WO2021094272A1

WO2021094272A1 - Method and system for determining correlations between detected product defects and detected state variables of a production system

Info

Publication number: WO2021094272A1
Application number: PCT/EP2020/081555
Authority: WO
Inventors: Stefan KLANKE; Christian Plociennik; Markus Reifferscheid; Holger Beyer-Steinhauer; Volker PICHLER
Original assignee: Sms Group Gmbh
Priority date: 2019-11-14
Filing date: 2020-11-10
Publication date: 2021-05-20
Also published as: EP4058858A1; DE102020210967A1

Abstract

The invention relates to a method, system and computer program for optimizing a production process (2) in a production system (3) from the metal production industry, the non-ferrous industry or the steel industry for producing semi-finished products or finished products, comprising the steps of: detecting state variables of the production system (3), detecting product defects of the products produced in the production system (3) and determining correlations between the detected product defects and the detected state variables of the production system (3), monitoring the state variables of the production system (3) in the future production of products and comparing them with the state variables of the determined correlations between the detected product defects and the detected state variables of the production system (3), and modifying the detection of product defects of the products produced in the production system (3) in the future if the comparison of the monitored state variables of the production system (3) in the future production of products with the state variables of the determined correlations between the detected product defects and the detected state variables of the production system (3) provides a sufficient match.

Description

PROCESS AND SYSTEM FOR DETERMINING RELATIONSHIPS BETWEEN DETECTED PRODUCT DEFECTS AND DETECTED STATE VARIABLES OF A PRODUCTION PLANT

The invention relates to a method and a system for optimizing a production process in a metal-producing production plant

Industry, the non-ferrous industry or the steel industry for the production of semi-finished or finished products, in particular for monitoring the product quality of rolled or forged metal products.

The invention generally relates to a plant in the iron and steel industry, such as a plant in the metal-producing industry, the non-ferrous (non-ferrous) or steel industry, or master alloy production. Examples of such industrial systems are blast furnaces, direct reduction systems, electric arc furnaces, converters or systems for ladle processes, systems for primary or forming of metals such as continuous or billet casting systems and hot and / or cold rolling systems, or upstream or downstream of these systems

Systems such as ovens, e.g. B. reheating or holding furnaces, finishing facilities, coating lines, cooling sections, pickling or annealing.

The products manufactured in the production plant are subject to certain quality fluctuations, for example caused by external influences, quality differences in the raw material, wear and tear of individual components of the production plant or similar impairments of the production process in the production plant. For the further processing of the manufactured product, however, compliance with the specified product qualities is imperative, since otherwise the manufactured product is not suitable for the intended use. In order to ensure a given product quality of the products manufactured in the production plant, it is known to monitor the state variables of the production plant and to intervene in the production process if the state variables deviate from given areas in order to ensure the given product quality. State variables of the production plant include, for example, the following process parameters: pressure, temperature, speed, forces, torques, damage, vibrations, functional impairments, cooling performance, rolling parameters, device and / or component wear, contamination, buildup, or the like. The aim is to avoid state variables, that have a negative impact on product quality. For this purpose, models of the production process running in the production plant are often used in order to determine the specified ranges of state variables.

It is also known from the prior art to monitor the product quality of the products manufactured in the production facility. Depending on the type of product quality to be monitored and the production and / or testing process, monitoring takes place continuously or randomly. The surface of a manufactured product can often be checked by means of continuous monitoring using camera systems, while checking for inclusions regularly includes taking a sample and checking the sample manually or automatically. The aim is to identify products that do not meet the specified product quality so that they can be sorted out.

However, automatic monitoring of product quality in a production plant in the metal-producing industry, the non-ferrous industry or the steel industry for the production of flat products or finished products is difficult to implement for the following reasons: Different error groups must be used for different product surfaces (glossy, matt, smooth, rough, etc.) can be recognized can, which requires differentiated device settings and costly additional equipment;

Detection and classification parameters must be continuously maintained and adapted to changing requirements (e.g. new customer requirements);

Vibrations in the production system (e.g. belt vibrations) or unevenness can only be tolerated to a limited extent;

Oil and emulsion mist affect optical systems; and

Low-contrast errors can only be recorded automatically to a limited extent; for example, cold roll marks from Sendzimir roll stands are currently still recorded manually by an inspector.

Limits and disadvantages of manual control by an inspector in the context of a continuous or random check are, for example: 100% surface control of all manufactured products, such as belts, in high-performance production systems with coupled inspection and production is due to the high process speeds of e.g. up to 130 m / min not possible; the entire defect classification is based on a subjective defect classification without reproducible parameters; decreased control performance on textured surfaces;

Risk of changeable quality standards in the assessment due to fatigue; in the case of random checks at discrete points, errors can be overlooked; and when measurements are taken by a person being measured, their personal characteristics are added, in particular specialist knowledge, manual skills, experience, current physical / psychological state.

All of the aforementioned influencing variables can be causes of errors, which can result in systematic or random measurement errors.

In the following, the technological background is shown as an example using a surface test of flat products.

An essential quality feature of rolled and forged products is the surface quality. The surface structure is set and influenced at numerous process stages in steel production. The surface is therefore tested in numerous process stages - from slab and hot strip to processes in the cold rolling mill such as pickling, rolling, annealing and coating.

The test task is to check the surfaces of steel belts for damage, soiling and other irregularities. This check can take the form of a random check on the belt, e.g. B. every umpteenth band per melt or production lot as well as in further, differently defined intervals. This test can be carried out seamlessly from the beginning of the strip to the end of the strip (100 percent control of the random sample) or at discrete points, such as the control of the beginning and end of the strip after cold strip scratching.

Since in the area of the cold rolling mill even the smallest changes in the surface structure can lead to reworking and rejects, and such deviations can often not or only very poorly be recognized with the naked eye under normal ambient conditions, various methods and systems are used to deal with both metallurgical and production-related To be able to determine defects according to the respective production stage. These are: 1. Surface inspection with special light sources: purely optical inspection

2. Worker inspection with a whetstone: usually limited to random sample inspection

3. Optical surface inspection systems: Modern surface inspection systems today allow extensive

Automation of the surface inspection with the help of digital camera systems and suitable downstream data processing.

In particular in the inspection stands of rolling and strip processing plants that work with endless strips (e.g. preparation lines) or even fully continuous (e.g. galvanizing lines), continuous testing of the

Belt surface carried out.

Conventional surface inspection involves a visual inspection (often referred to as a "visual inspection"); H. In an inspection stand or booth, the conveyor belt is subjected to a visual surface inspection by experienced personnel. Detected surface defects are classified by the inspector and entered into the accompanying defect log using the manual input desk, which is then available for subsequent evaluation and processing.

With the automated surface inspection (OIS), the strip surface is recorded online by cameras with high resolution and all defects are detected (detecting anomalies), classified (depending on the type of surface defect, e.g. scratches, marks, ...) and documented . Suspect areas of the tape are stored as an image and, after evaluation, provide a large number of features, on the basis of which the recorded areas are divided or classified into (defect) classes.

In the steel industry, the surface quality of strips is increasingly automated with the help of automatic surface inspection systems (OIS) carried out. This already applies to pickled, continuously annealed, hot-dip galvanized products and, to a lesser extent, to hot-rolled products.

Based on this prior art, the invention is based on the task of improving the checking of product qualities of semi-finished products or finished products produced in a production plant in the metal-producing industry, the non-ferrous industry or the steel industry and, on the basis of the checking of the product quality, of the production process in to optimize the production plant.

The object is achieved according to the invention by a method for optimizing a production process in a production plant in the metal-producing industry, the non-ferrous industry or the steel industry for the production of semi-finished or finished products, in particular for monitoring product qualities of rolled or forged metal products, comprising the steps: Acquisition of state variables of the production plant,

Detecting product defects in the products manufactured in the production facility, and

Determination of connections between the recorded product defects and the recorded state variables of the production plant, monitoring of the state variables of the production plant in the future manufacture of products and comparison with the state variables of the determined connections between the recorded product defects and the recorded state variables of the production plant, and adaptation of the recording of product defects of the products manufactured in the production plant in the future if the comparison of the monitored state variables of the production plant in the future manufacture of products with the state variables of the determined relationships provides a sufficient correspondence between the recorded product defects and the recorded state variables of the production plant.

In a first step of the method according to the invention, state variables of the production plant are recorded. State variables within the meaning of the invention include process variables, machine and apparatus-related state variables in the system, system, process and / or model parameters, characteristic values or the like. A condition monitoring system can be used to record the status variables. This step includes, in particular, the monitoring of the process automation of the production plant.

In a second step, product defects in the products manufactured in the production facility are recorded. The detection of product defects usually takes place when the product quality achieved is checked, i.e. in the area of quality testing.

According to the invention, relationships between the recorded product defects and the recorded state variables of the production plant are determined. The method according to the invention thus determines which recorded state variables have an influence on recorded product defects. As a result, an effective cycle is created between condition monitoring and quality inspection.

During the future manufacture of products, the state variables of the production plant are monitored. The status variables recorded in this way can, as described above, be related to the correspondingly recorded product defects. In particular, the status variables recorded during the future manufacture of products are compared with the status variables of the determined relationships between the recorded product defects and the recorded status variables of the production plant. The product qualities determined by the quality inspection, in particular the recorded product defects, are reported back to the condition monitoring system. The feedback of the information is a feedback (feedback loop). It includes the feedback of realized product qualities, in particular of quality features and product defects.

If the comparison of the monitored status variables of the production plant in the future production of products with the status variables of the determined relationships between the recorded product defects and the recorded status variables of the production plant provides a sufficient match, the recording of product defects in the production plant is adjusted. If status variables occur in future production which essentially correspond to status variables from previous productions to which product defects have already been assigned, then the current detection of product defects can be matched to this specific product defect. According to the invention, the detection of product defects is improved in the context of the quality check. In particular, this improves the accuracy and the speed of the detection of product defects.

By comparing the status variables from the condition monitoring system with the status variables from the already determined relationships between status variables and product defects and adapting the recording of product defects in the quality inspection, a feedforward loop is created which provides a prognosis of probable product defects. Information about expected product defects, such as, for example, the type of defect and / or the defect location, can be provided to the quality inspection system in advance, before the manufactured product, such as a coil, reaches the quality inspection. Since the information on the expected product defect was made available to the quality inspection in advance, the inspection process can be tailored to this, for example by means of more targeted metrological recording and fault detection. According to the invention, an active circuit with a closed information flow via the feedbackward and feedforward loops is thus created. The method according to the invention thereby enables the quality for finding and measuring product defects (quality defects) and / or the quality of the interpretation of status information about machines, systems and their components to be influenced and / or optimized in a targeted manner. The basic idea of the method according to the invention is to link the identification of quality-relevant, machine condition-related deviations by a condition monitoring system with the processes for measuring product defects (quality defects), in particular the test processes, using the previously described operating cycle. An integrative method is thus created for online assessment of the effects of process deviations due to the machine condition with regard to the required product quality, as well as an optimization of the targeted metrological recording during quality testing through defined test features and to improve the analysis, diagnosis and / or prognosis quality of condition monitoring Systems, for example by optimizing threshold values.

The relationship between the measured quantity and the quantity to be diagnosed (status variable and / or product error) can be formed differently

In the simplest case, (additional) sensors are installed for the variable to be diagnosed, for direct status control (e.g. optoelectronic sensors such as a video camera for monitoring the function of cooling nozzles).

In many cases, a direct measurement of the variable to be diagnosed (condition control) is not possible, so that one has to rely on indirect features and draw conclusions about the condition of the machine or one of its components. For example, it is usually not possible to measure the state of wear on gear drives directly ("assessment without dismantling"). In such a case, one has to rely on signals that are related to the variable to be diagnosed, such as those coming from a transmission and are related to the state of wear of the gear transmission. In the case of the gear drive of a drive, such a signal is represented by the vibrations that arise during operation, which are recorded by vibration sensors (additional sensors and machine-internal sensors)

Furthermore, non-measurable quantities can be derived by means of some measurable quantities in combination with mathematical processes from very different subject areas: from the classic parameter estimation processes (e.g. condition observer), traditional control technology, as well as indirect measurement processes and model-based measurement technology, via data-based modeling / identification (including Soft sensors and neuro-fuzzy methods) to tree search methods from computer science. According to a preferred variant of the invention, the method comprises the step of adapting the setting method in the production plant in order to avoid states with state variables that are related to detected product defects. For this purpose, the product quality determined by the quality inspection, in particular the recorded product defects, is reported back to the condition monitoring system. The reported product quality can be compared by the condition monitoring system with a previous prognosis (calculation / model) of a product quality by the condition monitoring system. A predicted product quality is therefore compared with an achieved product quality. On the basis of the comparison, the setting method in the production plant can be adapted in order to avoid states with state variables that are related to detected product defects. In this way, the quality of the interpretation of status information about machines, systems and their components can be optimized. In a variant of the invention, the method comprises the step of maintaining the production plant in order to avoid states with state variables that are related to detected product defects. If the method according to the invention identifies state variables which result in a certain product defect and if the state variables can be avoided by repairing the production plant, then the production plant is repaired accordingly in order to avoid the state variables and the resulting product defects in manufactured products. According to an advantageous variant, the method according to the invention comprises the step of improving a model for specifying parameters for the manufacturing method in the production plant on the basis of the determined relationships between the detected product defects and the detected ones

State variables of the production plant. It is known from the prior art to predict parameters for a production method in a production plant by means of a model. This is intended to optimize the production process in the production plant, in particular by predicting parameters for the manufacturing process, taking into account target criteria such as product quality. The relationships determined by the method according to the invention between the detected product defects and the detected one

According to this variant, state variables of the production plant are used to improve the model.

In a variant according to the invention, the detection of state variables of the production plant relates to one or more of the following parameters: pressure, temperature, speed, forces, torques, damage,

Vibrations, functional impairments, cooling performance, rolling parameters, device and / or component wear, dirt, buildup, or the like.

According to a variant of the invention, the recording of status variables of the production plant takes place by means of suitable sensors, automatic or manual test methods, manual tests by an operator, or the like. The Sensors are designed in particular for direct or indirect measurements of state variables. In the case of an indirect measurement, a parameter or state of the production plant is measured, from which the state variable is derived. According to an expedient variant of the invention, the detection of product defects in the products manufactured in the production plant relates to one or more of the following defect parameters: type of defect, location of the defect, size of the defect, or the like.

In an expedient variant according to the invention, product defects in the products manufactured in the production facility are detected by means of surface inspections, chemical analyzes, sampling, Fland measuring devices, visual inspections, measuring devices, test devices or the like.

According to an advantageous variant according to the invention, the adaptation of the detection of product defects of the products manufactured in the production plant in the future takes place in that a targeted search is made for product defects that are related to the detected state variables of the

Production plant for the future production of products. A product defect in the manufactured product is therefore no longer only searched for in general, but rather a specific product defect is searched for in particular.

According to a particularly preferred variant, the detection of product defects in the products that will be manufactured in the production facility in the future is adapted in that the detection is focused on certain defect parameters. Error parameters are, for example, the type of error, location of the error, error pattern, or the like. In the case of manual quality control, in particular, the indication of the probable fault location is an important information, since the inspector does not have to check the entire manufactured product, but can concentrate on the probable fault location. But also with one Automatic quality control, the location and type of error are important information, since, for example, the automatic error detection can be focused accordingly.

In a particularly advantageous variant according to the invention, the method further comprises the steps:

Saving the recorded status variables of the production plant over the runtime of the production plant, and

Compare the stored status variables with status variables assigned to a newly detected product defect. As a result, product defects found later in the value chain can also be associated with the state variables recorded at the time of manufacture. By storing the recorded state variables, the detection of the product defects, i.e. the quality check, can take place with a time delay and thus also locally independent of the manufacture of the product.

According to an advantageous variant of the invention, in addition to the status variables, further information on the manufactured product is stored, such as product identifiers, product recipients, storage location, delivery information, required product quality, or the like. Thus, if a product defect is subsequently assigned to status variables recorded, the product that has already been manufactured can be identified and, in particular, its current location can be more easily determined.

According to an expedient variant, the method comprises the identification of a product that has already been manufactured if the comparison of the stored status variables with status variables assigned to a newly detected product defect has yielded one or more hits. The already manufactured product identified in this way can be checked for the newly recorded product defect and possibly sorted out from the subsequent value chain or sent to alternative further processing.

In a further variant according to the invention, the method comprises the step of recording the type and / or product properties of the product manufactured in the production facility. The type of product manufactured differentiates, for example, between rolled products (strips, sheets) and long products (pipes, profiles / beams, wire). Product properties are, for example, geometry, material alloy, mechanical properties, surface, or the like. According to an advantageous variant of the invention, the method includes the step of taking into account the detected type and / or product properties of the product manufactured in the production plant when determining the relationships between the detected product defects and the detected state variables of the production plant. The relationships between the recorded product defects and the recorded state variables of the production plant are therefore recorded separately for different product types and / or product properties. This increases the accuracy of the method according to the invention, since the same state variables for different product types and / or product properties do not necessarily result in the same product defect, or a product defect is not essential for every product type and / or product characteristic.

According to an expedient variant of the invention, the detection of state variables of the production plant and the detection of product defects in the products manufactured in the production plant take place spatially separately. The detection of product defects during the quality inspection is thus carried out in a different part of the plant than the actual production of the product to be manufactured. In this way, for example, the production plant avoids influencing the detection of the product defects. In this variant, the connection between the recorded product defects and the recorded status variables of the production plant is determined with a time delay, since the The manufactured product must first be transported to the quality inspection to record the product defects. Only after the product defects have been recorded can the relationships to the recorded state variables of the production plant be determined. In an expedient variant of the invention, production in the production plant takes place in several plant parts, the method according to the invention being carried out in each plant part and / or for the entire production plant. The status variables can be recorded in one or more parts of the system. Also, after each or part of the system parts, product defects in the (intermediate) product manufactured at that point in time can be detected. The corresponding recorded status variables and recorded product defects can each be taken into account individually over a plant section or variably over several plant sections and / or the entire production plant when determining the relationships. For this purpose, the system parts expediently exchange information on the recorded and / or ascertained data with one another and / or with a central data processing device.

According to an advantageous variant according to the invention, the method comprises the step of transmitting the determined relationships between the recorded product defects and the recorded state variables of the production plant to a production planning system of the production plant

Production control system of the production plant or other systems of the production plant that can evaluate and / or utilize the determined relationships between the recorded product defects and the recorded state variables of the production plant. The determined relationships between the recorded product defects and the recorded state variables of the production plant are therefore used to further optimize the

The production process in the production plant is transmitted to other systems in the production plant, which take the determined relationships into account for optimization. According to an expedient variant of the invention, the method comprises the step of outputting a warning message, a proposed measure, an instruction or the like to an operator, a subsystem of the production plant or a further processing device when a product defect is detected.

In an advantageous variant, the method according to the invention comprises the step of adapting the further processing in order to maintain product quality after the further processing. If product defects are detected in the quality check, the method according to the invention tries to adapt the further processing of the product so that the

Quality requirements for the product are met after further processing. Otherwise, the further processing is stopped and the product with the detected product defect is sorted out or sent to another further processing for a product with quality requirements, the quality requirements of which can still be met despite the detected quality defect.

According to a further advantageous variant of the invention, the method includes the step of defining permissible ranges for the state variables of the production plant and outputting a warning message if

State variables outside the defined areas are recorded. The warning message is issued to other systems in the production facility, so that appropriate measures to comply with the permissible

State variables can be executed.

According to an expedient variant, the defined permissible ranges are adapted, in particular continuously, on the basis of the determined relationships between the recorded product defects and the recorded state variables of the production plant.

The object is also achieved by a system for optimizing a production process in a production plant in the metal-producing industry, the non-ferrous industry or the steel industry for the production of Semi-finished or finished products, in particular for monitoring the product quality of rolled or forged metal products, including:

Sensors for detecting state variables of the production plant and / or interface to a production monitoring system of the

Production plant for receiving status variables of the production plant,

Quality inspection for recording product defects in the products manufactured in the production facility and / or interface to a quality inspection system for receiving product defects in the

Production plant manufactured products, and

Computing device for determining relationships between the detected product defects and the detected state variables of the production plant, monitoring the state variables of the production plant by means of the

Sensors or via the interface to the production monitoring system in the future production of products and comparing with the state variables of the determined relationships between the detected product defects and the detected state variables of the production system by means of the computing device, and

Adjustment of the quality check for the detection of product defects of the products manufactured in the production plant in the future if the comparison of the monitored state variables of the production plant in the future manufacture of products with the state variables of the determined relationships between the detected product defects and the detected ones

State variables of the production plant provide a sufficient match. In an expedient variant, the system according to the invention is designed to carry out the method according to the invention.

The production monitoring system is also referred to as a condition monitoring (CM) system. All components, systems and the like of the system according to the invention are expediently networked with one another, for example by means of a wired or wireless communication network.

The system comprises sensors for detecting status variables of the production plant or an interface to a production monitoring system of the production plant for receiving status variables of the production plant. The status variables are recorded, for example, by machine and / or process monitoring, which are part of an automation system of the production plant. The machine and / or process monitoring or the automation system is expediently coupled to a so-called condition monitoring system.

The quality check for detecting product defects in the products manufactured in the production facility or the interface to a quality control system for receiving product defects in the products manufactured in the production facility include facilities at a test location for quality measurement and / or quality testing. The facilities for quality measurement and / or quality testing range from simple

Hand measuring equipment up to complex measuring devices and from

Measuring devices up to visual inspection. As far as possible, devices preferably comprise an automation solution.

The data from the sensors and the quality check are evaluated by a computing device to determine relationships between the recorded product defects and the recorded state variables of the production plant. The computing device has in an inventive Variant Access to a quality information system such as a quality database for evaluating tape logs, historical data (coil files) and / or archives for result and / or quality data, for example. This data is available, for example, in a quality management system (QMS tools, quality information or CAQ systems) and / or in BDE protocols in production planning systems (level 3 systems, PPS systems, MES systems). The computing device is preferably integrated in the aforementioned condition monitoring system, which in turn is coupled with the machine and / or process monitoring or the automation system. An exemplary system according to the invention combines a condition monitoring system with the measurement and testing of quality features and has access to a system with quality data, such as a quality management system (QMS tools, quality information or CAQ systems) and / or to a production planning and control system ("PPS systems", "MES systems"). The condition monitoring system with the corresponding sensors and the networking with the plant automation system collects and interprets the data on the status of the process as well as of machines, plants and their components.

The aforementioned components are networked with one another for data exchange. The quality test can also take place at several test locations, that is, spatially and / or temporally distributed. The data are then preferably processed centrally by the computing device.

The location of the machine and process monitoring (where quality defects arise) and the test location can be located at different distances from one another. The location of the test can be close to the machine, e.g. B. the measurement takes place during the process (inline and online measurement), or remote from the machine, e.g. B. Measurement takes place separately in special laboratory equipment / measuring room. Furthermore, the measuring and testing process can also be used in subsequent processes within the further processing in the factory for steel production or non-ferrous metal production take place or within the value chain of steel users and sheet metal processors or non-ferrous metal processors.

The object is also achieved by a computer program comprising commands which, when the program is executed by a computer, cause the computer to execute the method according to the invention, in particular comprising commands which cause the system according to the invention to execute the method according to the invention.

The present invention, i.e. the method according to the invention, the system according to the invention and / or the computer program according to the invention enables, for example by means of the condition monitoring system described, a new functionality for online assessment of the effects of (A) process deviations caused by the machine state and (B) unfavorable combinations of Machine and process states with regard to the required product quality. As a result, sorting criteria in particular for "conspicuous strips" (in general: conspicuous rolled or forged products) are formulated and a corresponding warning message is generated.

This is the starting point for the cycle of action according to the invention between the integrative condition monitoring system, measuring or testing system and / or measuring or testing process for quality measurement and the feedback options for verifying the condition monitoring system. In the case of an integrative condition monitoring system, the condition monitoring system is coupled to an automation system for process monitoring, which optionally includes at least partial machine monitoring. The types of error considered can be periodic and non-periodic errors. Examples of the causes of periodic errors are excitations proportional to the roll speed such as roll eccentricities or roll imbalances, non-periodic errors are caused, for example, by instability of the lubrication (eg change in the lubrication angle). When checking for quality errors, according to the invention, corresponding information on the type of error and the location of the error from the condition monitoring system is used. If this check is carried out for a quality feature that is checked continuously (e.g. as part of an automatic surface inspection of cold-rolled strips), the information from the condition monitoring system is used by the relevant test system (the OIS) in such a way that the software for Control of the measuring system (here the camera) the fault location and the type of fault during the quality inspection are recorded in a more targeted manner using measurement technology. Furthermore, this information should be used to

Evaluation software within the image processing chain, e.g. B. in the detection ("There is something"), the feature extraction and / or the classification ("It is a ...") to improve in order to achieve an optimized quality allocation.

If the inspection is carried out as part of a random sample or outside of a planned random sample (separate inspection), the inspector can use the information

Search specifically for the error (coil no., Affected strip meter, type of error, ...)

(confirmative analysis), since both the information about the type of error ("what should be searched for or discovered") and the error location ("where should be searched for") are available. Through a feedback (feedback loop) from the test process and the

Information about the actually measured test result for the integrative condition monitoring system, the cycle of action of the system, method and / or computer program according to the invention is closed. This feedback loop from the test process - with information about the quality feature measured in each case - always exists, even if the condition

Monitoring system has not given a warning (in such a case the threshold values are not reached and may be too high).

By connecting the information from the condition monitoring system about possible, expected quality errors (forecast / expectation calculation) with the A feedforward loop is created for the measurement or testing process and the cycle of action is formed. The products identified with the condition monitoring system or the relevant areas must now be checked by a test / measuring system (automated or by an expert / inspector), e.g. B. whether they are committee or not.

By means of the present invention, i.e. by means of the method according to the invention, the system according to the invention and / or the computer program according to the invention, conspicuous products such as ribbons can be found. For this purpose, the production is monitored for process deviations, in particular by determining quality-relevant process variables, for example in rolling mills and strip processing systems, and assigning product defects. Conspicuous products can, for example, be sorted out or monitored and / or treated separately in the subsequent manufacturing stages. Furthermore, inter alia By means of process restrictions (limit values for certain process parameters during the manufacturing process), the interaction between, for example, the machine and the production process running in the machine is mapped and taken into account.

The present invention can also be used to optimize measurement processes and / or measurement methods. This can also include the activities of a measurement person. In particular, the present invention enables a more targeted metrological detection of product defects (quality defects) from subsequent inspection activities (tests). The more targeted metrological detection takes place, for example, by means of tuning / fine-tuning of hardware components and / or software of a measuring system, or by specifying / predicting the fault location and / or fault image, for example as a calibration module for line scan cameras. The indication of an error can be, for example, “chatter marks at a distance of x mm on the upper and / or lower side of the belt”. For example, time-based data is recorded and converted into length-related data. Like other measurement systems need automatic

Surface inspection systems one maintenance: detection and

Classification parameters must be maintained and adapted to changing requirements (e.g. new requirements from the automotive industry as a steel processor) so that they continuously achieve good results. So are z. B. Improvements of 30% of the inspection performance after a tuning of only three weeks on many OIS systems realistic. Tuning or fine-tuning of hardware components and / or software of a measuring system by means of the present invention is therefore advantageous. Furthermore, the accuracy of a condition monitoring system of the production plant can be increased by means of the present invention, in particular by sharpening threshold values of state variables of the production plant. The threshold values are sharpened (a) by training KI algorithms for diagnosis or prognosis of the CM system in the early learning phase (no historical data are yet available) using physical-mathematical ones

Simulation models, (b) through the above-mentioned cycle of action

(Causal relationship, causal relationship) of the present invention. This is made possible by the feedback option of the active circuit between the CM system and the test processes. For example, it is recorded which errors have actually occurred, when certain threshold values are exceeded or also when they are not exceeded (= feedback loop for verification and optimization of the analysis, diagnosis and prognosis quality of the CM system).

A variant of the invention enables automatic recognition of process situations or machine states. Objective: To show which finished products (e.g. bundles) could also be affected.

Subsequent inspection of products in the warehouse (e.g. bundles in bundles). Complaints should be reduced, if necessary, complaint predictions should be derived. The online condition monitoring is additionally trained on the basis of the historical data with a known and verified classification. With this approach of a “non-static training phase” of the Condition Monitoring systems, the present invention is able to adapt to changed framework conditions. Changed framework conditions are, for example, the exchange of tools (rollers, molds, ...) or other machine components as well as the production of a wide variety of rolled products (steel types and their dimensional ranges), which have changed behavior (e.g. wear behavior of the tool or deformation behavior of the rolled product ) exhibit. This is synonymous with a large parameter space and great variability of the data, which must be covered. The present invention also offers the advantage of shortening fault distances, such as, for example, strip areas with a defective surface. This is a significant advantage, particularly in the case of production with an endless belt or in continuous operation. One example is the occurrence, detection and removal of impurities on the in-line skin pass mill of a galvanizing plant.

The length of the fault path depends on:

(A) the time span between occurrence and first discovery of the error (e.g. at time t = t * for coil no.N, the error is detected on strip meter x *). This includes the time and space between the place of origin and the place of recording / testing. In addition, the measurement or test accuracy and the detection capability of the test method or of the tester for the types of defects in order to ensure reliable and timely defect detection (time span between measurement and perception “this is a defect”). (B) the response time for a correction: time for analysis (cause?),

Recognition of the required measure and implementation of the measure.

The shortening of the error distance is achieved according to the invention by 1. Safe and earlier detection using verified information available at an early stage; concerns proportion (A) based on a prognosis, and

2. Shortening the response time; Part (B) can be shortened on the basis of a cause analysis of the invention. In addition to the cycle of action described above, the method, system or computer program according to the invention can be networked with other work areas of various operational organizational areas and / or their information systems (e.g. maintenance, the

Production planning, ...) or further process steps within metal production and processing up to use by the metal user (horizontal integration). Examples are:

• The flow of information and suggested measures to the maintenance department. This option should only be mentioned for the sake of completeness, since condition monitoring systems are traditionally used in maintenance areas.

• The information to the operational production, e. B. to process management, level 2 (production area) and suggestions for measures to eliminate the error.

• The warning message or the verified warning message about a quality error as well as about the corresponding system status also receives the

Production planning (production area) and / or quality planning (quality assurance area). These evaluate this information with regard to the planned production of strips. For example, based on knowledge of quality defects and / or the degree of damage to components (cardan shafts, etc.), the load caused by the planned production sequence can be adjusted accordingly.

• When the system, method or computer program according to the invention is networked with a further processing stage (horizontal Integration), for example, depending on the predicted quality errors determined during the test process, an assistance system could be used to analyze whether the quality features can be influenced in the subsequent process steps by adapting the process parameters accordingly so that the required specifications can be adhered to. Since no rescheduling of production is required, this forecast tool can work independently within a level 2 system or a "level 2.5".

The cycle of action according to the invention can also be used for long-term tasks within the operational organizational areas mentioned above.

Examples of such tasks are: performance analysis and monitoring of the production plant in the production area; Incident statistics and weak point analyzes in the area of maintenance, as well as long-term test data processing in the area of quality assurance. In the context of offline analyzes (medium and long-term considerations), further analysis of the causes of quality defects can be carried out, or evaluations and analyzes such as:

- Do certain defects accumulate on the drive or operating side? - what is the ratio of major to minor errors?

- could the product quality be improved?

By means of these offline analyzes in combination with the present invention, an improved process optimization of the production system can take place through the improved correlation of process and quality data. Furthermore, these offline analyzes in combination with the corresponding long-term machine states, such as For example, the annual progression of damage to drive elements, process management and production planning can be optimized. The achievable goals are: long-term Avoidance of impermissible operating conditions, reduction of maintenance and material costs as well as increase of the service life of components and machines, while maintaining the required quality.

The system, method or computer program according to the invention provide the information technology basis for this purpose, with a simultaneous increase in the accuracy of the statements based on the verified information from the cycle of action that is now available at an early stage. As a result, the diagnostic reliability can be increased significantly and the forecast quality can be increased for the aforementioned applications.

The invention is explained in more detail below with reference to the exemplary embodiments and further examples shown in the figures. Show it:

1 shows a schematic view of a first embodiment of a system according to the invention for optimizing a

Production process in a production plant of the metal producing industry, the non-ferrous industry or the steel industry for the manufacture of semi-finished products or

Finished products, FIG. 2 shows a schematic view of a second embodiment of a system according to the invention for optimizing a

Finished products, and

3 shows a detailed view of an embodiment of a production monitoring system in connection with an automation system of a production plant from metal producing industry, the non-ferrous industry or the steel industry for the production of semi-finished or finished products.

Fig. 1 shows a schematic view of a first embodiment of a system 1 according to the invention for optimizing a production process 2 in a production plant 3 of the metal-producing industry, the non-ferrous industry or the steel industry for the production of semi-finished or finished products, in particular for monitoring product qualities of rolled products or forged metal products. The system 1 from FIG. 1 comprises sensors 4 for detecting status variables of the production plant 3. Furthermore, the system 1 from FIG. 1 comprises an interface to a production monitoring system 5, also referred to as a condition monitoring system, for sending / receiving status variables of the production plant 3 The production monitoring system 5 cooperates with an automation system 9 for controlling the production process 2 in the production plant 3. The sensors 4 can be part of the production plant 3 or in the immediate vicinity of the production plant 3 for recording the state variables of the

Production plant 3. The state variables detected by the sensors 4 are transmitted, for example, via the interface to the production monitoring system 5 for further processing.

The system 1 from FIG. 1 further comprises a quality check 6 for detecting product defects in the products manufactured in the production plant 3 and an interface to a quality check system 7 for receiving product defects in the products manufactured in the production plant 3. In the exemplary embodiment shown, the quality check uses the quality check system 7 via an interface and receives the product defects detected by the quality check system 7 in the products manufactured in the production facility 3 via the interface. The quality testing system 7 can be a testing system 10 for controlling, implementing and / or evaluating the Include detection of product defects of the products manufactured in the production facility 3. As an alternative or in addition, the quality inspection system 7 can be based on the activities of a measurement person 11.

The system 1 according to the invention further comprises a computing device 8 which receives and or queries the status variables detected by the sensors 4 and / or the production monitoring system 5 and the product defects detected by the quality inspection 6 and / or quality inspection system 7. The computing device 8 is connected at least to these aforementioned components of the system 1 for the exchange of data, for example via a wired or wireless communication network (not shown).

The computing device 8 is designed to determine relationships between the recorded product defects and the recorded state variables of the production plant 3.

The system 1 according to the invention monitors the state variables of the production plant 3 during the future production of products by means of the sensors 4 and / or via the interface to the production monitoring system 5. These state variables of the production plant 3 determined during the monitoring of the future production of products are combined with the state variables of the Correlations between the recorded product defects and the recorded state variables of the production plant 3 by means of the

Computing device 8 compared.

If the comparison of the monitored status variables of the production plant 3 in the future production of products with the status variables of the determined relationships between the detected product defects and the detected status variables of the production plant 3 is sufficient

The quality check 6 for the detection of product defects in the products manufactured in the production plant 3 in the future is adapted. The system 1 according to the first exemplary embodiment of the invention shown in FIG. 1 is designed to carry out a method according to the invention for optimizing a production process 2 in a production plant 3 of the metal-producing industry, the non-ferrous industry or the steel industry for the production of semi-finished or finished products, in particular for monitoring the product quality of rolled or forged metal products. In particular, the method comprises the following steps:

Acquisition of state variables of the production plant 3,

Detecting product defects in the products manufactured in the production facility 3, and

Determination of relationships between the recorded product defects and the recorded state variables of the production plant 3,

Monitoring the state variables of the production plant 3 in the future production of products and comparing them with the state variables of the determined relationships between the detected product defects and the detected state variables of the production plant 3, and

Adaptation of the detection of product defects of the products manufactured in the production plant 3 in the future if the comparison of the monitored state variables of the production plant 3 in the future production of products with the state variables of the determined relationships between the detected product defects and the detected state variables of the production plant 3 provides a sufficient correspondence. The method according to the invention preferably further comprises the step of adapting the production method in the production plant 3 in order to avoid states with state variables which have a relationship have recorded product defects. This avoids repeated product errors that are generated by the same state variables.

In a further variant, the method includes the step of maintaining the production plant 3 in order to avoid states with state variables that are related to detected product defects.

According to a preferred variant, the method according to the invention further comprises the step of improving a model for specifying parameters for the manufacturing method in the production plant 4 on the basis of the determined relationships between the detected product defects and the detected state variables of the production plant 3.

The detection of state variables of the production plant 3 includes in particular one or more of the following parameters: pressure, temperature, speed, forces, torques, damage, vibrations, functional impairments, cooling performance, rolling parameters, device and / or component wear, contamination, buildup, or the like. The

Detection of state variables of the production plant 3 takes place in particular by means of suitable sensors 4, automatic or manual test methods 6, manual tests by an operator 11, or the like.

The detection of product defects in the products manufactured in the production facility 3 includes, for example, one or more of the following

Error parameters: type of error, location of the error, size of the error, or the like. The detection of product defects in the products manufactured in the production system 3 takes place in particular by means of surface inspections, chemical analyzes, sampling, hand-held measuring devices, visual inspections, measuring devices, testing devices or the like.

The adaptation according to the invention of the detection of product defects of the products manufactured in the future in the production plant 3 takes place in that a targeted search is made for product defects that are related to the recorded state variables of the production plant 3 in the future Manufacture of products. The detection of product defects in the products manufactured in the production facility (3) in the future is adapted, for example, in that the detection is focused on certain defect parameters. In a particularly advantageous variant of the invention, the method further comprises the steps:

Saving the recorded state variables of the production plant 3 over the running time of the production plant 3, and

Compare the stored status variables with status variables assigned to a newly detected product defect.

For this purpose, the system 1 comprises a storage device 12 which is connected to other components of the system 1 for data exchange, in particular to the computing device 8, the production monitoring system 5, the sensors 4 and / or the quality inspection system 7. In addition to the status variables, further information about the The manufactured product is stored, such as product identifiers, product recipients, storage location, delivery information, required product quality, or the like.

If the comparison of the stored status variables with status variables assigned to a newly detected product defect has delivered one or more hits, a manufactured product can be identified subsequently. The identified, already manufactured, product can then be checked for the newly recorded product defect.

According to an expedient variant, the method according to the invention records the type and / or product properties of the product manufactured in the production facility 3. The type of product manufactured differentiates, for example, between rolled products such as flat products (strips, sheets) and long products (pipes, profiles / beams, wire). Product properties are, for example, geometry, material alloy, mechanical properties, surface, or the like.

The recorded type and / or the recorded product properties of the product manufactured in the production system 3 can be taken into account when determining the relationships between the recorded product defects and the recorded state variables of the production system 3. According to a variant according to the invention, the detection of state variables of the production plant 3 and the detection of product defects of the products manufactured in the production plant 3 take place spatially separately. The production plant 3 and the quality inspection 6 are therefore spatially separated.

Due to the spatial separation, the connection between the recorded product defects and the recorded state variables of the production plant 3 is determined with a time delay.

In a variant according to the invention, the multiple system parts exchange information on the recorded and / or ascertained data with one another and / or with the central data processing device 8. According to a particularly preferred variant of the invention, the method includes the step of outputting a warning message, a suggested measure, an instruction or the like to an operator, a subsystem of the production plant (3) or a further processing device when a product defect is detected. The method according to the present invention expediently additionally comprises the step of adapting the further processing in order to maintain product quality after the further processing. Furthermore, according to the invention, permissible ranges for the state variables of the production plant 3 can be defined and a warning message is output if state variables outside the defined ranges are detected.

Fig. 2 shows a schematic view of a second embodiment of a system 1 according to the invention for optimizing a production process 2 in a production plant 3 of the metal-producing industry, the non-ferrous industry or the steel industry for the production of flat products or finished products, in particular for monitoring product qualities of rolled products or forged metal products. The second embodiment from FIG. 2 differs from the first embodiment from FIG. 1 in that quality data determined by the quality check are transmitted to a quality management system 13 and / or a production planning / control system.

Furthermore, the determined relationships between the recorded product defects and the recorded state variables of the production plant 3 are transmitted, in this variant according to the invention, from the computing device 8, to the production planning system 14 of the production plant 3, the production control system 14 of the production plant 3 and / or the quality management system 13 of the production plant 3 . The determined relationships between the recorded product defects and the recorded status variables of the production plant 3 can be evaluated and / or used by these systems to optimize the production process in production plant 3 of the metal-producing industry, the non-ferrous industry or the steel industry for the production of flat products or Finished products, in particular for monitoring the product quality of rolled or forged metal products.

FIG. 3 shows a detailed view of an embodiment of a production monitoring system 5 in connection with an automation system 9 of a production plant 3 of the metal-producing companies Industry, the non-ferrous industry or the steel industry for the production of semi-finished or finished products.

It can be seen from FIG. 3 that the integrative production monitoring system 5 (condition monitoring (CM) system) determines the setpoints and / or control variables for the automation system 5, by means of which the automation system 9 controls the production process 2 in the production plant 3. For this purpose, the production monitoring system 5 uses, for example, simulations, models and / or expert knowledge. The setpoints and / or control variables can be determined either from the (i) condition monitoring system itself and / or from (ii) the automation system and / or with the help of (iii) additional simulations and models as well as expert knowledge.

The production monitoring system 5 receives the status variables (plant and process signals from the sensors 4 and the quality data (detected product defects) from the quality check 6. The production monitoring system 5 evaluates the received

State variables and quality data. In connection with the computing device 8, the relationships between the recorded product defects and the recorded status variables of the production plant 3 are determined in a training phase or, in normal operation, the status variables of the production plant 3 are monitored during the (future) manufacture of products and with the status variables of the determined correlations between the recorded product defects and the recorded state variables of the production plant 3 are compared.

If the comparison of the monitored status variables of the production plant 3 in the future production of products with the status variables of the determined relationships between the detected product defects and the detected status variables of the production plant 3 provides a sufficient match, a diagnosis is carried out according to FIG. 3, i.e. a search for possible mistakes. Furthermore, a short-term and / or long-term prognosis (prediction) and a short-term and / or long-term therapy (proposed measure) can be created.

The therapy suggestions can be recommendations that process parameters should be changed, for example with the current condition of the rollers (the bearings, the drive spindle, etc.) the desired or required quality is not achievable, please change the operating point of the control ", up to concrete specification of the setpoints and / or control variables.

The prognosis and / or therapy proposals can be based on special simulations and / or by using special models, such as B. Remaining lifetime models as well as based on expert knowledge. For example, by means of a simulation-based forecast, based on the observed past load changes, the values of future load changes for the production process can be forecast (simulation into the future).

If it is determined that the required product quality cannot be achieved with the current process parameters, (A) production in production plant 3 can be rescheduled, (B) the process parameters can be recalculated, or (C) the quality can be downgraded to the best possible quality become.

The invention is explained in more detail below with the aid of specific operating examples. In general, there are basically two approaches to recording quality features in the production systems under consideration:

1. continuous recording by means of an online measurement method (this corresponds to a 100% test), if possible in the process sequence (often also referred to as inline test) 2. random sampling for (sample testing)

Example 1 - continuous testing In the following, the sequence is described using the example of a continuous test using an automatic surface inspection system (OIS) within a galvanizing line in a cold rolling mill, as an example of the application of the invention for continuous detection (100% test of the surface feature):

IA) There is a border crossing, z. B. for the vibration in the roll stand, determined for the first time. For example, at time t *, an overflow occurs for the first time in the condition monitoring system. The information about the bundle or coil no. , as well as the corresponding tape section (tape meter) is available in the condition monitoring system.

I B) There is a warning message about the process capability for the desired quality, i. H. there is a conspicuous band or a conspicuous band section with a possible quality problem.

This warning message is sent to the relevant test system (in this example the automatic surface inspection) at the corresponding test location

(Inspection status in the outlet area of the galvanizing plant). This message can take the following form, for example:

Specification of the error to be expected, e.g. B. chatter vibration, chatter marks at a distance of x mm on the top or bottom of the strip on the coil no. X and tape meter Y.

For this purpose, the recorded time-based data was converted into length-related data (if necessary, a change in length of the strip sections due to the change in thickness in the subsequent roll stands must be taken into account by the condition monitoring system).

Furthermore, this information can be passed on to other information systems, e.g. B. to the automation system, ie the operator in the control room receives this warning message, as well as to higher-level systems, such as For example, a quality assurance system and quality analysis system that encompasses process levels.

IC) This information is used by the relevant test system (the automatic surface inspection system) in such a way that the software for controlling the measuring system (here the camera) identifies the fault location and the

The type of error in the quality inspection is recorded in a more targeted manner using measurement technology and the evaluation software in the image processing chain, e.g. B. in the detection ("There is something"), the feature extraction and / or the classification ("It is a ...") is improved in order to achieve optimized quality allocation

Setting parameters of the measuring system to detect the defect or by specifying the relevant area on the strip surface.

ID) By feedback from the test system or test process and the information about the actually measured test result to the integrative condition monitoring system, the cycle of action of the system according to the invention is closed. Through this

Feedback loop, you automatically receive an improvement in the analysis, diagnosis and prognosis quality of the condition monitoring system, e.g. B. by optimizing the threshold values (automatic training and fine tuning at defined intervals).

Regardless of whether a warning message was given by the condition monitoring system and an error was discovered during testing, this feedback always took place, i. H. The condition monitoring system is trained with all data on tapes for which a quality check has taken place.

IE) Another procedural step is the subsequent review of historical data, e.g. B. for strips that are in further processing stages or in the coil store. The process for the condition monitoring search in the history is as follows: If the test result shows a quality error or shows a quality deviation, the condition monitoring system (or another higher-level system) automatically searches the history of the strips that have already been manufactured with similar features (symptoms) relevant for error detection, such as e.g. B. the vibration profile.

The process in a specific operational situation is, for example: (starting point) there is a quality error "chatter marks at a distance of 20mm for X tape meters" -> Condition Monitoring System searches for the corresponding spectrum and the run-out speed -> Finding the relevant

Amplitude at frequency Y (= run-out speed through

Mark spacing);

(Goal) Which belts had similar vibrations under similar process conditions (speed, roller diameter / ID, belt material, thickness and width). With this automatic recognition of process situations and

The "suspicious tapes" identified by the condition monitoring system are reported to the quality assurance department about the machine status, stating the coil number, the tape meter concerned, the defect category and other characteristics of the defect. These error features can be: The condition monitoring system is expected and forecasted, chatter marks at a distance of x mm.

These "conspicuous straps or bundles" must then be checked separately by experts for the warning message regarding a quality defect by the condition monitoring system (separate check). In the case of massive quality defects discovered in bundles, which have already been

Customers, there is also the option of a detailed quality warning message (bundle number, affected belt area, type of error) to deal with later complaints and, in particular, production costs, to avoid processing steps that have already been carried out at the customer's site on faulty primary material.

The results of this method step are also used to improve the analysis, diagnosis and prognosis quality of the condition monitoring system and represent a further feedback loop of the system and method according to the invention.

Example 2 - sample inspection

In the following, the sequence is shown using the example of a sample inspection of the surface inspection by an inspector on a skin pass mill stand in a cold rolling mill, as an example of the application of the invention for discontinuous detection (sample inspection of the surface feature):

The aim of regular inspections on the skin pass mill stand is to

Detecting irregularities on the belt or on the tools, which under certain circumstances would affect the entire batch over a longer period of time. For this purpose, manual inspections are carried out on the standing skin-pass mill at defined intervals (the production process is interrupted for this purpose) and the strip is checked on discrete strip sections. First a few meters of the strip are wound onto the reel mandrel. Then the band under tension or the considered band section is lightly sanded over a length of around one meter over the entire width. After peeling, abnormalities become clear, which are assessed and classified by the inspector. Grinding with a whetstone is usually done manually, while grinding robots are used in development and in the first operating phase. In addition, there are already initial considerations to use an automated surface inspection, which, however, has not yet been reliably technically possible due to the complexity of the error detection.

With manual inspection, humans do more than just perform any activity within the measurement process. A visual inspection requires the Inclusion of the measurement person in the measurement solution process, ie the human being is a direct component of a measurement chain - without him, no usable result is obtained.

2A) Step: "Sorting criteria for" suspicious tapes "- Finding these tapes": This step is carried out in analogy to operating example 1 - continuous

Examination, point 1A)

2B) Step: "Information flow / transfer - warning message: This step is carried out in analogy to operating example 1 - continuous testing, point 1B). The inspector can e.g. B. in the control room by means of a

Coil report application received.

2C) Step: "Carrying out the test": With the help of the information (coil number, affected strip meter, type of defect, ...) the inspector can search for the defect in a targeted manner (confirmative analysis), since both the information about the

The type of error ("what should be searched for or discovered") and the location of the error ("where should be searched for") are present. The test takes place, for example, in an inspection stand outside the production line (offline) or within the production line when the line is at a standstill. When using robots for grinding and for

Surface testing, this step can be carried out analogously to step (1C).

2D) to 2E): These steps are carried out in analogy to operating example 1 - continuous testing, points (1D) to (1E)

Example 3 - Cross-process consideration A special feature in the steel industry is that quality features are influenced in several processing stages. For example, the strength and surface finish is determined by almost all processing stages (albeit different) influenced or generated. In addition, there are process stages (e.g. the "liquid phase" of steel production) in which the quality feature is not generated or developed, but the respective process is essentially responsible for the future development of these quality features and significantly influences the feature that has not yet been developed .

Ultimately, these quality features can only be reliably set by means of a quality test that spans the processing stages.

The process sequence of the patent for a cross-process level consideration is described below using the example of a continuous surface inspection from continuous casting to the finished, cold-rolled and coated strip (end product).

3A) Step: "Sorting criteria for" suspicious tapes "- Finding these tapes": This step is carried out in analogy to operating examples 1 and 2, point (1A) or (2A)

For example, the condition monitoring system detected an abnormality “servo valves hanging” (cause: diaphragm accumulator leaking, leakage) on the oscillating cylinders for the mold oscillation. In our example, this abnormality is found on 5 slabs.

3B) Step: "Information flow / transfer - warning message: According to the previous operating examples, there are two options: There is an automatic surface inspection (OIS) of the slabs: this step then takes place in analogy to operating example 1 - continuous testing, point (1B) es a manual surface inspection is carried out (for example, visual inspection by means of scarfing control lines) on the slabs: this step is then carried out analogously to operating example 2 - random sample inspection, point 2B) 3C) Step: "Carrying out the test": see step (1C) or (2C) according to the options mentioned in step (3B).

Two scenarios are now possible:

I. An error is found during the test according to the warning message

II. No error is found in the test process, i. H. (i) there are no surface defects or (ii) there are currently no surface defects because the formation has not yet taken place in the process stage under consideration. For example, there are metallic inclusions in the form of rows below the slab surface (with reduced quality in terms of cleanliness levels close to the surface), which are strongly stretched and tear open by subsequent deformation during rolling.

The further steps of the procedure for scenario I are then: 3D / scenario I) to 3E / scenario I): These steps are carried out in analogy to the

Operating examples 1 and 2, point (1D) to (1E) or point (2D) to (2E)

The further procedural steps for scenario II are initially:

3D / Scenario II) to 3E / Scenario II): These steps are carried out in analogy to the

Operating examples 1 and 2, points (1D) to (1E) or points (2D) to (2E) for the process stage currently under consideration. Furthermore, the steps under points (3A) to (3C) are run through for each test at the test centers of the respective subsequent production stage. The warning message from the upstream process stage (here: the continuous caster in the steelworks) is available and is taken into account accordingly by the condition monitoring system. In case (ii) of scenario II), these steps are carried out until a surface defect corresponding to that in a process stage (e.g. after cold rolling) Warning message (the elongation of the strip caused by rolling is taken into account when specifying the fault location).

After the surface defect has been found for the first time in accordance with the warning message, there is the following feedback according to the procedure (similar to the step under point (1D) or (2D), but now across the process level (cross-process level feedback).

The system A / experienced according to the invention carries out the subsequent examination of historical data from upstream process stages. In our example, the abnormality “servo valves hanging” (oscillating cylinder, mold oscillation) occurred on 5 slabs. Based on the verified prognosis of the Condition Monitoring System “surface defect” for the already manufactured, suspicious slab, the system / method gives recommendations for action for the slabs that have not yet been manufactured. This could be, for example, "Check slab with warning message ..." at the point (slab meter) for inclusions ". For this purpose, a sample would be taken from the conspicuous slab. B. by means of ultrasound or etching (Baumann prints from sample discs) determine the internal and near-surface degree of purity. This means that due to the expected error, a special incoming inspection of a quality feature is carried out for the subsequent "rolling" process. This feedback loop even enables proactive quality control: with regard to the downstream process, preventive error detection takes place before the start of the production process, which is intended to prevent defective strips from being manufactured in the first place.

Furthermore, a retrospective review of historical data can be carried out in accordance with the steps under point (1 E) or (2E).

Further examples for finding or avoiding quality errors or unwanted deviations from quality features with the help of the invention are:

• Cold rolling area: "heat scratches" (surface defects) as well as the generation of unsuitable surface roughness of the strip (surface structure) During rolling, for example, by monitoring the condition of the roll gap lubrication (e.g. the viscosity of the rolling emulsion) and the process variables (e.g. rolling force), possibly combined with comparisons with tribological model calculations

The flushing in of casting powder slag can lead to non-metallic inclusions under the slab surface. The consequences are so-called shell defects on the strip, which occur during the later rolling process. · Area of rolling of long products: internal defects (longitudinal and transverse cracks) from e.g. B. Round material such as tubes, rods and wires can be found optimized or avoided by means of the invention in combination with eddy current tests (inline or offline testing).

In the specific operating examples listed above, only the condition monitoring functions “Monitoring” and “Diagnosis” of the machine and process status have been considered so far.

As already described above, the functions of the condition monitoring system of the overall system described above can be expanded in further embodiments of the invention (see FIG. 3). The prognosis functionality of a condition monitoring system with regard to a quality error or quality feature due to the present machine and process status is also intended to be used by the invention. For this purpose, the condition monitoring functions “Predictive” and “Prescriptive” of the expected machine wear due to future production events (e.g. predicted degree of wear for the planned production batches) are to be included. This means that a prognosis can now be carried out, for example, on the quality-optimal, condition-oriented exchange of interchangeable parts (e.g. tools such as work rolls). Here is a comparison between the future achievable product quality in relation to the planned product quality.

The following are possible scenarios for the online application of the system and method according to the invention.

Under the boundary conditions:

• Phase “trained system”: a connection between the recorded product defects and the recorded state variables of the production plant was learned.

• Planned order sequence for the products: Coil no. N, coil number N + 1, etc.

• At time t = t * for coil no. N, strip meter x *, an overrun occurs for the first time, the system issues a warning message “Quality error”. for t> t * the scenarios result from the possible combinations of the following factors:

• Factor A - "affected product": The warning message "Quality error due to the machine / process status" concerns: i. the currently manufactured rolled product (coil no. N) or ii. the subsequent rolled product (coil no. N + 1)

• Factor B - "The source of the quality defect can be eliminated within the current process stage" i. yes ii. No

• Factor C - "Quality errors can be corrected within the current process stage" i. Yes No Examples of the application scenarios are:

Scenario A (i) and B (i): For the currently manufactured rolled product (coil no. N), the source of the error is eliminated and thus the quality error (according to t> tcorr, with tcorr = t ^* +

corresponding tape meter x (t> tcorr). With the system A / experienced according to the invention, the error distance (Dc = x (t = tcorr) −x *) can be significantly shortened, and the remaining error distance is then cut out. E.g. advance warning and confirmed feedback of skin pass marks by the surface inspection system, by activating an automatic floch pressure cleaning on the rollers, the corresponding points can be cleaned in a targeted manner.

Scenario A (i) and C (i): After t> t *, new process parameters or new dynamic reference variable curves are determined for the currently manufactured rolled product (coil no. N), with which the quality error can be eliminated. Examples for scenarios A (ii) and C (i) (current process stage) as well as for scenarios A (ii) and C (ii) (cross-process stage applications) relate to the level of production planning and process management. For example, the warning message “quality error” can lead to a check as to whether the planned quality can be achieved with the planned process parameters. If the prediction is made: planned production (e.g. a casting sequence with crack-prone grades) is not possible with the current machine status (e.g. faulty segment 7) or the planned quality cannot be achieved with the planned process parameters (casting speed. , Spray plan, ...), the following measures can be proposed at an early stage based on the system and method according to the invention:

Variant A: Sending an early warning signal: "The production plans must be changed", e.g. B. in level 3 (production planning)

Variant B: Optimization (recalculation of the process parameters), e.g. B. in level 2 (process management), with the goal “how can the quality anyway be achieved? ". The forecast error (with 80% probability, with 90% probability, etc.) can also be specified, as well as the information “at which operating costs”. Lowering the casting speed leads, for. B. to a reduction in production output, by means of the connection with a cost module, the specified optimization measure can also be evaluated economically.

Variant C: with what level of quality can the planned production plan be used for production.

These variants are summarized in FIG. 3.

Examples for eliminating the source of the quality error:

In the case of chatter vibrations during rolling: the control center is informed of the error message (= warning message that has now been verified), as well as the cause of the error. If necessary, there is a corresponding regulation to compensate for the discovered quality defect, e.g. regulation for the vibration problem 3rd octave chatter vibration (chatter vibration with a harmonic vibration of the third order) otherwise proposed measure: Recommendation - change of process parameters (e.g. lowering the rolling speed, braking) to eliminate the quality error in the current machine / process status.

Based on the knowledge of quality defects and / or the degree of damage to components (cardan shafts, etc.), the stress on the components can also be optimized accordingly through the planned process management, in which inadmissible operating states are now avoided. For example, by leveling the pass schedules, the Peak loads at the first stitches are reduced (reduction of the component stress) so as not to increase the existing degree of damage and to extend the service life in such a way that an unplanned system downtime is avoided.

After determining a quality error predicted with the trained CM system, the process parameters are changed in such a way that the quality deviations can be corrected. For example, the condition monitoring system detects a reduced nozzle output within a heating or cooling section, which causes a deviation from the setpoints and leads to quality deviations. The corresponding heating or cooling output would now be changed within the process control or a control system in order to achieve the desired quality result.

List of reference symbols

1 system

2 production process

3 production plant 4 sensor

5 Production monitoring system (Condition Monitoring System)

6 quality inspection

7 quality inspection system

8 Computing device 9 Automation system

10 test system

11 measuring person

12 Storage facility

13 Quality management system 14 Production planning / control

Claims

Expectations

1. A method for optimizing a production process (2) in a production plant (3) of the metal-producing industry, the non-iron industry or the steel industry for the production of semi-finished or finished products, in particular for monitoring product qualities of rolled or forged metal products, comprising the steps :

Acquisition of state variables of the production plant (3),

Detecting product defects in the products manufactured in the production facility (3), and

Determination of relationships between the recorded product defects and the recorded state variables of the production plant (3),

Monitoring the state variables of the production plant (3) in the future production of products and comparing with the state variables of the determined relationships between the detected product defects and the detected state variables of the production plant (3), and

Adjustment of the detection of product defects of the products manufactured in the production plant (3) in the future if the comparison of the monitored state variables of the production plant (3) in the future production of products with the state variables of the determined relationships between the detected product defects and the detected state variables of the production plant (3 ) provides a sufficient match.

2. The method of claim 1, further comprising the step of:

Adaptation of the manufacturing process in the production plant (3) to avoid states with state variables that are related to detected product defects.

3. The method of claim 1 or claim 2, further comprising the step of:

Maintaining the production plant (3) to avoid states with state variables that are related to detected product defects.

4. The method according to any one of claims 1 to 3, further comprising the step:

Improving a model for specifying parameters for the manufacturing process in the production plant (3) on the basis of the determined relationships between the recorded product defects and the recorded state variables of the production plant.

5. The method according to any one of claims 1 to 4, further comprising the step: wherein the detection of state variables of the production plant (3) comprises one or more of the following parameters: pressure, temperature, speed, forces, torques, damage, vibrations, functional impairments, Cooling performance, rolling parameters, equipment and / or component wear, contamination, buildup, or the like.

6. The method according to any one of claims 1 to 5, wherein the detection of state variables of the production plant (3) by means of suitable sensors (4), automatic or manual test methods (6), manual tests by an operator (11), or the like takes place.

7. The method according to any one of claims 1 to 6, wherein the detection of product defects of the products manufactured in the production plant (3) comprises one or more of the following defect parameters: type of defect, location of the defect, size of the defect, or the like.

8. The method according to any one of claims 1 to 7, wherein the detection of product defects of the products manufactured in the production plant (3) is carried out by means of surface inspections, chemical analyzes, sampling, Fland measuring devices, visual inspections, measuring devices, test devices or the like.

9. The method according to any one of claims 1 to 8, wherein the adaptation of the detection of product defects of the products manufactured in the production plant (3) in the future takes place in that a targeted search is made for product defects that are related to the detected state variables of the production plant (3 ) in the future production of products.

10. The method according to claim 9, wherein the detection of product defects in the future products manufactured in the production facility (3) is adapted in that the detection is focused on specific defect parameters.

11. The method according to any one of claims 1 to 10, further comprising the steps:

Saving the recorded state variables of the production plant (3) over the running time of the production plant (3), and

12. The method according to claim 11, wherein, in addition to the state variables, further information on the manufactured product is stored, such as product identifiers, product recipients, storage location, delivery information, required product quality, or the like.

13. The method of claim 11 or claim 12, further comprising the

Step:

Identifying a product that has already been manufactured if the comparison of the stored status variables with status variables assigned to a newly detected product defect has yielded one or more hits.

14. The method of claim 13, comprising the step of:

Check the already manufactured product for the newly detected product defect.

15. The method according to any one of claims 1 to 14, further comprising the

Step:

Detecting the type and / or product properties of the product manufactured in the production plant (3).

16. The method of claim 15, comprising the step of:

Taking into account the recorded type and / or product properties of the product manufactured in the production plant (3) when determining the relationships between the recorded product defects and the recorded state variables of the production plant (3).

17. The method according to any one of claims 1 to 16, wherein the detection of state variables of the production plant (3) and the detection of product defects of the products manufactured in the production plant (3) take place spatially separately.

18. The method according to claim 17, wherein the connection between the recorded product defects and the recorded state variables of the production plant (3) is determined with a time delay.

19. The method according to any one of claims 1 to 18, wherein the production in the production plant (3) in several

Plant parts takes place, the method according to the invention being carried out in each plant part and / or for the entire production plant (3).

20. The method according to claim 19, wherein several parts of the plant information about the recorded and / or determined data with each other and / or with a central

Exchange data processing device (8).

21. The method according to any one of claims 1 to 20, further comprising the step of: transmitting the determined relationships between the recorded

Product defects and the recorded state variables of the production plant (3) to a production planning system (14) of the production plant (3), a production control system (14) of the production plant (3) or other systems of the production plant (3) that determine the relationships between the recorded product defects and the captured

Evaluate and / or utilize state variables of the production plant (3).

22. The method according to any one of claims 1 to 21, further comprising the step: Output of a warning message, a proposed measure, an instruction or the like to an operator, a subsystem of the production plant (3) or a further processing device when a product defect is detected.

23. The method of claim 22, further comprising the step of:

Adjustment of further processing to maintain product quality after further processing.

24. The method according to any one of claims 1 to 23, further comprising the step:

Defining permissible ranges for the state variables of the production plant (3) and issuing a warning message if state variables outside the defined ranges are detected.

25. The method according to claim 24, wherein the defined permissible ranges are adapted, in particular continuously, on the basis of the determined relationships between the recorded product defects and the recorded state variables of the production plant.

26. System (1) for optimizing a production process (2) in a production plant (3) of the metal-producing industry, the non-iron industry or the steel industry for the production of semi-finished or finished products, in particular for monitoring the product quality of rolled or forged metal products, full: Sensors (4) for detecting state variables of the production plant (3) and / or an interface to a production monitoring system (5) of the production plant (3) for sending / receiving state variables of the production plant (3), quality inspection (6) for detecting product defects in the in the

Products manufactured in the production facility (3) and / or an interface to a quality inspection system (7) for receiving product defects in the products manufactured in the production facility (3), and

Computing device (8) for determining relationships between the detected product defects and the detected state variables of the

Production plant (3),

Monitoring of the state variables of the production plant (3) by means of the sensors (4) and / or via the interface to the production monitoring system (5) during the future manufacture of products and comparisons with the state variables of the determined ones

Correlations between the recorded product defects and the recorded state variables of the production plant (3) by means of the computing device (8), and

Adjustment of the quality inspection (6) for the detection of product defects in the products manufactured in the production plant (3) in the future if the

Comparison of the monitored state variables of the production plant (3) in the future production of products with the state variables of the determined relationships between the detected product defects and the detected state variables of the production plant (3) provides a sufficient correspondence.

27. System (1) according to claim 26, wherein the system (1) is designed to carry out the method according to one of claims 1 to 25.

28. A computer program comprising instructions which, when the program is executed by a computer, cause the computer to execute the method according to one of claims 1 to 25, in particular comprising instructions which cause the system according to claim 26 or claim 27 to use the method according to a the claims 1 to 25 carries out.