling stock - ie before the execution of rolling passes - specified.
- the initial profile is spatially resolved at least across the width. It may additionally be spatially resolved over the length of the rolling stock.
- the resolution over the width of the rolling stock is such that, viewed over the rolling stock width, the respective thickness of the rolling stock is predetermined for at least 5 places. Preferably, a default is made for at least 10 digits. Viewed over the length of the rolling stock, either no spatial resolution or at least one occurs
- Spatial resolution for the head, the center piece and the foot of the rolling stock In particular, in the case of a strip, it is still possible to make a spatial resolution over the length of the rolling stock over many (often more than 100) sections.
- the individual sections may in this case be determined, for example, by a uniform length, a uniform mass or a working cycle.
- the initial profile of the rolling stock can be given to the artificial intelligence 9 in particular by an image which is displayed to the operator 4 via the man-machine interface 3.
- the flatness and / or the profile of the finished rolled rolled stock are given to the artificial intelligence 9 with spatial resolution at least over the width of the rolling stock, preferably also over the length of the rolled stock.
- the evaluation results A determined by the artificial intelligence 9 may include, for example, a strategy S1 for displacing the work rolls and / or intermediate rolls of the rolling stands executing the respective rolling passes for a subsequent rolling operation of another flat rolled stock.
- the determined displacements can then be used for the next rolling process - possibly for the next similar rolling process.
- the determination of the displacement takes place separately for each rolling pass.
- the rolling stock is usually a band.
- the rolling of the strip is usually done in a tandem mill in which the strip is cold rolled.
- the number of rolling passes is usually between 3 and 8, for example between 4 and 7, in particular 5 or 6.
- the turn list data include in particular for each rolling pass the respective nominal rolling gap, the respective rolling force, possibly a respective return force, possibly one Wedge adjustment and possibly a work roll shift. They also include static and dynamic data of the rolling stock for each rolling pass.
- the static data of the rolling stock may comprise, for example, its chemical composition and possibly also its width.
- the dynamic data include the thickness of the rolling stock before and after the respective rolling pass and possibly also the width of the rolling stock before and after the respective rolling pass.
- the artificial intelligence 9 is given the contour of the roll gap and / or the roll bales of the work rolls of the roll stand carrying out the respective roll pass in a spatially resolved manner over the rolling stock width.
- the spatial resolution is such that, viewed over the rolling stock width, the roll gap or the respective radius (or diameter) of the two work rolls is predetermined over the rolling stock width for at least 5 places. Preferably, a default is made for at least 10 digits.
- Artificial Intelligence 9 will become the profile and / or planarity of the flat
- Artificial Intelligence 9 determines during operation - ie during rolling of the rolling stock a control for a segmented work roll cooling of at least one of the rolling stands, a set displacement of the work rolls of at least one of the rolling stands and / or a set deflection of the work rolls of at least one of the rolling stands.
- the Artificial Intelligence 9 thus implements a strategy S2 for segmented cooling of rolls of at least one of the rolling mills and / or a work roll shifting strategy S3 and / or a work roll bending strategy S4.
- the strategies S2, S3 and / or S4 thus represent an intelligent controller.
- the data received by the artificial intelligence 9 include, among other things, at least one image.
- the picture shows the area between two successive stands 15 of a multi-stand rolling train.
- the image (visible spectrum or, preferably, infrared) need not necessarily include the rolling stands 15. It is only necessary to be able to determine the lateral position of the metal strip 14 from the image.
- the artificial intelligence 9 is supplied with a difference manipulated variable of the upstream of the two rolling stands 15.
- the differential control variable may be, for example, a differential rolling force or a differential rolling gap.
- the evaluation results A determined by the artificial intelligence 9 can be a threading-in control strategy S5 of the metal band 14.
- the strategy S5 can be determined in such a way that the metal strip 14 enters the downstream rolling stand 15 in the center and possibly without any transverse component of the speed. This procedure makes it possible in particular to avoid high-altitude walkers.
- the evaluation results A determined by the artificial intelligence 9 may include a strategy S6 for controlling the meandering during the "normal" rolling of the metal strip 14.
- the artificial intelligence 9 is further supplied with a difference train, if possible which prevails in the metal strip 14 between the two rolling stands 15.
- the differential train can be detected, for example, by means of a loop lifter which is arranged between the two rolling stands 15.
- the corresponding procedure can also be carried out simultaneously for several such rolling stands at the same time.
- the artificial intelligence 9 for each rolling stand in which the appropriate procedure is carried out, the corresponding data (image, difference control variable and possibly Differenzzug) supplied.
- the artificial intelligence 9 can in particular perform a rolling stand-spanning determination of the difference manipulated variable.
- the data received by the artificial intelligence 9 include, inter alia, dimensional data describing wear of a component of the plant ANL of the basic industry.
- the evaluation results A determined by the artificial intelligence 9 can include a prediction VI of a remaining lifetime of the component of the plant ANL of the basic industry.
- the component may be a work roll of a rolling stand.
- the data received from the artificial intelligence 9 comprises a roll bale contour as a function of the roll travel of the work roll as well as the stitch plan data of the roll stand.
- roll travel has a fixed significance for the person skilled in the art: It is the total distance traveled since the roll has been installed in the roll stand on the rolling stock (or another rolling stock)
- the roll bale contour is preferably broken down into a basic contour, supplemented by a contour change by backbending and a contour change by temperature effects and wear 9 predicted when to change the respective work roll (or roll set, respectively) .This prediction can be used, for example, to tailor and optimize production planning, particularly in continuous rolling and semi-continuous rolling.
- the data received by the artificial intelligence comprise at least one of the following dimensional data:
- the substance may be, for example, a gas (oxygen, air, nitrogen, argon,...) Or a solid such as lime, ore or others.
- a gas oxygen, air, nitrogen, argon, etc.
- a solid such as lime, ore or others.
- it may be sufficient to specify only individual times and the amount of solid fed in each case instead of a continuous time course.
- a position of a lance in the converter vessel as a function of time may include the location as such and / or the orientation of the lance.
- a cooling water temperature of the converter as a function of time.
- At least one coming from the region of the converter vessel acoustic vibration and / or their spectrum.
- the acoustic vibration can be detected for example by means of a conventional microphone (sound, noise) or as structure-borne noise (acoustic body vibration).
- At least one image containing the area of a converter hatch (visible spectrum or, preferably, infrared), often even a temporal sequence of such images.
- the image of the converter hat can contain, for example, the flame image of the converter. From the flame pattern conclusions can be drawn about the progress of the process taking place in the converter vessel, similar to the quantity and composition of the exhaust gas.
- these data may be locally spatially resolved data.
- the spatial resolution can in particular be two- or three-dimensional.
- artificial intelligence 9 is supplied with several of the above-mentioned data. However, not all data have to be supplied to Artificial Intelligence 9. Furthermore, the artificial intelligence 9 are often also supplied with other, non-dimensional data such as the so-called scrap rate, the type of scrap, the amount of scrap and the chemical analysis of the scrap, comparable information including the temperature of the pig iron, data on a subjective assessment of expected and / or or already made by the operator 4 ejection behavior or other boundary conditions such as humidity, ambient temperature and more. Which data is actually fed into artificial intelligence 9 can vary from case to case.
- the evaluation results A determined by the artificial intelligence 9 in this case may include a prediction V2 to V4 about a probability, a time and / or an extent of an expected ejection from the converter vessel.
- a prediction V2 to V4 about a probability, a time and / or an extent of an expected ejection from the converter vessel.
- the artificial intelligence 9 can, for example, compare the quantities supplied to it. evaluate that it proposes control and regulation commands or possibly even directly implemented, for example, to avoid converter ejection.
- the control and regulation commands may include, for example, an adjustment of the lance position, a volume or mass flow of oxygen, timing and amount of addition of solid additives and the like.
- the addition of solid aggregates such as lime or dolomite and the like, which are used for dephosphorization, takes place only in the lower temperature range of the process of the converter (up to a maximum of approx.
- the task for Artificial Intelligence 9 may be based on (among other things) the dimensional data - i. Among other things, based on the originating from the area of the converter vessel acoustic vibrations and / or their spectrum and / or based on the converter hat showing image or the corresponding sequence of images - to determine the right time for the addition of the right amount of aggregate and pretend the operator 4. The determination can be made, for example, based on the time profile of the measured or model-based determined mass flow of carbon in the exhaust gas, the temperature profile as a function of time (calculated or measured) and / or the time course of at least one vibration spectrum. In an analogous manner, the artificial intelligence 9 can also determine other variables, such as a specification for the lance position or the supply of oxygen. Known time profiles for the melt of similar boundary conditions can be used in addition to increasing the probability of the statement made by artificial intelligence 9.
- the artificial intelligence 9 can be used to make target sizes in temperature and chemical composition of the molten metal at the time of tapping more reproducible, to avoid Nachblasraten and the like.
- a target window for the temperature and the carbon content for the time point of tapping can be achieved better.
- the time course of the mass flow of oxygen, the time course of the position of the lance in conjunction with a measurement of the oxygen content and the temperature measurement can be utilized.
- a reclassification of a melt (due to a non-achieved chemical composition) or a return to the pig iron mixer or the converter (with less deviation of the actual chemical composition of the desired) can be avoided.
- a transformer temperature as a function of time a state of the supply network as a function of time, a cooling water temperature of the arc furnace as a function of time,
- At least one image containing the area of the upper side of the arc furnace (visible spectrum or, preferably, infrared, for example for evaluating a flame image, in particular as a function of time),
- artificial intelligence 9 is supplied with several of the above-mentioned data. However, not all data have to be supplied to Artificial Intelligence 9. Furthermore, the artificial intelligence 9 are often also supplied with other, non-dimensional data such as the so-called scrap rate, the type of scrap, the amount of scrap and the chemical analysis of the scrap, comparable information including the temperature of the pig iron, data on a subjective assessment of expected and / or or foaming behavior of the slag in the electric arc furnace by the operator 4 or others has already taken place
- Boundary conditions such as humidity, ambient temperature and more. Which data is actually supplied to Artificial Intelligence 9 can vary from case to case.
- control and regulation commands may include, for example, an adjustment of the electrode position, a switching stage of the furnace transformer, an operating voltage or an operating current and the like. Also, a supply of a fuel gas (for example, natural gas) and / or oxygen is possible.
- a fuel gas for example, natural gas
- THD Total Harmonie Distortion
- artificial intelligence 9 it is also possible for artificial intelligence 9 to supply dimensional data (and, in addition, often non-dimensional data) across systems, for example data from a blast furnace or an electric arc furnace as well as data from a ladle furnace as well as data from a continuous casting plant as well as data from one rolling mill. As a result, artificial intelligence 9 can even be used to optimize plant operation across plants. It is also possible for Artificial Intelligence 9 in addition to the dimensional data (and possibly also non-dimensional data) of the respective plant of the primary industry to supply comparable data to another plant of the basic industry and additionally to have these evaluated by Artificial Intelligence 9.
- An overall intelligence 16 may include a single Artificial Intelligence 9 or, as shown in FIG. 11, multiple Artificial Intelligents 9.
- the individual artificial intelligences 9 can be designed as explained above. In principle, any desired combinations of different embodiments of artificial intelligence 9 are possible.
- the embodiment of artificial intelligence 9 shown in FIG. 11 as neural networks is therefore only an example.
- the artificial intelligences 9 can be arranged parallel to one another in accordance with FIG. Additionally, within the overall intelligence 16, there may be 16 other components 17, such as a model or rule-based system. It is possible to pre-process the data (this not only, but also for the dimensional data) before feeding it to the total intelligence 16 in a preprocessing device 18.
- the preprocessing can be, for example, a filtering or a static or dynamic averaging. Also, an error correction, normalization and / or structuring of the input data can take place.
- the total intelligence 16 can be arranged downstream of an evaluation device 19. In this case, based on the results determined by means of the total intelligence 16, a further evaluation is carried out. As such, this evaluation does not have to meet the requirements placed on Artificial Intelligence.
- the control of the overall arrangement of FIG. 11 is effected by a data flow control device 21.
- An automation system 1 determines control data S ', outputs these to controlled elements 5 of the system ANL and thereby controls the system ANL.
- Sensor devices 2 acquire measurement data M of the system ANL, which they supply at least partially to the automation system 1 and to a human-machine interface 3.
- the man-machine interface 3 furthermore receives planning data (P) from a production planning system (11) and / or from the automation system 1 receives control data S 'and / or internal data I. It outputs the data M, S ', I to a person 4. Furthermore, it receives control commands S from the person 4 and forwards them to the automation system 1.
- the automation system 1 processes the measurement data M and the control commands S as part of the determination of the control data S '.
- An artificial intelligence 9 at least partially accepts the measurement data M, the control data S 'and / or the internal data I and the data output to the person 4. It also accepts the control commands S.
- the data M, S ', I received by the artificial intelligence 9 are at least partially dimensional data the man-machine interface 3 output a picture, a part of such an image, a temporal sequence of such images or a temporal sequence of a part of such images or an acoustic oscillation or a spectrum of an acoustic oscillation.
- the present invention has many advantages. In particular, the robustness of the plant operation is increased. Productivity and product quality are guaranteed at a high level.
- the manual control actions of person 4 required in the prior art are reduced. in the Ideally, they can even be completely eliminated.
- the procedure according to the invention additionally reduces the psychological burden on the person 4. There is therefore a lower risk of incorrect operation by the person 4.
- the invention is not only feasible for new systems ANL, but can also be retrofitted to existing systems ANL. It can no longer happen that knowledge about the operation and operation of the plant ANL of the basic industry is lost through a generational change of the operating personnel. This also applies to knowledge that is accumulated in the prior art in a few experienced operators 4, but otherwise is not documented. For example, commissioning a new ANL system is often faster. Furthermore, employees with less knowledge of the ANL system can also be used. Furthermore, employees can also be trained faster. The time within which it responds to changing operating conditions is often shortened. Finally, personnel expenses are significantly reduced overall.
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Quality & Reliability (AREA)
- Manufacturing & Machinery (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Mechanical Engineering (AREA)
- Chemical & Material Sciences (AREA)
- Chemical Kinetics & Catalysis (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
Description
Claims
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| EP16188584.3A EP3293594A1 (de) | 2016-09-13 | 2016-09-13 | Verwendung umfassender künstlicher intelligenz bei anlagen der grundstoffindustrie |
| PCT/EP2017/071827 WO2018050438A2 (de) | 2016-09-13 | 2017-08-31 | Verwendung umfassender künstlicher intelligenz bei anlagen der grundstoffindustrie |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| EP3513263A2 true EP3513263A2 (de) | 2019-07-24 |
Family
ID=57144731
Family Applications (2)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| EP16188584.3A Withdrawn EP3293594A1 (de) | 2016-09-13 | 2016-09-13 | Verwendung umfassender künstlicher intelligenz bei anlagen der grundstoffindustrie |
| EP17762085.3A Ceased EP3513263A2 (de) | 2016-09-13 | 2017-08-31 | Verwendung umfassender künstlicher intelligenz bei anlagen der grundstoffindustrie |
Family Applications Before (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| EP16188584.3A Withdrawn EP3293594A1 (de) | 2016-09-13 | 2016-09-13 | Verwendung umfassender künstlicher intelligenz bei anlagen der grundstoffindustrie |
Country Status (4)
| Country | Link |
|---|---|
| US (1) | US11294338B2 (de) |
| EP (2) | EP3293594A1 (de) |
| CN (1) | CN109874338A (de) |
| WO (1) | WO2018050438A2 (de) |
Families Citing this family (23)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3575890B1 (de) * | 2018-05-28 | 2022-03-30 | Siemens Aktiengesellschaft | Assistenzsystem zur unterstützung der planung von automatisierungssystemen |
| DE102019103012A1 (de) * | 2019-02-07 | 2020-08-13 | Thyssenkrupp Ag | Verfahren zur Produktion eines Metallwerkstücks |
| JP2020131248A (ja) * | 2019-02-21 | 2020-08-31 | Jfeスチール株式会社 | 圧延荷重予測方法、圧延荷重予測装置、及び圧延制御方法 |
| EP3705965A1 (de) * | 2019-03-04 | 2020-09-09 | Siemens Aktiengesellschaft | Bildbasierte systemüberwachung |
| DE102019209110A1 (de) * | 2019-06-24 | 2020-12-24 | Sms Group Gmbh | Industrielle Anlage, insbesondere Anlage der metallerzeugenden Industrie oder der Aluminium- oder Stahlindustrie und Verfahren zum Betreiben einer industriellen Anlage, insbesondere einer Anlage der metallerzeugenden Industrie oder der Aluminium- oder Stahlindustrie |
| WO2021048984A1 (ja) * | 2019-09-12 | 2021-03-18 | 東芝三菱電機産業システム株式会社 | 絞り発生予測システム |
| CN110909508B (zh) * | 2019-10-29 | 2024-11-26 | 中国石油化工股份有限公司 | 基于卷积长短期记忆网络的加热炉温度场实时预测方法 |
| EP3825789A1 (de) * | 2019-11-20 | 2021-05-26 | Primetals Technologies Germany GmbH | Fernsteuerung einer anlage zum herstellen und/oder behandeln eines walzguts aus metall |
| JP2021089526A (ja) * | 2019-12-03 | 2021-06-10 | 株式会社Preferred Networks | 推定装置、訓練装置、推定方法、訓練方法、プログラム及び非一時的コンピュータ可読媒体 |
| DE102020102863A1 (de) * | 2020-02-05 | 2021-08-05 | Festo Se & Co. Kg | Parametrierung einer Komponente in der Automatisierungsanlage |
| EP3862104A1 (de) | 2020-02-06 | 2021-08-11 | Primetals Technologies Germany GmbH | Vorrichtung zur steuerung und/oder zur überwachung einer technischen anlage |
| EP4117836A1 (de) * | 2020-03-09 | 2023-01-18 | SMS Group GmbH | Verfahren zur reduzierung von prozessstörungen bei der herstellung eines walzproduktes |
| CN111651931B (zh) * | 2020-05-19 | 2022-05-17 | 浙江大学 | 基于深度神经网络的高炉故障诊断规则导出方法 |
| EP4102311B1 (de) * | 2021-06-11 | 2024-01-24 | Tata Consultancy Services Limited | Verfahren und system zum adaptiven lernen von modellen in fertigungssystemen |
| CN113380334B (zh) * | 2021-07-03 | 2024-01-12 | 河津市康庄新材料有限公司 | 一种钢渣游离氧化钙的快速消解方法及系统 |
| EP4124398B1 (de) * | 2021-07-27 | 2024-04-10 | Primetals Technologies Austria GmbH | Verfahren zur bestimmung mechanischer eigenschaften eines walzgutes mit hilfe eines hybriden modells |
| DE102021210234A1 (de) * | 2021-09-15 | 2023-03-16 | Sms Group Gmbh | System, Verfahren und Computerprogramm zur Optimierung eines Herstellungsprozesses in einer industriellen Produktionsanlage, insbesondere einer Anlage der metallerzeugenden Industrie, der nicht-Eisen- oder Stahlindustrie oder der Vorlegierungsherstellung |
| TWI824834B (zh) * | 2022-11-22 | 2023-12-01 | 國立高雄科技大學 | 連續輥軋製程檢測方法 |
| EP4647188A1 (de) * | 2023-04-06 | 2025-11-12 | JFE Steel Corporation | Verfahren zur festlegung eines kaltwalzzustands, kaltwalzverfahren, verfahren zur herstellung eines kaltgewalzten stahlblechs, vorrichtung zur berechnung eines kaltwalzzustands und kaltwalzwerk |
| DE102023124815B3 (de) | 2023-09-14 | 2025-02-27 | Röwaplan Ag | System zur fortlaufenden akustischen Überwachung von wenigstens einer technischen Anlage |
| CN117312388B (zh) * | 2023-10-08 | 2024-03-19 | 江苏泰赋星信息技术有限公司 | 一种人工智能模型控制系统 |
| CN117406668B (zh) * | 2023-12-14 | 2024-02-23 | 邢台金昆冶金机械有限公司 | 基于数值分析的轧辊质量控制数据模型及其应用 |
| WO2025137778A1 (en) * | 2023-12-29 | 2025-07-03 | Stas Inc. | Aluminum manufacturing process with retroaction loop |
Family Cites Families (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5930136A (en) * | 1990-06-04 | 1999-07-27 | Hitachi, Ltd. | Control device for controlling a controlled apparatus, and a control method therefor |
| JPH04231106A (ja) | 1990-12-28 | 1992-08-20 | Nippon Steel Corp | 帯鋼形状自動認識制御装置 |
| JPH05318222A (ja) | 1992-05-15 | 1993-12-03 | Toshiba Corp | 圧延材端部切断制御装置 |
| US5566092A (en) | 1993-12-30 | 1996-10-15 | Caterpillar Inc. | Machine fault diagnostics system and method |
| DE19547010C2 (de) | 1994-12-19 | 2001-05-31 | Siemens Ag | Verfahren und Vorrichtung zur Überwachung des Prozeßablaufs bei der Strahlerzeugung nach dem Sauerstoffaufblasverfahren |
| DE19508476A1 (de) * | 1995-03-09 | 1996-09-12 | Siemens Ag | Leitsystem für eine Anlage der Grundstoff- oder der verarbeitenden Industrie o. ä. |
| DE19508474A1 (de) | 1995-03-09 | 1996-09-19 | Siemens Ag | Intelligentes Rechner-Leitsystem |
| DE19731980A1 (de) * | 1997-07-24 | 1999-01-28 | Siemens Ag | Verfahren zur Steuerung und Voreinstellung eines Walzgerüstes oder einer Walzstraße zum Walzen eines Walzbandes |
| US7526463B2 (en) | 2005-05-13 | 2009-04-28 | Rockwell Automation Technologies, Inc. | Neural network using spatially dependent data for controlling a web-based process |
| JP4427074B2 (ja) * | 2007-06-07 | 2010-03-03 | 株式会社日立製作所 | プラントの制御装置 |
| DE102007035283A1 (de) * | 2007-07-27 | 2009-01-29 | Siemens Ag | Verfahren zur Einstellung eines Zustands eines Walzguts, insbesondere eines Vorbands |
| DE102008028777A1 (de) * | 2008-06-17 | 2009-12-24 | Siemens Aktiengesellschaft | Leitsystem einer Anlage mit mehrstufiger Modelloptimierung |
| EP2527053A1 (de) | 2011-05-24 | 2012-11-28 | Siemens Aktiengesellschaft | Steuerverfahren für eine Walzstraße |
| EP2701020A1 (de) * | 2012-08-22 | 2014-02-26 | Siemens Aktiengesellschaft | Überwachung einer ersten Ausrüstung einer technischen Anlage zur Herstellung eines Produkts |
-
2016
- 2016-09-13 EP EP16188584.3A patent/EP3293594A1/de not_active Withdrawn
-
2017
- 2017-08-31 US US16/332,873 patent/US11294338B2/en active Active
- 2017-08-31 EP EP17762085.3A patent/EP3513263A2/de not_active Ceased
- 2017-08-31 WO PCT/EP2017/071827 patent/WO2018050438A2/de not_active Ceased
- 2017-08-31 CN CN201780056211.9A patent/CN109874338A/zh active Pending
Also Published As
| Publication number | Publication date |
|---|---|
| WO2018050438A3 (de) | 2018-05-31 |
| WO2018050438A2 (de) | 2018-03-22 |
| US11294338B2 (en) | 2022-04-05 |
| CN109874338A (zh) | 2019-06-11 |
| US20190361409A1 (en) | 2019-11-28 |
| EP3293594A1 (de) | 2018-03-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP3513263A2 (de) | Verwendung umfassender künstlicher intelligenz bei anlagen der grundstoffindustrie | |
| DE4008510C2 (de) | Regeleinrichtung mit Optimal-Entscheidungseinheit | |
| AT414316B (de) | Verfahren und einrichtung zur steuerung einer hüttentechnischen anlage | |
| DE19508476A1 (de) | Leitsystem für eine Anlage der Grundstoff- oder der verarbeitenden Industrie o. ä. | |
| EP4150419B1 (de) | System und verfahren zur steuerung einer aus mehreren anlagenteilen bestehenden produktionsanlage, insbesondere einer produktionsanlage zur erzeugung von industriegütern wie metallischem halbzeug | |
| EP2428288B1 (de) | Verfahren zum Herstellen von Stahlbändern durch Endloswalzen oder Semi-Endloswalzen | |
| EP2697001A1 (de) | Steuerverfahren für eine walzstrasse | |
| EP0813700A1 (de) | Intelligentes rechner-leitsystem | |
| WO2004076086A2 (de) | Verfahren zur regelung der temperatur eines metallbandes, insbesondere in einer fertigstrasse zum walzen von metallwarmband | |
| EP1596999B1 (de) | Verfahren zur regelung der temperatur eines metallbandes, insbesondere in einer kühlstrecke | |
| EP2603333A1 (de) | Echtzeit-ermittlungsverfahren für temperatur und geometrie eines metall-warmbandes in einer fertigstrasse | |
| DE19618995A1 (de) | Verfahren und Einrichtung zur Beeinflussung des Profils eines Walzbandes | |
| WO2022106707A1 (de) | VERFAHREN ZUR EINSTELLUNG DER EIGENSCHAFTEN EINES WARMBANDES MIT EINER BESTIMMTEN CHEMISCHEN ZUSAMMENSETZUNG IN EINER WARMBANDSTRAßE | |
| DE19623671C2 (de) | Verfahren und Einrichtung zum Führen eines Prozesses in der Grundstoffindustrie | |
| DE102020201215A1 (de) | Verfahren zum Betreiben einer industriellen Anlage | |
| EP4630958A1 (de) | Verfahren und system zur computerimplementierten gestaltung und/oder (weiter)entwicklung und/oder auslegung und/oder lebendaueroptimierung von maschinenelementen einer hüttenechnischen anlage | |
| EP1481742B1 (de) | Steuerrechner und rechnergestütztes Ermittlungsverfahren für eine Profil- und Planheitssteuerung für eine Walzstrasse | |
| EP1014239A1 (de) | Verfahren zur Berechnung eines Stichplanes | |
| EP4320489B1 (de) | Optimierungsverfahren für den betrieb von anlagen der grundstoffindustrie | |
| EP1030747B1 (de) | Verfahren und einrichtung zum warmwalzen dünner stahlbänder | |
| EP4107594A1 (de) | Verfahren zur automatisierung einer hüttentechnischen anlage, insbesondere einer anlage zum walzen von metallbändern | |
| EP3862104A1 (de) | Vorrichtung zur steuerung und/oder zur überwachung einer technischen anlage | |
| DE102022212627A1 (de) | Verfahren zum Herstellen eines Stahlbandes aus einem Vorprodukt, bei dem die Sollwerte über die Länge eines einzelnen Stahlbandes und / oder zeitlich in Bezug auf eine einzelne Produktionsanlage einer Walzstraße variabel vorgegeben werden | |
| EP4130895A1 (de) | Verfahren zum ermitteln eines steuerparameters zum steuern einer walzanlage | |
| EP4311606A1 (de) | Verfahren zur regelung einer walzstrasse sowie walzstrasse |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: UNKNOWN |
|
| STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: THE INTERNATIONAL PUBLICATION HAS BEEN MADE |
|
| PUAI | Public reference made under article 153(3) epc to a published international application that has entered the european phase |
Free format text: ORIGINAL CODE: 0009012 |
|
| STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: REQUEST FOR EXAMINATION WAS MADE |
|
| 17P | Request for examination filed |
Effective date: 20190415 |
|
| AK | Designated contracting states |
Kind code of ref document: A2 Designated state(s): AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR |
|
| AX | Request for extension of the european patent |
Extension state: BA ME |
|
| DAV | Request for validation of the european patent (deleted) | ||
| DAX | Request for extension of the european patent (deleted) | ||
| STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: EXAMINATION IS IN PROGRESS |
|
| 17Q | First examination report despatched |
Effective date: 20211007 |
|
| REG | Reference to a national code |
Ref country code: DE Ref legal event code: R003 |
|
| STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: THE APPLICATION HAS BEEN REFUSED |
|
| 18R | Application refused |
Effective date: 20250128 |