Method and system for adapting a manufacturing process

Abstract

The invention relates to a method (100) and a system (1) for adapting a manufacturing process (P), in particular a heat rolling process, and to the use of the method (100) for a conventional operation of a facility (10). According to the method, i) a product property (E*) of a process product (2) is specified (S1), ii) an operating starting point (X) of a facility (10), in particular a hot rolling facility, which is designed to carry out the manufacturing process (P) in order to produce the process product (2), is provided (S2), and iii) the process product (2) is produced (S3) by carrying out the manufacturing process (P) by means of the facility (10) starting from the specified operating starting point (X). According to the invention, iv) an actual product property (E) of the produced process product (2) is ascertained (S4), and v) an optimized operating point (X*) of the facility (10) is ascertained (S5) on the basis of the specified product property (E*) and the ascertained actual product property (E) using an optimizing unit (30).

Description

[0001] Description

[0002] Method and system for adapting a manufacturing process

[0003] field of technology

[0004] The present invention relates to a method and a system for adapting a manufacturing process, in particular a hot rolling process.

[0005] State of the art

[0006] When commissioning plants, for example hot rolling mills, it is usually necessary to correctly adjust various parameters of the plant in order to be able to obtain a product with the desired properties during later operation of the plant.

[0007] For this purpose, it is known to create a model of the plant and calculate it using a target value that represents the desired product properties. The input variables of the model are expediently assumed external influences, for example in the form of parameterized environmental conditions, and model parameters (so-called commissioning parameters, or IBS parameters for short, which control the model behavior or parameterize control curves). Such a model can output manipulated variables with which the plant can be operated. If necessary, at least individual model parameters can be optimized by operating the plant several times with the initially calculated manipulated variables and comparing the properties of the resulting product with the desired properties.

[0008] For example, WO 2018 / 192798 A1 discloses an optimization method for a model of a plant in a basic materials industry. Using the model and specified target values ​​for an output product, a control device determines the corresponding desired target operation of the plant (setup calculation). The model is parameterized with a large number of key parameters (= model parameters) of the plant to be modeled. After a large number of output products have been manufactured, the actual values ​​of the output products are compared with the expected values ​​of the output products. Based on this comparison, the first model parameters are recalculated and the model reparameterized accordingly. The expected values ​​are determined using the model based on the actual operation of the plant.

[0009] The problem here, however, is that once the plant has been commissioned based on the selected model input variables, it is usually exposed to changing external influences and / or the same control variables lead to different product properties over longer periods of time, for example due to aging / wear of the plant. Furthermore, there are also manufacturing processes that are too complex for sufficiently precise modeling or whose modeling is extremely time-consuming or computationally intensive. If a sufficiently precise model cannot be found or used, the commissioning parameters cannot be automatically optimized using an algorithm. In such cases, the commissioning parameters must be determined manually. However, this requires considerable experience from the relevant personnel and is also very time-consuming.

[0010] For optimized control of thickness quality in a rolling process, it is known from EP 1 488 863 A2 to provide an online process model that adapts model parameters, including by identifying variable process parameters. Using an adaptation unit that communicates with the online process model, the control parameters are adapted based on the adapted model parameters. The adapted control parameters are then applied online to a model-based multivariable control unit for implementing and adapting control functions and / or loops.

[0011] DE 195 08 476 A1 discloses a control system for a plant in the basic materials or processing industry. Based on input prior knowledge, the control system is computer-based and designed to automatically detect the plant's status and details of a manufacturing process occurring within the plant and to issue situation-appropriate instructions, such as setting values, to achieve reliable production success. The situation-appropriate instructions are optimized in predefined optimization routines, with the input prior knowledge being continuously improved by knowledge acquired internally through computer-based calculations from the process model during production.

[0012] Summary of the invention

[0013] Against this background, it is an object of the present invention to provide an improved method and an improved system with which even complex manufacturing processes can be automatically adapted and / or properties of process products can be kept constant over long periods of time.

[0014] These objects are achieved by a method and a system for adapting a manufacturing process according to the independent claims.

[0015] Preferred embodiments are subject of the dependent claims and the following description.

[0016] In the method according to the first aspect of the invention for, in particular automatically, adapting a manufacturing process, in particular a hot rolling process, i) a product property of a process product is specified. Furthermore, ii) an operating starting point of a plant, in particular a hot rolling plant, which is configured to carry out a manufacturing process for producing the process product is specified, and iii) the process product is produced by executing the manufacturing process using the plant starting from the specified operating starting point. iv) an actual product property of the produced process product is determined, and v) an optimized operating point of the plant is determined based on the specified product property and the determined actual product property using an optimizer.

[0017] A manufacturing process within the meaning of the invention is preferably a technical process in which a product is manufactured, i.e., produced. The product does not necessarily have to be a "finished" (end) product; rather, it can also be a semi-finished product. A manufacturing process can therefore also be part of a larger manufacturing process for an end product, such as the hot rolling of a slab, in particular a width forming process carried out in the roughing train using a vertical stand.

[0018] An optimizer within the meaning of the invention is preferably a means for solving, in particular iteratively, an optimization problem. For example, an optimizer can be configured to determine one or more values ​​for process parameters with which a process delivers a desired result. The optimizer expediently uses a predetermined optimization strategy to solve the optimization problem. For example, selected process parameters can be varied according to the optimization strategy, and based on the resulting process result, the variation of the process parameters can be adjusted so that the process result converges toward the desired result.

[0019] Preferably, an optimizer is based on an optimization algorithm. The optimization algorithm can embody the optimization strategy.

[0020] Executing a manufacturing process using a system starting from a predefined operating point within the meaning of the invention preferably means operating the system at the predefined operating point or at least at an operating point determined based on the predefined operating point. However, it may happen that, despite appropriate adjustment of operating parameters, the system does not operate at the desired operating point, e.g., the predefined operating point. Such an execution of the manufacturing process can advantageously also be understood as execution using the system starting from the predefined operating point.

[0021] One aspect of the invention is based on the approach of using an optimizer to determine an optimal operating point at which a plant produces a process product with a specified product property, in conjunction with the (real) operation of the plant. In particular, an optimization strategy underlying the optimizer can be implemented by operating the plant. The plant can therefore be operated iteratively by the optimizer to determine the optimal operating point (or an operating point that is at least very close to the optimal operating point). The optimizer expediently specifies framework conditions, for example a variation of the plant's operating parameters or the corresponding operating parameter values ​​under which the plant is to be operated.By analyzing the manufacturing process carried out during plant operation, in particular the resulting process product, the optimizer can propose optimized operating parameter values ​​that are likely to enable the production of the process product with a product property that is closer to the specified product property. It is not necessary for the optimizer to evaluate a model or a so-called (mathematically formulated) objective function that characterizes the manufacturing process. Implementing the optimization strategy through (real) plant operation or evaluating the associated manufacturing process, particularly over a longer period of time, also advantageously allows for the consideration of changing operating conditions. For example, changes in external influences and / or changes in the plant condition, for example due to aging, can be considered.

[0022] In order to adapt a manufacturing process, an optimization algorithm can, for example, be divided into individual calculation steps. At the end of each calculation step, the manufacturing process is expediently evaluated using process parameter values ​​specified by the optimization algorithm, i.e., at a specified operating point. This evaluation is preferably carried out by operating the plant starting from the specified operating point, which is characterized by the specified process parameter values. In this respect, instead of calling up an objective function for a specific parameter combination, an evaluation of an actual physical process for precisely this parameter combination can be carried out. The result of the evaluation can then form the basis for calculating updated process parameter values, i.e., an optimized operating point, in the next calculation step.

[0023] The evaluation of the manufacturing process is preferably carried out by determining an actual product property of the process product produced by the manufacturing process. In particular, the determined actual product property can be compared with the specified product property. The comparison result can represent a quality of the actual product property of the produced process product. Based on this quality, the optimizer can determine optimized process parameter values, i.e., an optimized operating point for the plant. Using this type of method for adapting a manufacturing process, the effort—and thus also the costs—for commissioning a plant can be significantly reduced compared to manual methods. At the same time, a consistently high quality of the process products produced using the plant can be ensured.In particular, an optimal operating point of the plant can be easily adjusted when new product spectra are produced or the plant status changes.

[0024] Preferred embodiments of the invention and their further developments are described below. These embodiments can be combined with each other and with the aspects of the invention described below, unless expressly excluded.

[0025] In a preferred embodiment, the production of the process product is carried out using the plant instead of the optimization of a mathematical objective function. This allows the manufacturing process to be adapted using a conventional optimizer, in particular a conventional optimization algorithm. For example, a plant for which a model or mathematical objective function cannot be created can be put into operation using a conventional optimizer.

[0026] According to the invention, the process is preferably repeated continuously with the specified product properties, i.e., while maintaining the specified product properties. In particular, an optimal operating point can be continuously redetermined or updated during regular operation of the plant. This allows for any changes in operating conditions, such as changes in external influences or changes in plant properties due to aging or wear, to be compensated for, thus achieving consistently high quality of the process products.

[0027] It is advisable to specify the optimized operating point determined in each case as the starting point for the subsequent repetition, especially in the corresponding subsequent repetition. In this respect, the optimizer can check, based on a previously determined optimized operating point, whether the plant can be operated at an operating point more suitable for the specified product property. This ensures, particularly permanent, convergence of the actual product property with the specified product property.

[0028] Alternatively, the respectively determined optimized operating point can be used as the basis for the, in particular subsequent, repetition of a specification of the operating starting point. In other words, the operating starting point can be specified by determining the operating starting point based on the determined optimized operating point, for example, by deriving it from the determined optimized operating point. It is conceivable, for example, that one or more operating points in the vicinity of the determined optimized operating point are specified as the operating starting point. An operating point in the vicinity of the determined optimized operating point can be determined taking into account predetermined rules, i.e., for example, by a predetermined selection or determination process, for example, according to a predetermined optimization strategy.This may accelerate the convergence of the actual product characteristics with the specified product characteristics.

[0029] However, it is also conceivable that the operating starting point is chosen randomly based on the determined optimized operating point. In principle, the specified operating starting point does not have to be close to the determined optimized operating point.

[0030] In a further preferred embodiment, multiple process products are produced by carrying out the manufacturing process with the aid of the system at different operating points, starting from the specified operating point. In other words, the behavior of the system—and thus the corresponding manufacturing process—can be tested at multiple operating points before the optimizer determines an optimized operating point. For example, the manufacturing process can be carried out with the aid of the system at different operating points close to the specified operating point. This allows, in particular, so-called gradients for various process parameters to be determined, on the basis of which the optimizer ultimately determines the optimized operating point.

[0031] The different operating points can be selected starting from the specified operating point, e.g. from a previously determined optimized operating point, by varying at least one operating parameter of the system. It should be noted that if the manufacturing process is carried out multiple times at the same operating point of the system, the actual product properties are never exactly the same. Such variations, which are also referred to as natural process noise, are generally unavoidable because each system component is only "precise" within a certain tolerance range. Consequently, a standard deviation can be determined as a measure of the variation of the product properties. To determine the operating points, the operating parameter is therefore preferably changed so drastically that the product properties change by more than the standard deviation of the product properties.Against this background, an operating point close to the specified operating starting point is preferably an operating point which, starting from the specified operating starting point, is obtained by varying an operating parameter of the plant by a maximum of 10%, preferably a maximum of 5%, in particular a maximum of 2%, as long as the associated change in the product property is greater than the standard deviation of the product property. Other operating parameters expediently remain unchanged. Expediently, the actual product property is determined for each process product produced. Accordingly, it is preferred that the optimized operating point is determined on the basis of the specified product property and the several determined actual product properties. In this respect, the optimizer can check at which of the different operating points - iewhich process parameter variations achieve the best result with regard to the specified product properties. The corresponding operating point can be provided as the most promising candidate for achieving the specified product properties as an optimized operating point, for example, as a new starting point for the optimization process.

[0032] In order to enable the method to run effectively, in a further preferred embodiment the scatter of actual product properties of process products obtained when the manufacturing process is carried out multiple times at the same operating point of the plant is determined. For example, the standard deviation can be determined as a measure of the scatter of the product property. The measure of the determined scatter is preferably used as the basis for determining the different operating points. In other words, a distance, in particular a minimum distance, between the various operating points and / or to the predetermined operating starting point, for example the most recently determined optimized operating point, can be defined on the basis of the determined scatter. For example, the strength of the variation of at least one operating parameter of the plant can depend on the strength of the determined scatter.This ensures that a difference between the actual product property determined and the specified product property is due to a change in the operating point and not just to “noise”.

[0033] Preferably, the different operating points near the specified operating starting point are determined by varying one of several operating parameters. This expediently corresponds to the variation of a process parameter. Each variation of an operating parameter can represent a gradient formation in one direction in the corresponding parameter space. The gradients determined in this way for different operating parameters are expediently linearly independent, and in one embodiment, even orthogonal to one another, in the operating parameter space. Based on the respective determined actual product properties corresponding to the variations or the formed gradients, the optimizer can decide in which direction in the parameter space the search for an optimal operating point should continue.

[0034] In a further preferred embodiment, multiple process products are produced by performing the manufacturing process multiple times at the same operating point using the system. This allows the product property corresponding to the operating point to be determined more precisely. In particular, a standard deviation, which serves as a measure for any possible variations in the product property of the respective process product produced, can be determined.

[0035] Conveniently, the actual product property is determined for each process product produced, and the optimized operating point is determined based on the specified product property and a location parameter of the multiple determined actual product properties. A location parameter is preferably understood to be a mathematical summary of the multiple determined actual product properties, in particular a key figure of the sample embodied by the multiple determined actual product properties. Such a location parameter can be, for example, an arithmetic mean, a median, a mode, or the like of the multiple determined actual product properties.

[0036] For example, the system can be used to perform the manufacturing process at least three times, preferably at least four times, and especially at least five times, at the same operating point. It has been shown that with such a number of manufacturing processes per operating point, i.e., of manufactured process products, variations in the actual product properties can be reliably detected and taken into account accordingly, for example, by averaging. At the same time, with this number of manufacturing processes per operating point, the resulting effort is still manageable.

[0037] In a further preferred embodiment, the optimizer determines the predetermined operating starting point as the optimized operating point if a difference between the predetermined product property and the determined actual product property, in particular a target value representing the actual product property and a predetermined target value representing the predetermined product property, reaches or falls below a threshold value. In other words, the optimizer does not propose a new operating point as the operating starting point if the determined actual product property essentially corresponds to the predetermined product property. In this respect, a barrier can be introduced that can prevent a continuous change of the operating point during successive production runs of the process product in the range of the optimal operating point.As a result, a consistent quality of the process product can be ensured during regular operation of the plant.

[0038] It is advantageous if the number of times the manufacturing process is repeated using the system at the same operating point is increased to determine the actual product property as precisely as possible at that operating point, depending on the difference between the specified product property and the determined actual product property, preferably until the threshold value is reached or undershot. The number of repetitions is expediently selected such that the standard deviation of the product property is smaller than the difference. The threshold value can accordingly be selected as a value toward which the standard deviation of the product property converges with a high number of repetitions of the manufacturing process.

[0039] In a further preferred embodiment, the optimizer is configured to detect whether more than one optimal operating point exists for producing a process product with the specified product property. For example, it is conceivable that a particular material quality of a longitudinally rolled flat stock can be achieved by operating a rolling stand at different operating points. In this case, at least one of the determined operating points that is equivalent with regard to the specified product property is expediently output as the optimized operating point. Detection can generally be performed online during the production process or offline using recorded data from the production process.

[0040] If the specified operating starting point lies at the limit of the system's performance capability, it may happen that the system does not operate at the desired operating point despite setting operating parameters corresponding to the operating starting point. Therefore, in a further preferred embodiment, it is checked whether the operation of the system during the execution of the production process meets at least one specified criterion. In particular, it can be checked whether the production process was carried out at the specified operating starting point or at least close to it. For this purpose, recorded data from the production process, for example, control variables of the system, can be evaluated.

[0041] If it is determined that the production process using the system was not carried out at the desired operating point, for example, not at the specified operating point or not even close to it, the optimized operating point can be determined in various ways. Consequently, it is preferable to determine the optimized operating point of the system based on a test result.

[0042] For example, it is conceivable that the production of the process product is repeated at the specified operating starting point in the hope that the plant's operation now meets the specified criterion. Alternatively, a new operating starting point can be specified that is closer to an operating point at which the plant's operation meets the criterion. For example, if the optimizer specifies the operating starting point based on an operating point, such as a previously determined optimized operating point, and it turns out that the criterion is not met at this specified operating starting point, the optimizer can specify an operating point as the new operating starting point that lies between the specified operating starting point and the previously determined optimized operating point.

[0043] However, it is also conceivable that the determined actual product property is accepted and used to determine the (next) optimized operating point, even though the criterion is not met. For example, the determined actual product property can be compared with a prediction from the optimizer. If the actual product property is closer to the specified product property than a product property predicted for the specified operating point, the (next) optimized operating point is preferably determined based on the actually determined product property.

[0044] In a further preferred embodiment, the optimized operating point is determined by the optimizer using a trust region method and / or a line search method. In particular, the optimized operating point can be determined using a sequential quadratic programming (SQP) variant of the trust region method. Starting from the specified operating starting point, the optimizer preferably determines a "correction" of the operating point in the sense of the trust region method, in particular with the SQP variant by solving a quadratic minimization problem under constraints. The "usability" of this correction is expediently subsequently measured by generating the process product using the system at the optimized operating point and, for example, comparing the corresponding actual product property with the specified product property.Depending on the result of this “measurement”, the constraints of the quadratic minimization problem can be changed in order to be able to make a more qualified correction of the (new) operating starting point in a further iteration of the procedure.

[0045] In this respect, the optimizer preferably has an optimization algorithm, in particular in the form of the SQP variant of the trust region method or reinforcement learning. However, other optimization algorithms are also conceivable. In order to enable adaptation to changed operating conditions, for example, changed external influences or aging of the plant, the optimization algorithm preferably does not have a termination criterion. The optimization of the operating point or the process preferably does not end even if the determined actual product property essentially corresponds to the specified product property. This is because changed operating conditions may result in the plant having to be operated at a different (optimal) operating point at a later point in time in order to continue producing the process product with the specified product property.In this respect, the method is suitable, in particular in this embodiment, but also in all other embodiments described above or combinations thereof, for the regular, in particular permanent, operation of a system.

[0046] According to a second aspect of the invention, the method according to the first aspect of the invention is therefore used for the regular operation of a plant, in particular a hot rolling mill. The plant can operate essentially continuously in regular operation, while the optimizer iteratively determines optimized operating points, and the plant is operated at the determined optimized operating point after each iteration. This process can be fully automated, so that effort and thus costs can be significantly reduced not only for commissioning but also during regular operation of the plant.

[0047] A system according to a third aspect of the invention for, in particular automatically, adapting a manufacturing process, in particular a hot rolling process, comprises a plant for producing a process product. The plant is preferably a hot rolling plant, in particular a vertical rolling stand for rolling longitudinal flat products with a downstream horizontal rolling stand for longitudinal flat products. The process product can expediently be produced by carrying out a manufacturing process using the plant in the vicinity of a predetermined operating starting point. The system further comprises a means for determining an actual product property of the produced process product and an optimizer configured to determine an optimized operating point for the plant based on a predetermined product property of the process product and the determined actual product property.Such a system allows the evaluation of the manufacturing process under conditions specified by the optimizer, particularly those selected for testing, in order to find optimal or at least close-to-optimal values ​​for operating parameters. The system also allows for the consideration of changing operating conditions during regular operation, such as changes in external influences, changes in the condition of the plant, for example, due to aging, and / or changes in the processed materials, such as rolled materials.

[0048] The plant may be an industrial plant, such as a hot rolling mill. The plant may, for example, comprise a casting machine for producing a metallic strip from molten metal, particularly slabs, a vertical rolling stand (sometimes also referred to as an edge former), and a horizontal rolling stand for reducing the thickness of the slabs.

[0049] The means for determining an actual product property can comprise a sensor arrangement that can sense the process product, for example the slab, and determine the actual product property from the generated sensor data. The optimizer can be implemented in hardware and / or software. In particular, the optimizer can comprise a processing unit, preferably data- or signal-connected to a memory and / or bus system. For example, the optimizer can comprise a microprocessor unit (CPU) or a module thereof and / or one or more programs or program modules. The optimizer can be configured to execute instructions implemented as a program stored in a memory system, to acquire input signals from a data bus, and / or to output signals to a data bus.A storage system may comprise one or more, in particular different, storage media, in particular optical, magnetic, solid-state, and / or other non-volatile media. The program may be designed in such a way that it at least partially embodies or is capable of executing the methods described here, so that the optimizer can execute at least individual steps of such methods and thus, in particular, determine an optimized operating point of the system.

[0050] Short description of the drawings

[0051] The above-described properties, features, and advantages of this invention, as well as the manner in which they are achieved, will become clearer and more readily understood in connection with the following description of an embodiment, which is explained in more detail in conjunction with the drawings. These show, at least partially schematically:

[0052] FIG 1 shows an example of a conventional method for commissioning a plant;

[0053] FIG 2 shows an example of a system for adapting a manufacturing process; and

[0054] FIG 3 shows an example of a method for adapting a manufacturing process.

[0055] Where appropriate, the same reference numerals are used in the figures for the same or corresponding elements of the invention.

[0056] Description of the embodiments

[0057] FIG 1 shows an example of a conventional method for commissioning a plant which is set up to carry out a production process P. Such production processes P are generally dependent on operating conditions B and operating parameters of the plant, which can assume different operating parameter values ​​a. Depending on the operating conditions B and, for example, operating parameter values ​​a selected by a user, a product property E of the process product produced by the production process P varies. In order to be able to produce process products with a given product property E* using the plant or the production process P which can be carried out with the plant, the “correct” operating parameter values ​​a must be found. This process is referred to as commissioning.

[0058] For this purpose, the plant or production process P can be modeled using a model M. In addition to the operating conditions B and the desired, specified product property E*, which can be characterized by a so-called target value, model parameters Z can be entered into this model M as input variables. The model parameters Z, which are sometimes also referred to as commissioning parameters (IBS parameters), can be parameters inherent in the model M, but alternatively or additionally can also serve for the setup calculation and / or influence the execution of an algorithm for evaluating the model M.

[0059] The evaluation of model M in process step V1 with these operating conditions B, the specified product property E*, and the model parameters Z yields operating parameter values ​​a. In addition, such a model M can also be used to determine useful sensitivities for controlling the plant or the production process P (not shown).

[0060] However, practice has shown that operating the plant with the operating parameter values ​​a determined in this way generally does not exactly achieve the desired, specified product property E*. This is due, among other things, to the usually incomplete recording of all operating conditions B and the inherent limitations of the model M in describing reality (a model M can never describe reality with 100% accuracy).

[0061] Therefore, an attempt can be made to optimize the model parameters Z in a further process step V2. The optimization takes place subsequently, i.e., after the production process P has been carried out using the system at an operating point defined by the operating parameter values ​​a provided by the model M. In this respect, it is also referred to as offline optimization.

[0062] For this offline optimization, all relevant data D relating to the manufacturing process P are recorded while the system is running the manufacturing process P. Using this data D, the model M can be evaluated again – however, this time, instead of calculating operating parameter values ​​a, it is an expected product property E' that is calculated. A comparison of this expected product property E' with the product property E actually achieved by the manufacturing process P provides a quality G for the product property that can be achieved by the manufacturing process P running with the model-determined operating parameter values ​​a.

[0063] If the relationship between the model parameters Z and the quality G for the achievable product property, i.e., the achievable target value, is known, the model parameters Z can be optimized using the data D with conventional optimization methods (offline). The model parameters Z optimized in this way ultimately allow the plant to be commissioned with the corresponding (optimized) operating parameter values ​​a.

[0064] However, the described method is not applicable in cases where there is no model M that can adequately describe the manufacturing process P with reasonable effort. An example of such a manufacturing process P is the production of strip using a hot rolling mill. A typical specified product property E* is a rectangular shape of the strip ends with a specified width at the end of the rolling process. In practice, the strip end shape is influenced by hydraulically adjusting an edger upstream of a roughing stand of the hot rolling mill. The goal during commissioning is therefore to find the correct settings for the edger.

[0065] However, due to the material flow associated with rolling, the relationship between the upsetting process and the resulting strip end shape after rolling is extremely complex. Therefore, it requires extremely high computing power—if at all possible—to simulate the shape of the strip ends with specific edger settings, for example, using a finite element calculation. Due to computing time and accuracy constraints, such a model M is unsuitable for use in the commissioning of a plant, and certainly not for continuous control of the upsetting process.

[0066] In practice, a parameterized curve is therefore specified as the manipulated variable function for the edger. The curve parameters must be manually adjusted to the plant's operating conditions during commissioning. This requires considerable experience from the commissioning engineer and a significant amount of time until the parameters are set for the plant's entire production spectrum.

[0067] FIG 2 shows a system 1 for adapting a manufacturing process, which can be carried out using a system 10 and with which a process product 2 can be produced. The system 10 is designed here as a hot rolling mill, which is set up for hot rolling a metal strip—which in this example represents the process product 2. The hot rolling mill comprises a casting machine 11 for the preferably continuous casting of the metal strip, a cutting device 12 for dividing the cast metal strip, a vertical rolling stand 13 for reducing the width of the divided metal strip, a roughing stand 14 for a first thickness reduction of the divided metal strip, and a finishing stand 15 for a second thickness reduction of the divided metal strip.The hot rolling mill may comprise additional components (not shown), for example, one or more furnaces for (re-)heating the cast or cut metal strip, descaling stations for removing scale layers, additional separating devices for cutting the metal strip downstream of the roughing stand 14 and / or the finishing stand 15, and one or more reels for winding the cut metal strips onto coils. A transport device for transporting the process product 2 along a transport direction T is not shown.

[0068] The system 1 also has a means 20 for determining an actual product property E of the process product 2 and an optimizer 30 for determining an optimized operating point on the basis of the determined actual product property E and a predetermined product property E*.

[0069] The means 20 preferably comprises a sensor arrangement with which the process product 2 can be detected by sensors. For example, the means 20 can comprise one or more sensors in the form of cameras, (distance) ultrasonic sensors, and / or the like.

[0070] Conveniently, the actual product property E of the process product 2 can be determined based on sensor data generated during the sensory detection of the process product 2. In the variant shown, the means 20 can be configured for this purpose. Alternatively, the evaluation of the sensor data with regard to the actual product property E of the process product 2 can also take place elsewhere, for example, by the optimizer 30. Conveniently, the actually determined product property E is provided at the optimizer 30 in any case.

[0071] In the example shown, the means 20 is arranged between the roughing stand 14 and the finishing stand 15 with respect to the transport direction T. However, other arrangements of the means 20 are also conceivable, for example downstream of the finishing stand 15.

[0072] The optimizer 30 can be implemented in hardware and / or software. The optimizer 30 can, for example, comprise a computing device, such as a microprocessor, and / or a memory device connected to the microprocessor via signals or data. The optimizer 30 can, in particular, comprise a dedicated hardware architecture designed to determine an optimized operating point. The optimizer 30 can, for example, be implemented as an ASIC. Alternatively or additionally, the optimizer 30 can, for example, comprise one or more software modules that embody an optimization algorithm.

[0073] The optimizer 30 is expediently connected to the system 10 via an interface in order to initiate operation of the system 10 at the (newly) determined optimized operating point, in particular in order to be able to send correspondingly optimized operating parameter values ​​a to the system 10.

[0074] The vertical rolling stand 13 is designed to shape the side surfaces, sometimes also referred to as the longitudinal sides, of the cut metal strip according to a setting curve, so that the side surfaces do not run in a straight line along the length of the cut metal strip. In other words, the vertical rolling stand 13 can influence the width of the cut metal strip. This predetermined course of the side surfaces is intended to compensate for the material flow generated during rolling in the roughing stand 14—and possibly also in the finishing stand 15—so that a rectangular shape of the cut metal strip can be achieved downstream of the roughing stand 14 or the finishing stand 15.

[0075] The setting curve is expediently characterized by one or more operating parameters, the values ​​of which can be specified by the optimizer 30. Therefore, in the present example, the optimizer 30 is signal- or data-connected to the vertical rolling stand 13. After each cut metal strip has been detected by the means 20, in particular a measurement of the strip width in the initial and final regions of each metal strip, the optimizer 30 can vary at least one of the operating parameters depending on the correspondingly determined actual product property E, thereby changing the setting curve. The resulting shape of the cut metal strips can be detected again by the means 20, and the optimizer 30 can once again vary one of the operating parameters. As a result, for example, after twenty such passes, the cut metal strips form an essentially square shape.

[0076] Figure 3 shows an example of a method 100 for adapting a manufacturing process P.

[0077] In a method step S1, a product property E* of a process product 2 is specified, which can be produced using the manufacturing process P. In a further method step S3, a system 10, which is configured to produce the process product 2 by executing the manufacturing process P, is operated at a specified operating starting point X. The operating starting point X is previously specified in a method step S2.

[0078] In a further method step S4, an actual product property E of the process product 2 produced by the manufacturing process P is determined. For this purpose, a means 20 for determining the actual product property E can be provided. For example, in method step S4, the produced process product 2 is scanned by a sensor, and the actual product property E is derived from the sensor data generated thereby. The determined actual product property E is made available to an optimizer 30. In a further method step S5, the optimizer 30 determines an optimized operating point X* based on the specified product property E* and the determined actual product property E, for example, based on a quality of the determined actual product property E with regard to the specified product property E*.

[0079] When determining the optimized operating point X*, the optimizer 30 expediently uses an optimization algorithm 31, for example, a trust region method or a line search method. Starting from the current operating starting point X of the plant 10, which is characterized by a specific combination of values ​​for the operating parameters A of the plant 10, the optimization algorithm 31 can vary individual operating parameters A. The variation expediently occurs according to an optimization strategy underlying the optimization algorithm 31. For example, the optimization algorithm 31 can be configured to find a minimum in a potential landscape from the set of all operating points, which reflects the difference AE,E* between the actual product property E and the specified product property E*.In the present example, this is indicated one-dimensionally by a gradient determination when varying an operating parameter A.

[0080] The optimized operating point X* determined in method step S5 is expediently specified in a further iteration of method 100 as the operating starting point X in method step S2. This means that method 100 can be repeated starting from method step S2.

[0081] Unlike conventional optimization methods, no termination criterion is provided for method 100. This means that plant 10 can also be operated in regular mode using method 100. This ensures that plant 10 is automatically transferred to this modified operating point even when operating conditions change, which require a modified operating point to produce process product 2 with the specified product property E*.

[0082] However, to prevent the newly specified operating point X from continuously fluctuating within a range C around the optimal operating point X', in which the plant 10 produces a process product 2 with the specified product property E* or at least one very close to it, limits can be specified. For example, the specified operating point X can be output as the optimized operating point X* if a difference between the specified product property E* and the determined actual product property E reaches or falls below a threshold value. This avoids a "readjustment" of the operating point of the plant 10 once the range C, for example, the "floor" of the minimum shown in Figure 3, is reached.

[0083] Although the invention has been illustrated and described in detail by the preferred embodiments, the invention is not limited by the disclosed examples and other variations may be derived therefrom by those skilled in the art without departing from the scope of the invention.

[0084] List of reference symbols

[0085] 1 system

[0086] 2 Process product

[0087] 10 Appendix

[0088] 11 Casting machine

[0089] 12 Separator

[0090] 13 vertical rolling stand

[0091] 14 Roughing stand

[0092] 15 finishing rolling stands

[0093] 20 Means of determining an actual product characteristic

[0094] 30 optimizers

[0095] 31 Optimization algorithm

[0096] 100 procedures

[0097] S1-S5 process steps

[0098] V1, V2 process steps

[0099] A operating parameter a operating parameter value

[0100] B Operating condition

[0101] E Product characteristic

[0102] E* Specified product property

[0103] E' expected product property

[0104] Z model parameters

[0105] M model

[0106] P Manufacturing process

[0107] D Data

[0108] G quality

[0109] T Transport direction

[0110] X Operating starting point

[0111] X* optimized operating point

[0112] X' Optimal operating point c range

[0113] AE,E* Difference

Claims

Claims 1. Method (100) for adapting a manufacturing process (P), in particular a hot rolling process, comprising the steps: - specifying (S1) a product property (E*) of a process product (2) characterized by a target value; - specifying (S2) an operating starting point (X) of a plant (10), in particular a hot rolling plant, which is set up to carry out a manufacturing process (P) for producing the process product (2); - producing (S3) the process product (2) by carrying out the manufacturing process (P) using the plant (10) starting from the predetermined operating starting point (X); - determining (S4) an actual product property (E) of the produced process product (2); - determining (S5) an optimized operating point (X*) of the plant (10) on the basis of the predetermined product property (E*) and the determined actual product property (E) with the aid of an optimizer (30) for iteratively solving an optimization problem using an optimization strategy; - repetition of the process (100) with the specified product property (E*), wherein the respectively determined optimized operating point (X*) is specified as the operating starting point (X) during the repetition or is used as the basis for a specification of the operating starting point (X).

2. Method (100) according to claim 1, wherein - several process products (2) are produced by carrying out the production process (P) with the aid of the system (10) starting from the specified operating point (X) at different operating points, - for each process product (2) produced, the actual product property (E) is determined, and - the optimized operating point (X*) is determined on the basis of the specified product property (E*) and the several determined actual product properties (E).

3. Method (100) according to claim 2, wherein the scatter of actual product properties (E) of process products (2) obtained when the manufacturing process (P) is carried out several times at the same operating point of the plant (2) is determined and a measure of the scatter determined is used as a basis for determining the different operating points.

4. Method (100) according to claim 2 or 3, wherein the different operating points in the vicinity of the predetermined operating starting point (X) are determined by varying one of a plurality of operating parameters (A).

5. Method (100) according to one of the preceding claims, wherein - several process products (2) are produced by carrying out the manufacturing process (P) several times at the same operating point using the plant (10), and - for each process product (2) produced, the actual product property (E) is determined, and - the optimized operating point (X*) is determined on the basis of the specified product property (E*) and a position parameter of the several determined actual product properties (E).

6. The method (100) according to claim 5, wherein the manufacturing process (P) is carried out using the system (10) at least three times, preferably at least four times, in particular at least five times, at the same operating point.

7. Method (100) according to one of the preceding claims, wherein the optimizer (30) determines the predetermined operating starting point (X) as the optimized operating point (X*) when a difference (AE,E*) between the predetermined product property (E*) and the determined actual product property (E) reaches or falls below a threshold value.

8. The method (100) according to claim 7, wherein a number of executions of the manufacturing process (P) using the system (10) at the same operating point is increased as a function of the difference (AE,E*) between the predetermined product property (E*) and the determined actual product property (E).

9. Method (100) according to one of the preceding claims, wherein the optimized operating point (X*) is determined by the optimizer (30) by means of a trust region method and / or a line search method.

10. Method (100) according to one of the preceding claims, wherein it is checked whether the operation of the system (10) when carrying out the manufacturing process meets at least one predetermined criterion, and the optimized operating point (X*) of the system (10) is determined depending on a result of the test.

11. Method (100) according to one of the preceding claims, wherein the optimizer (30) comprises an optimization algorithm (31) and the optimization algorithm (31) has no termination criterion.

12. System (1) for adapting a manufacturing process (P), in particular a hot rolling process, with - a plant (10), in particular a hot rolling plant, for producing a process product (2) by carrying out a manufacturing process (P) with the aid of the plant (10) starting from a predetermined operating starting point (X); - a means (20) for determining an actual product property (E) of the produced process product (2); and - an optimizer (30) for iteratively solving an optimization problem using an optimization strategy, which is set up to determine an optimized operating point (X*) for the plant (10) on the basis of a predetermined product property (E*) of the process product (2), characterized by a target value, and the determined actual product property (E), so that the determined optimized operating point (X*) can be specified as the operating starting point (X) when the production process (P) is carried out again or can be used as the basis for specifying the operating starting point (X).

Images