DE102018006035A1 - Method for the automated generation of setting marks and for process monitoring in cyclical production processes - Google Patents

Abstract

Are known methods for operating point optimization and quality monitoring based on the generation of a relationship between different setting parameters and sensor data of the production process and the part quality. The disadvantage here is the required cost-intensive measurement of the parts produced to adapt the parameters of the process model.
In order to avoid this costly measurement of the manufactured parts, an automated generation of setting parameters is proposed on the basis of sensor data (2) acquired online. In a first adjustment phase (3), the setting parameters are changed on the basis of a predetermined test plan and a set of online sensor data is recorded for each setting and optimized setting parameters (12) are determined in a module designed to be adaptive. In a second adjustment phase (14), the sensor data obtained from the optimized adjustment parameters can be related to the quality features of the parts produced, so that in the module working phase (15) the respective quality features can be deduced from the sensor data. The process key figures generated from the online sensor data can be used for process monitoring as well as process control (16).
The method can be used to generate setting parameters and process monitoring in all cyclical production processes, such as plastic injection molding, aluminum die casting, spot welding of components.

Description

The invention relates to a method for the automated generation of setting parameters and process monitoring in cyclical production processes according to the preamble of claim 1.

From the literature a variety of methods and process variants for operating point optimization and quality monitoring of cyclical production processes are known. Most of the methods are based on the generation of a relationship between different setting parameters of the cyclical production process and the part quality produced in the production process.

According to the company Engel, it is known to analyze a process during the injection process in real time, the pressure profile over the screw position and to compare the measured values online with a reference cycle. On this basis, the system calculates new process parameters that directly reveal fluctuations in the amount of melt and the viscosity of the material. In the event of deviations from the target values, the relevant process parameters are automatically readjusted. The method is not used to automatically determine the relevant setting parameters of a process via online recorded measurement data, but to reduce fluctuations of the molded part quality on the basis of a reference cycle determined at any time via comparative data. However, finding an operating point of process optimized in terms of stability and / or cycle time is of fundamental importance to any manufacturing process and thus can not be replaced in any way by this process.

Also known is a method of the company Krauss Maffei, which keeps the specified by the setter molding weight in the current production even when parameters change due to external influences. For this purpose, the installer sets the ideal process. The method adjusts the switching point and the holding pressure profile to the present melt viscosity and the current flow resistance in the tool, so that deviations are compensated online and still in the same cycle. Also in this method, from an already optimized (ideal) process, i. starting from an optimized operating point, which is regarded as a desired state. The process eliminates any interference. The presented method does not serve for the automated generation of setting parameters of the process on the basis of sensor data recorded online and the use of these sensor data for automated process monitoring like the method according to the invention. In particular, the method according to the invention significantly reduces the required set-up time of the process and thereby reduces the costs of the process device.

Another method of Wittmann Battenfeld is based on the analysis of target-actual deviations and their assignment to the temperature control circuits in the tool using thermal imaging cameras. Here, too, an already optimized operating point is assumed and the thermal images of different production cycles are recorded and compared with reference images. However, the method for determining the component quality is very complex. The working point of the process can not be optimized.

It is known according to DE 10241746 a method for quality evaluation and process monitoring in cyclical production processes, wherein a distinction is made between a setting phase I, a setting phase II and a working phase and a quality assessment of the products produced in the cyclical production process based on a set of quality characteristics, wherein in the first setting phase an automatic operating point optimization, generation a first training data set, automatic characteristic selection and a self-generating process model, which is adopted in the work phase. In addition to the process model, this includes a quality evaluation module and a process monitoring module and, in the case of impermissible deviations in an adjustment phase II, provides a sampling that leads to the generation of another training data record, with subsequent renewed characteristic selection and adaptation of the self-generating process model, which is subsequently taken over into the work phase. In this method, the operating point is determined by means of a process model, which creates a relationship between the setting parameters of the production process and the part quality. This process is extremely time-consuming and therefore cost-intensive, because the part quality has to be measured for all process settings and all manufactured parts. This is necessary because the process setting process does not have online data acquisition and therefore relies on part measurement.

It is known according to DE 19743600 a method for monitoring a cyclical production process in which signal curves are recorded at several points in the production process and at least one quality statement is derived for the products produced, wherein the waveforms of a process relevant for the production of the process section using envelopes are examined on the basis of permissible signal curves of the decisive for the production of the process section process process codes are determined such that changes correlate the process parameters with changes in quality characteristics of the manufactured products and that in an evaluation phase, the process parameters occurring during the production of the products are assigned to the associated quality characteristics of the products produced and / or used to sort the products produced and / or to regulate the production process.

In DE 19961631 A method and a device for displaying and monitoring at least two functional parameters of a technical system are described, which is a generalization of the method of envelope analysis DE 19743600 corresponds to hyperplane. It is, as well as in analogous such methods of envelope analysis, approximately according to DE 19962967 Assuming that if the process as such runs within preset limits, the quality of the manufactured components will also be within specified tolerances.

In DE 29617200 a device for predictive diagnosis of the current quality of the technical work result of a technical system is described, consisting of a device for the cyclical detection of sets of measured values, the actual values of which the desired quality of the technical work result of the technical system is affected, an investment model, which at least one current set of measured values of the technical system simulates an actual value for the current quality of the technical work result of the technical system, and a device at least for parameterizing the system model, which contains a database for storing selected sets of measured values and associated characteristic values and by successive evaluation the measurement value sets and the associated quality characteristic values generated and / or optimized by at least the parameters of the plant model by an iterative optimization.

According to Michaeli, Bluhm, Vaculik and Wybitul (Kunststoffe 84 (1994)), a method for quality assurance and statistical process monitoring is known, in which signals such as the injection time or the internal molding pressure during injection molding are detected and individual quality characteristics of the molded parts are determined by a process model which process parameters are received, which are calculated from the process curves.

The invention has for its object to provide a method for the automated generation of optimized setting parameters in cyclic production processes with online measurement data acquisition, wherein only the online generated sensor data, as well as on-line measurable variables such as cycle time received in the determination of the setting parameters of the production process. As a result, time-consuming and expensive measuring processes of the manufactured parts and thus the disadvantages of the known methods are avoided. The process setting becomes much easier to handle and use. The process is robust, provides reliable results and can be used in a variety of cyclic production processes. If necessary, the quality features of the parts produced in the production process can be directly linked with the online sensor data and other process data through a second, downstream adjustment phase. This enables a pre-calculation of the part quality with minimal effort. The process key figures generated over the course of time from the sensor data recorded online can be monitored and used for process monitoring and process control.

This inventive task is achieved by a method having the features of claim 1, further developments of the invention are illustrated in the subclaims.

• 1 eine schematische Darstellung des Gesamtaufbaus des Verfahrens zur automatisierten Erzeugung von Einstellparametern und Prozessüberwachung bei zyklischen Produktionsprozessen anhand von online erfassten Sensordaten.
• 2 eine schematische Darstellung der ersten Einstellphase des Verfahrens zur automatisierten Erzeugung von Einstellparametern und Prozessüberwachung bei zyklischen Produktionsprozessen anhand von online erfassten Sensordaten.
• 3 eine schematische Darstellung der zweiten Einstellphase und der Arbeitsphase des Verfahrens zur automatisierten Erzeugung von Einstellparametern und Prozessüberwachung bei zyklischen Produktionsprozessen anhand von online erfassten Sensordaten.

Embodiments of the invention are described below by way of example with reference to the accompanying drawings. Likewise, the features mentioned above and the features which have been further elaborated can each be used individually or in any combination with one another. The mentioned embodiments of the method are not to be understood as exhaustive enumeration, but rather have exemplary character. In detail show:

• 1 a schematic representation of the overall structure of the method for the automated generation of setting parameters and process monitoring in cyclic production processes using online recorded sensor data.
• 2 a schematic representation of the first adjustment phase of the method for the automated generation of setting parameters and process monitoring in cyclic production processes based on sensor data recorded online.
• 3 a schematic representation of the second adjustment phase and the working phase of the method for the automated generation of adjustment parameters and process monitoring in cyclical production processes using online recorded sensor data.

in Abhängigkeit des Produktionszyklus zu erfassen, als auch digitale und kontinuierliche Mess- und generelle Einstellparameter des Produktionsprozesses ${\overrightarrow{PS}}_{\alpha },$

wie beispielsweise Ein- und/oder Umschaltzeiten, Zykluszeiten sowie Druck- und/oder Temperatureinstellungen. Das erfindungsgemäße Verfahren beinhaltet eine erste Einstellphase 3, eine zweite Einstellphase 14 sowie eine Arbeitsphase 15.In 1 the method according to the invention for the generation of setting parameters and for process monitoring in cyclical production processes is shown schematically. It is with 1 denotes a possible cyclical production process. The illustrated method can basically used for the generation of setting parameters and for process monitoring in all cyclical production processes. This concerns, for example, plastic injection molding, aluminum die casting, spot welding of components, to name but a few possible applications. It is necessary that the respective cyclical production process 1 with an online data acquisition 2 is equipped, which allows both signal waveforms of different sensor data $\overrightarrow{\text{S}}\left(\text{t}\right)$

depending on the production cycle, as well as digital and continuous measuring and general setting parameters of the production process ${\overrightarrow{PS}}_{\alpha }.$

such as input and / or switching times, cycle times and pressure and / or temperature settings. The method according to the invention includes a first adjustment phase 3 , a second adjustment phase 14 as well as a work phase 15 ,

mit α = 1,2,...,L die jeweils zu jedem Zyklus α mit einer bestimmten Abtastrate aufgenommenen Sensordaten ${\overrightarrow{S}}_{\alpha }\left(t\right)$

des Produktionsprozesses erfasst und darauf basierend ein Satz optimaler Einstellparameter ${\overrightarrow{PS}}_{opt}$

auf Basis dieser Daten für den zyklischen Produktionsprozess generiert. Dabei kann erfindungsgemäß als ein Optimierungskriterium der Einstellparameter die Prozessstabilität, bezogen auf die aufgenommenen Sensordaten gewählt werden, wobei es erfindungsgemäß denkbar und möglich ist, zusätzlich Einstelldaten und Signalverläufe aus anderen zyklischen Produktionsprozessen, die in einer Datenbank oder Cloud 8 gespeichert sind über lernfähige Verfahren, die in 3 integriert sind, in der Analyse und Optimierung der Einstellparameter zu berücksichtigen. Hierdurch können erfindungsgemäß wichtige Erfahrungen, die in einem Zusammenhang mit dem Produktionsprozess stehen zur Stärkung der Aussagekraft und Stabilität der optimalen Einstellung beitragen.In the first adjustment phase 3 of the cyclical production process are used for a number of different process setting parameters ${\overrightarrow{PS}}_{\alpha }$

with α = 1, 2,..., L the sensor data recorded in each cycle α at a specific sampling rate ${\overrightarrow{S}}_{\alpha }\left(t\right)$

of the production process and based on this a set of optimal adjustment parameters ${\overrightarrow{PS}}_{Opt}$

generated on the basis of this data for the cyclical production process. In this case, according to the invention, the process stability, based on the recorded sensor data, can be selected as an optimization criterion of the adjustment parameters, it being conceivable and possible according to the invention to additionally set adjustment data and signal curves from other cyclical production processes, which are stored in a database or cloud 8th are stored via adaptive methods that are in 3 integrated into the analysis and optimization of the setting parameters. As a result, according to the invention, important experiences which are connected with the production process can contribute to strengthening the informational value and stability of the optimal setting.

For generating the optimized setting parameters 12 it is not required according to the invention in the cyclical production process 1 to measure finished parts. This means a significant time and cost savings compared to prior art methods and significantly supports the acceptance of the method according to the invention. In particular, the measurement of parts is usually time-critical and perform in a measuring room or laboratory. Therefore, empirical values have often been used in practice to date and the production process has been started with non-optimized process setting parameters and no possible time, energy and cost savings have been made. The optimized setting parameters 12 can directly into the work phase 15 be taken over.

online gewonnenen Messdaten ${\overrightarrow{S}}_{opt}\left(t\right),$

bzw. aus von diesen Messdaten über eine automatisierte Selektion gewonnenen dynamischen Kenngrößen ${\overrightarrow{SYN}}_{optj},$

mit den N Qualitätsmerkmalen ${\overrightarrow{Q}}_{opt}={\left(Q_1,Q_2,\dots ,Q_N\right)}_{opt}$

des im zyklischen Produktionsprozess 1 erzeugten Produkts in Beziehung gebracht werden, so dass im Modul Arbeitsphase 15 aus den online erfassten Sensordaten $\overrightarrow{S}\left(t\right)$

und den daraus online erzeugten dynamischen Kenngrößen ${\overrightarrow{SYN}}_{opt\gamma },$

wobei γ die einzelnen zyklischen Produktionsprozesse in der Arbeitsphase kennzeichnet, in einem Modul Qualitätsanalyse 22 auf die jeweiligen Qualitätsmerkmale ${\overrightarrow{Q}}_{\gamma }={\left(Q_1,Q_2,\dots ,Q_M\right)}_{\gamma }$

geschlossen werden kann und eine entsprechende Meldung über die prognostizierte Teilequalität erfolgt und/oder Signale zu einer Weiterverarbeitung der Information ausgegeben werden, etwa zur Ansteuerung einer Ausschussweiche 17. Erfindungsgemäß erfolgt der Aufbau des Moduls Qualitätsanalyse 22 mittels eines selbstgenerierenden neuronalen Netzwerks, so dass der Anwender nicht in die sich entwickelnde Netzwerkarchitektur eingreifen oder Parameter zu dessen Aufbau vorgeben muss.According to the invention, however, it is also conceivable and possible, if desired, directly to the first adjustment phase 3 a second adjustment phase 14 and connect to the optimized setting parameters 12 . ${\overrightarrow{PS}}_{Opt}$

online measured data ${\overrightarrow{S}}_{Opt}\left(t\right).$

or from these measured data via an automated selection obtained dynamic characteristics ${\overrightarrow{SYN}}_{Optj}.$

with the N quality features ${\overrightarrow{Q}}_{Opt}={\left({\mathrm{<figure-callout\; id="1"\; label="Q"\; filenames="DE102018006035A1\_0091.png,DE102018006035A1\_0092.png"\; state="\{\{state\}\}">Q</figure-callout>}}_1.{\mathrm{<figure-callout\; id="2"\; label="Q"\; filenames="DE102018006035A1\_0091.png,DE102018006035A1\_0092.png"\; state="\{\{state\}\}">Q</figure-callout>}}_2.....Q_N\right)}_{Opt}$

of the cyclical production process 1 produced product, so that in the module working phase 15 from the online sensor data $\overrightarrow{S}\left(t\right)$

and the dynamic parameters generated online ${\overrightarrow{SYN}}_{Opt\gamma }.$

where γ characterizes the individual cyclical production processes in the working phase, in a module quality analysis 22 on the respective quality characteristics ${\overrightarrow{Q}}_{\gamma }={\left({\mathrm{<figure-callout\; id="1"\; label="Q"\; filenames="DE102018006035A1\_0091.png,DE102018006035A1\_0092.png"\; state="\{\{state\}\}">Q</figure-callout>}}_1.{\mathrm{<figure-callout\; id="2"\; label="Q"\; filenames="DE102018006035A1\_0091.png,DE102018006035A1\_0092.png"\; state="\{\{state\}\}">Q</figure-callout>}}_2.....Q_M\right)}_{\gamma }$

can be closed and a corresponding message about the predicted part quality is carried out and / or signals are issued for further processing of the information, such as to control a reject gate 17 , According to the construction of the module quality analysis 22 by means of a self-generating neural network, so that the user does not have to intervene in the developing network architecture or must specify parameters for its construction.

und/oder relevanten dynamischen Kenngrößen ${\overrightarrow{SYN}}_{\gamma }$

zu aktivieren, um damit die Einstellparameter des zyklischen Produktionsprozesses 1 systematisch zu verändern, so dass die gewünschten Qualitätsmerkmale innerhalb der vorgegebenen Toleranzwerte bleiben.According to the invention is also conceivable and possible on the module working phase 15 a process control 16 Significant changes in the process-related sensor data ${\overrightarrow{S}}_{\gamma }\left(t\right)$

and / or relevant dynamic characteristics ${\overrightarrow{SYN}}_{\gamma }$

in order to activate the setting parameters of the cyclical production process 1 systematically change so that the desired quality characteristics remain within the specified tolerance values.

mit α = 1,2,...,L die jeweils zu jedem Zyklus, das heißt zu jeder Prozesseinstellung α und zu jedem innerhalb einer Prozesseinstellung gefertigten Teil mit einer bestimmten Abtastrate Sensordaten ${\overrightarrow{S}}_{\alpha }\left(t\right)$

wie Druck- und Temperaturkurven aber auch Messgrößen wie Zykluszeit τ_α und/oder Maschinendaten, die im folgenden unter der Bezeichnung ${\overrightarrow{P}}_{\alpha }\left(t\right)$

zusammengefasst werden und die zugeordneten Einstellparameter ${\overrightarrow{PS}}_{\alpha }$

aufgenommenen. Weitere digitale Mess- und Einstellgrößen, die während der Einstellphase nicht verändert werden, jedoch für den Produktionsprozess und dessen Charakterisierung von Bedeutung sein können werden ebenfall erfasst und zusammen mit den Sensordaten $\overrightarrow{S_{\alpha }}\left(t\right),$

weiteren Prozessdaten ${\overrightarrow{P}}_{\alpha }\left(t\right)$

und den Einstellparametern ${\overrightarrow{PS}}_{\alpha }$

in einer Datenbank oder einem sonstigen Speichermedium 9 abgelegt. In diesem Speichermedium 9 können sämtliche während der Einstellphasen I und II registrierten, erzeugten oder veränderten Daten abgelegt werden und bei Bedarf für eine nachhaltige und weitere Auswertung an die Datenbank oder Cloud 8 weiter transferiert werden. 2 shows a schematic representation of the method for the automated generation of setting parameters (module adjustment phase I) 3 and process monitoring in cyclical production processes. By means of online data acquisition 2 be during the adjustment phase 3 the cyclical production process for a number of different process setting parameters ${\overrightarrow{PS}}_{\alpha }$

with α = 1, 2,..., L the sensor data produced in each case for each cycle, that is to say for each process setting α and for each part produced within a process setting, with a specific sampling rate ${\overrightarrow{S}}_{\alpha }\left(t\right)$

such as pressure and temperature curves but also variables such as cycle time τ _α and / or machine data, which in the following under the name ${\overrightarrow{P}}_{\alpha }\left(t\right)$

summarized and the assigned adjustment parameters ${\overrightarrow{PS}}_{\alpha }$

recorded. Other digital measurement and setting variables that are not changed during the adjustment phase, but can be of importance for the production process and its characterization are also recorded and together with the sensor data $\overrightarrow{S_{\alpha }}\left(t\right).$

further process data ${\overrightarrow{P}}_{\alpha }\left(t\right)$

and the setting parameters ${\overrightarrow{PS}}_{\alpha }$

in a database or other storage medium 9 stored. In this storage medium 9 All stored, modified or modified data can be stored during the adjustment phases I and II and, if required, for sustained and further evaluation to the database or cloud 8th be transferred further.

mit α = 1,2,...,Z für die Ermittlung der optimalen Einstellparameter ${\overrightarrow{PS}}_{opt}$

auf Basis dieser Daten, kann erfindungsgemäß über ein herkömmliches Modul zur Versuchsplangenerierung 4 erfolgen oder direkt vom Bediener/Einrichter 24 des Prozesses 5 vorgegeben werden. Ebenso kann die Anzahl der zu fertigenden Teile M bei gleicher Prozesseinstellung α vom Bediener/Einrichter vorgegeben werden oder über das Modul 4 nach statistischen Verfahren ermittelt werden.The number of different process settings L and the choice of different setting parameters ${\overrightarrow{PS}}_{\alpha }.$

with α = 1,2, ..., Z for the determination of the optimal adjustment parameters ${\overrightarrow{PS}}_{Opt}$

Based on this data, can according to the invention via a conventional module for experimental plan generation 4 or directly from the operator / installer 24 of the process 5 be specified. Likewise, the number of parts to be manufactured M can be specified by the operator / setter at the same process setting α or via the module 4 be determined by statistical methods.

mit α = 1,2,...,L wird der zyklische Produktionsprozess 5 (Einstellphase I) gestartet und es werden erfindungsgemäß zu jedem Fertigungszyklus, das heißt zu jeder Prozesseinstellung α und jedem gefertigten Teil mittels der online Messdatenerfassung 6 (Einstellphase I), die mit einer bestimmten Abtastrate aufgenommenen Sensordaten ${\overrightarrow{S}}_{\alpha }\left(t\right),$

weiteren Prozessdaten ${\overrightarrow{P}}_{\alpha }\left(t\right)$

und zugeordneten Einstellparameter ${\overrightarrow{PS}}_{\alpha }$

aufgezeichnet und im Datenträger 9 abgespeichert.After specification of the setting parameters ${\overrightarrow{PS}}_{\alpha }$

with α = 1,2, ..., L becomes the cyclical production process 5 (Adjustment phase I) is started and it is according to the invention for each production cycle, that is to each process setting α and each manufactured part by means of the online measurement data acquisition 6 (Adjustment phase I), the sensor data recorded at a certain sampling rate ${\overrightarrow{S}}_{\alpha }\left(t\right).$

further process data ${\overrightarrow{P}}_{\alpha }\left(t\right)$

and associated adjustment parameters ${\overrightarrow{PS}}_{\alpha }$

recorded and in the disk 9 stored.

eines gefertigten Teils mittels eines Expertensystems im Zeitbereich in charakteristische Segmente unterteilt. Dynamische Merkmale des Sensorsignals $\overrightarrow{S}\left(t\right),$

beispielsweise Extremwerte, Ableitungen, Einzelwerte als auch Integrale über Flächenabschnitte bestimmter Zeitbereiche der Sensorsignale $\overrightarrow{S}\left(t\right)$

oder charakteristische Frequenzen und Zeiten werden berechnet. Für jedes einzelne Sensorsignal $\overrightarrow{S}\left(t\right)$

erhält man damit einen Satz dynamischer Kenngrößen ${\overrightarrow{SYN}}_{\alpha j,}$

wobei α die jeweilige Prozesseinstellung α = 1,2,...,L bezeichnet und j, j = 1,2,...,M die Anzahl an gefertigten Teilen je Prozesseinstellung α angibt. Nachdem entsprechend dem Modul Versuchsplangenerierung 4 oder nach hinreichend vielen Einstellungen, ausgelöst durch den Bediener/Einrichter für sämtliche L gewünschten Einstellparameter ${\overrightarrow{PS}}_{\alpha }$

Teile gefertigt und Sensorsignale $\overrightarrow{S}\left(t\right)$

und weitere Prozessdaten ${\overrightarrow{P}}_{\alpha }\left(t\right)$

aufgenommen und auch die daraus ermittelten dynamischen Kenngrößen ${\overrightarrow{SYN}}_{\alpha j}$

der Sensorsignale und deren zugeordnete Mittelwerte ${\overrightarrow{SYN}}_{\alpha }$

über die einzelnen Produktionszyklen j, j =1,2,...,M ermittelt wurden, erfolgt erfindungsgemäß im Modul Einstellparameteroptimierung 11 die Ermittlung des Satzes optimaler Einstellparameter ${\overrightarrow{PS}}_{opt}$

des zyklischen Produktionsprozesses 1 auf Basis der online Messdatenerfassung 2.In the module Process Codes 7 According to the invention, the sensor signals $\overrightarrow{S}\left(t\right)$

of a manufactured part by means of an expert system in the time domain divided into characteristic segments. Dynamic features of the sensor signal $\overrightarrow{S}\left(t\right).$

For example, extreme values, derivatives, individual values as well as integrals over surface portions of certain time ranges of the sensor signals $\overrightarrow{S}\left(t\right)$

or characteristic frequencies and times are calculated. For every single sensor signal $\overrightarrow{S}\left(t\right)$

This gives you a set of dynamic characteristics ${\overrightarrow{SYN}}_{\alpha j.}$

where α denotes the respective process setting α = 1,2, ..., L and j, j = 1,2, ..., M indicates the number of manufactured parts per process setting α. After according to the module design layout 4 or after a sufficient number of settings, triggered by the operator / installer for all L desired setting parameters ${\overrightarrow{PS}}_{\alpha }$

Parts made and sensor signals $\overrightarrow{S}\left(t\right)$

and further process data ${\overrightarrow{P}}_{\alpha }\left(t\right)$

and also the dynamic parameters determined therefrom ${\overrightarrow{SYN}}_{\alpha j}$

the sensor signals and their associated averages ${\overrightarrow{SYN}}_{\alpha }$

determined according to the individual production cycles j, j = 1,2, ..., M according to the invention in the module Einstellparameteroptimierung 11 the determination of the set of optimal adjustment parameters ${\overrightarrow{PS}}_{Opt}$

the cyclical production process 1 based on online data acquisition 2 ,

der Sensorsignale auf Basis eines lernfähigen Verfahrens, etwa eines neuronalen Netzwerks oder eines anderen mathematischen Verfahrens derart, dass die Einstellparameter ${\overrightarrow{PS}}_{\alpha }$

des zyklische Produktionsprozess 1 im Hinblick auf auswählbare Optimierungskriterien wie beispielsweise Prozessstabilität und/oder Zykluszeit τ_α oder diese nach Bedeutung gewichtet, ermittelt werden. Da je nach Anwendung unterschiedliche zeitliche Bereiche der Sensorsignale von Bedeutung sein können, werden erfindungsgemäß die einzelnen dynamischen Kenngrößen ${\overrightarrow{SYN}}_{\alpha j},$

über ein statistisches und/oder lernfähiges Verfahren unterschiedlich gewichtet und damit die jeweilige Streuung Σ²SYN_α der dynamischen Kenngrößen ${\overrightarrow{SYN}}_{\alpha j},$

in den einzelnen zeitlichen Abschnitten der Sensorsignale $\overrightarrow{S}\left(t\right)$

nach Bedeutung für die Stabilität des zyklischen Produktionsprozesses gewichtet.According to the invention, setting parameter generation takes place in the module 11 an analysis of the determined dynamic characteristics ${\overrightarrow{SYN}}_{\alpha }$

the sensor signals on the basis of an adaptive method, such as a neural network or other mathematical method such that the adjustment parameters ${\overrightarrow{PS}}_{\alpha }$

the cyclical production process 1 in terms of selectable optimization criteria such as process stability and / or cycle time τ _α or weighted by importance, are determined. Since, depending on the application, different temporal ranges of the sensor signals can be of importance, according to the invention the individual dynamic characteristics become ${\overrightarrow{SYN}}_{\alpha j}.$

weighted differently by a statistical and / or learning method and thus the respective scatter Σ ² SYN _{α of} the dynamic characteristics ${\overrightarrow{SYN}}_{\alpha j}.$

in the individual temporal sections of the sensor signals $\overrightarrow{S}\left(t\right)$

weighted by importance for the stability of the cyclical production process.

nicht mehr den möglichen Einstellwerten, gemäß dem Modul Versuchsplangenerierung 4 oder den durch den Bediener/Einrichter vorgegebenen Einstellungen. Vielmehr ist es denkbar und möglich, dass durch das erfindungsgemäße Verfahren ein zuvor noch nicht ausgeführter oder bekannter Satz an Einstellparametern ${\overrightarrow{PS}}_{opt,}$

ermittelt wird. Die optimierten Einstellparameter ${\overrightarrow{PS}}_{opt}$

werden in der Meldeeinrichtung 12 ausgegeben und in der Datenbank oder einem sonstigen Speichermedium 9 bzw. in der Cloud 8 abgespeichert.Through the learning process in 11 and the possible use of further process data from earlier production processes 8th or 9 , correspond to the optimized setting parameters ${\overrightarrow{PS}}_{Opt}$

no longer the possible setting values, according to the Module Experimental Generation 4 or the settings specified by the operator / installer. Rather, it is conceivable and possible that by the method according to the invention a previously not yet executed or known set of setting parameters ${\overrightarrow{PS}}_{Opt.}$

is determined. The optimized setting parameters ${\overrightarrow{PS}}_{Opt}$

be in the notification device 12 issued and in the database or other storage medium 9 or in the cloud 8th stored.

können entweder direkt als Eingangsgrößen in der Arbeitsphase 15 des zyklischen Produktionsprozesses verwendet werden oder es ist denkbar und möglich zusätzlich bei Bedarf eine zweite Einstellphase 14 an die erste Einstellphase 3 anzuschließen und die zu den optimierten Einstellparametern ${\overrightarrow{PS}}_{opt}$

online gewonnenen Messdaten $\overrightarrow{S}\left(t\right)$

und/oder dynamischen Kenngrößen ${\overrightarrow{SYN}}_{opt\gamma }$

mit den N Qualitätsmerkmalen ${\overrightarrow{Q}}_{opt}={\left(Q_1,Q_2,\dots ,Q_N\right)}_{opt}$

der im zyklischen Produktionsprozess 1 erzeugten Teile über ein mathematisches Verfahren in Beziehung gebracht werden, so dass im Modul Arbeitsphase 15 aus den online erfassten Sensordaten ${\overrightarrow{S}}_{\gamma }\left(t\right)$

und den daraus online erzeugten Prozesskenngrößen ${\overrightarrow{SYN}}_{opt\gamma }$

wobei γ die einzelnen zyklischen Produktionsprozesse in der Arbeitsphase kennzeichnet, in einem Modul Qualitätsanalyse 22 auf die jeweiligen Qualitätsmerkmale ${\overrightarrow{Q}}_{\gamma }={\left(Q_1,Q_2,\dots ,Q_N\right)}_{\gamma }$

geschlossen werden kann und beim Unter- oder Überschreiten vorgegebener Grenzwerte der einzelnen Prozesskennzahlen und/oder Qualitätsmerkmale eine Meldung und die Ansteuerung einer Ausschussweiche 17 erfolgt. 3 shows a schematic representation of the method for the automated generation of setting parameters (adjustment phase II) and process monitoring in cyclic production processes using online recorded sensor data. The first adjustment phase in the module 3 , determined optimized setting parameters ${\overrightarrow{PS}}_{Opt}$

can either directly as input variables in the work phase 15 be used in the cyclical production process or it is conceivable and possible in addition, if necessary, a second adjustment phase 14 to the first adjustment phase 3 and connect to the optimized setting parameters ${\overrightarrow{PS}}_{Opt}$

online measured data $\overrightarrow{S}\left(t\right)$

and / or dynamic characteristics ${\overrightarrow{SYN}}_{Opt\gamma }$

with the N quality features ${\overrightarrow{Q}}_{Opt}={\left({\mathrm{<figure-callout\; id="1"\; label="Q"\; filenames="DE102018006035A1\_0091.png,DE102018006035A1\_0092.png"\; state="\{\{state\}\}">Q</figure-callout>}}_1.{\mathrm{<figure-callout\; id="2"\; label="Q"\; filenames="DE102018006035A1\_0091.png,DE102018006035A1\_0092.png"\; state="\{\{state\}\}">Q</figure-callout>}}_2.....Q_N\right)}_{Opt}$

in the cyclical production process 1 produced parts are related via a mathematical method, so that in the module working phase 15 from the online sensor data ${\overrightarrow{S}}_{\gamma }\left(t\right)$

and the process parameters generated online ${\overrightarrow{SYN}}_{Opt\gamma }$

where γ characterizes the individual cyclical production processes in the working phase, in a module quality analysis 22 on the respective quality characteristics ${\overrightarrow{Q}}_{\gamma }={\left({\mathrm{<figure-callout\; id="1"\; label="Q"\; filenames="DE102018006035A1\_0091.png,DE102018006035A1\_0092.png"\; state="\{\{state\}\}">Q</figure-callout>}}_1.{\mathrm{<figure-callout\; id="2"\; label="Q"\; filenames="DE102018006035A1\_0091.png,DE102018006035A1\_0092.png"\; state="\{\{state\}\}">Q</figure-callout>}}_2.....Q_N\right)}_{\gamma }$

can be closed and when undershooting or exceeding predetermined limits of the individual process indicators and / or quality features a message and the control of a reject gate 17 he follows.

und die zugehörigen relevanten Prozesskenngrößen ${\overrightarrow{SYN}}_{opt\gamma }$

werden in Modul 23 in ihrer zeitlichen Entwicklung bewertet. Werden keine signifikanten Veränderungen in den relevanten Prozesskenngrößen ${\overrightarrow{SYN}}_{opt\gamma }$

oder Qualitätsmerkmalen ${\overrightarrow{Q}}_{\gamma }={\left(Q_1,Q_2,\dots ,Q_N\right)}_{\gamma }$

durch das Modul Prozessüberwachung 23 ermittelt, wird der nächste Produktionszyklus freigegeben. Erfindungsgemäß ist es denkbar und möglich bei einer signifikanten Veränderung in den relevanten Prozesskenngrößen, etwa durch Prozessdriften, Chargenschwankungen oder Änderungen in den Umgebungsbedingungen, je nach zugrundeliegendem Produktionsprozess, ein Steuersignal für eine Prozessregelung 16 auszugeben.A monitoring of the cyclical production process takes place in module 23 , The results of the process monitoring are in 9 logged. The during the cyclical production process with the help of the module quality analysis 22 registered discrete and / or continuous quality characteristics ${\overrightarrow{Q}}_{\gamma }={\left({\mathrm{<figure-callout\; id="1"\; label="Q"\; filenames="DE102018006035A1\_0091.png,DE102018006035A1\_0092.png"\; state="\{\{state\}\}">Q</figure-callout>}}_1.{\mathrm{<figure-callout\; id="2"\; label="Q"\; filenames="DE102018006035A1\_0091.png,DE102018006035A1\_0092.png"\; state="\{\{state\}\}">Q</figure-callout>}}_2.....Q_N\right)}_{\gamma }$

and the associated relevant process parameters ${\overrightarrow{SYN}}_{Opt\gamma }$

be in module 23 evaluated in their temporal evolution. Are no significant changes in the relevant process parameters ${\overrightarrow{SYN}}_{Opt\gamma }$

or quality characteristics ${\overrightarrow{Q}}_{\gamma }={\left({\mathrm{<figure-callout\; id="1"\; label="Q"\; filenames="DE102018006035A1\_0091.png,DE102018006035A1\_0092.png"\; state="\{\{state\}\}">Q</figure-callout>}}_1.{\mathrm{<figure-callout\; id="2"\; label="Q"\; filenames="DE102018006035A1\_0091.png,DE102018006035A1\_0092.png"\; state="\{\{state\}\}">Q</figure-callout>}}_2.....Q_N\right)}_{\gamma }$

through the process monitoring module 23 determined, the next production cycle is released. According to the invention, it is conceivable and possible with a significant change in the relevant process parameters, for example by process drifts, batch fluctuations or changes in the ambient conditions, depending on the underlying production process, a control signal for a process control 16 issue.

QUOTES INCLUDE IN THE DESCRIPTION

This list of the documents listed by the applicant has been generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Cited patent literature

• DE 10241746 [0006]
• DE 19743600 [0007, 0008]
• DE 19961631 [0008]
• DE 19962967 [0008]
• DE 29617200 [0009]

Claims (8)

Method for the automated generation of setting parameters and process monitoring in cyclical production processes using sensor data recorded online, distinguishing between a setting phase I, a setting phase II and a working phase and a quality analysis of the manufactured parts based on process metrics that generated automatically from the online recorded measurement data be, characterized in that in a setting phase I, an automated generation of the setting parameters of the cyclical production process in a module (3) is such that on the basis of an automatically predetermined test plan (4) or individual settings by the tuner (24) of the process the setting parameters are changed and for each setting a set of representative online measured data is taken over a module (6) and after sufficiently many settings α = 1,2, ..., L of the cyclical production process (1) in a learning configured module (7) a data evaluation of the sensor data recorded at a certain sampling rate $\overrightarrow{S}\left(t\right)$

and further process data ${\overrightarrow{P}}_{\alpha }\left(t\right)$

and automatically generated process metrics ${\overrightarrow{SYN}}_{\alpha j}$

such that optimized setting parameters ${\overrightarrow{PS}}_{Opt}$

with regard to selectable optimization criteria, such as process stability and / or cycle time, or these are weighted according to importance, whereby information of further stored data from cyclical production processes (8) can also be included in the optimization of the adjustment parameters and in an adjustment phase II, which is based on the first setting phase, which leads to the optimized setting parameters ${\overrightarrow{PS}}_{Opt}$

online from the recorded sensor data ${\overrightarrow{S}}_{\gamma }\left(t\right)$

obtained associated relevant process parameters ${\overrightarrow{SYN}}_{Opt\gamma }$

with the quality characteristics ${\overrightarrow{Q}}_{\gamma }={\left(Q_1.Q_2.....Q_N\right)}_{\gamma }$

the phase generated in the cyclical production process, and a module (15) working phase, which is controlled either directly after adjustment phase I or after adjustment phase II, so that in the module working phase (15) from the online recorded sensor data and other process data and the From this, process key figures generated in a quality analysis module (22) based on the process parameters determined from the sensor data ${\overrightarrow{SYN}}_{Opt\gamma }$

and / or by means of the associated quality features ${\overrightarrow{Q}}_{\gamma }={\left(Q_1.Q_2.....Q_N\right)}_{\gamma }$

directly or indirectly to the respective quality features can be concluded and when falling below or exceeding predetermined limits of the individual process indicators and / or quality features a message and the control of a rejects (17) and takes place when leaving a predetermined process window, a process control. Method according to Claim 1 , characterized in that for the automated generation of setting parameters and process monitoring in cyclic production processes in a learning configured module process codes (7), the sensor signals $\overrightarrow{S}\left(t\right)$

a manufactured part are divided into characteristic segments by means of an expert system in the time domain and dynamic features of the sensor signal $\overrightarrow{S}\left(t\right).$

For example, extreme values, derivatives, individual values as well as integrals over surface portions of certain time ranges of the sensor signals $\overrightarrow{S}\left(t\right)$

or characteristic frequencies and times are calculated and for each individual sensor signal $\overrightarrow{S}\left(t\right)$

a set of dynamic characteristics ${\overrightarrow{SYN}}_{\alpha j}$

where α denotes the respective process setting α = 1,2,..., L, and j, j = 1,2,..., M indicates the number of manufactured parts per process setting α and according to the module design of experimentation (FIG. 4) or after a sufficient number of settings, triggered by the operator / setter (24) for all L desired setting parameters ${\overrightarrow{PS}}_{\alpha }$

Parts made and sensor signals $\overrightarrow{S}\left(t\right)$

and further process data ${\overrightarrow{P}}_{\alpha }\left(t\right)$

and also the dynamic parameters determined therefrom ${\overrightarrow{SYN}}_{\alpha j}$

the sensor signals and their associated averages ${\overrightarrow{SYN}}_{\alpha }$

and further statistical moments were determined over the individual production cycles j, j = 1, 2,..., M, the module Setting parameter optimization (11) determines the set of optimum setting parameters ${\overrightarrow{PS}}_{Opt}$

the cyclical production process (5) based on the online data acquisition (6), with an analysis of the determined dynamic characteristics ${\overrightarrow{SYN}}_{\alpha }.{\displaystyle \Sigma {}^{2}SY{N}_{\alpha }}$

the sensor signals on the basis of an adaptive method, such as a neural network or other mathematical method is such that the setting parameters ${\overrightarrow{PS}}_{\alpha }$

the cyclical production process (1) with respect to selectable optimization criteria such as process stability and / or cycle time τ _α or weighted by meaning, are determined, wherein the individual dynamic characteristics ${\overrightarrow{SYN}}_{\alpha }.{\displaystyle \Sigma {}^{2}SY{N}_{\alpha }}$

about a statistical and / or or learnable method with different weighting in the determination of the optimized setting parameters (12), since depending on the application, different temporal ranges of the sensor signals may be of importance. Method according to Claim 1 or 2 , characterized in that a module second adjustment phase (14) to the module first adjustment phase (3) and connects to the optimized adjustment parameters ${\overrightarrow{PS}}_{Opt}$

online measured data ${\overrightarrow{S}}_{Opt}\left(t\right)$

and / or dynamic characteristics ${\overrightarrow{SYN}}_{Opt\gamma }$

with the N quality features ${\overrightarrow{Q}}_{Opt}={\left(Q_1.Q_2.....Q_N\right)}_{Opt}$

the parts produced in the cyclical production process are related via a mathematical method, so that in the module working phase (15) from the sensor data recorded online ${\overrightarrow{S}}_{\gamma }\left(t\right)$

and the process parameters generated online ${\overrightarrow{SYN}}_{Opt\gamma }$

where γ characterizes the individual cyclical production processes in the working phase, in a module quality analysis (22) on the respective quality characteristics ${\overrightarrow{Q}}_{\gamma }={\left(Q_1.Q_2.....Q_N\right)}_{\gamma }$

can be closed and when undershooting or exceeding predetermined limits of the individual process indicators and / or quality features a message and the control of a reject gate (17) takes place. Method according to one of Claims 1 to 3 , characterized in that in the modules (11) Einstellparametergenierung, and / or module (19) quality characteristics, and / or module (22) quality analysis, self-generating neural networks are used. Method according to one of Claims 1 to 4 , characterized in that the module setting phase I (3) at periodic intervals during the working phase of the cyclical production process and thus track the operating point by the respectively determined optimized setting parameters and when undershooting or exceeding permissible changes a message is triggered. Method according to one of Claims 1 to 5 , characterized in that the module setting phase II (14) at periodic intervals during the working phase of the cyclical production process and thus tracked the operating point based on the quality features by the respectively determined optimized setting parameters and when exceeding or exceeding admissible changes a message is triggered. Method according to one of Claims 1 to 6 , characterized in that a module process control (16) is connected to the module working phase (15) and when leaving a process window, which by the dynamic characteristics ${\overrightarrow{SYN}}_{Opt}.{\displaystyle \Sigma {}^{2}SY{N}_{Opt}}$

is determined, a process control is carried out, the different time scales of the cyclical production process are taken into account in changes of individual adjustment parameters in the control strategy, so that the operating point is adiabatically approximated to the optimum operating point. Method according to one of Claims 1 to 7 , characterized in that in the module setting phase I (3) information of other cyclical production processes stored in a database or cloud (8) are included in the optimization and the relevant data of the cyclical production process for further analysis in a database or cloud ( 8) are stored.

Images