Sensorlose online Zustandserfassung von Vorschubantriebskomponenten in Werkzeugmaschinen
| Autor: | Maier, Detlef |
|---|---|
| Přispěvatelé: | Bertsche, Bernd (Prof. Dr.-Ing.) |
| Jazyk: | němčina |
| Rok vydání: | 2015 |
| Předmět: | |
| DOI: | 10.18419/opus-4595 |
| Popis: | Die Verfügbarkeit ist wesentlich für die Profitabilität kapitalintensiver Fertigungsanlagen. Bei der Just-in-time-Produktion wirken sich Störungen schnell entlang der Lieferkette aus. Die zustandsabhängige Wartung als eine Instandhaltungsstrategie, die die Verfügbarkeit hoch hält und sich Schwankungen der Produktionsauslastung anpasst, gewinnt zunehmend an Bedeutung. Die ubiquitäre Informationstechnologie liefert die Plattform, aufwändige Strategien auf Bauelemente anzuwenden, bei denen Wirtschaftlichkeitserwägung dies bisher verhinderten. Kern der Zustandserkennung ist die Signalanalyse. Sie hängt von der Güte der Messdaten ab. Welche Güte benötigt wird, bestimmt der zugrunde liegende Abnutzungsvorgang, dessen Signalmoden identifiziert werden müssen. Hier wird nach der Betrachtung von linearen und empirischen Transformationen zur Modentrennung die Signalanalyse zu einem gänzlich neuen, das Messsignal linearisierende, raum- und auflösungsbasierten Konzept, das Modell, Messung, Datenverarbeitung und Auswertung harmonisiert, zusammengeführt. Das kinematische Verhalten von Vorschubantrieben lässt dabei sich sehr gut unabhängig von der Zeit, rein geometrisch beschreiben. Eine geometrische, physikalische Variable aus dem System als Führungsgröße der Messungen liefert unverzerrte Signale, die eine exakte, hochauflösende und detaillierte Bauteilerkennung und Abnutzungsbewertung durch eine sehr scharfe und robuste Modentrennung zulassen. Die Zeitunabhängigkeit ermöglicht es, gezielt instationäre und transiente Zustände unter Betriebsbedingungen zu betrachten. Die Erfassung von Abnutzung allein ist für zustandsabhängige Wartung nicht ausreichend. Vielmehr muss das langfristige Verhalten einer Kenngröße für eine Extrapolation zur Abschätzung der zukünftigen Entwicklung geeignet sein. Nur so lassen sich Instandhaltungsmaßnahmen zustandsabhängig terminieren. Wichtig ist die Einbettung des Konzepts in den Gesamtkontext des Instandhaltungsmanagements bzw. der Instandhaltungsstrategie. Der Kostendruck ist eine treibende Kraft. Ein weiteres Leitmotiv ist daher die Beschränkung auf die im Vorschubantrieb bereits vorhandenen mechatronischen Ressourcen. D.h. Ziel ist, den Verschleißzustand ohne den Einsatz zusätzlicher Sensorik und Hardware während des Regelbetriebs der Maschine zu erfassen. Das letztlich angestrebte Ziel lässt sich in das Schlagwort „Sensorless Online Condition Monitoring“ (SOCM) fassen. Condition Based Maintenance is a strategy to keep the availability up. Because it is capable to adopt to fluctuations in capacity utilization rates it’s gaining more and more acceptance. This PhD thesis deals with Condition Based Maintenance (CBM) of machine tools especially Condition Monitoring (CM) of feed drives. Every specific CM has to embed completely into CBM to provide appropriate data that can be processed by CBM. High data quality means that wear effects can be exactly quantified and affected machine elements can be identified. Further, for cost effective timing of maintenance measures the strategy of CBM requires forecasting of future wear rates sometimes even under changing boundary conditions such as different loads, processes and work pieces. Here, in detail feed drives with ball screws are examined exemplarily. A CM system itself is subject to cost effectiveness and reliability. At a mechatronic system such as machine tools, monitoring might be a part of the whole functionalities. Also, data might be acquitted during the normal operation of the machine without interrupting the production output. The CM system has to be resistant against signal noise and harsh running conditions. High price assessments and additional hardware can barely be asserted. That’s why a CM system must embed into the machine’s controller, its given platform. The above aspects define the requirements for the CM system. Identification of the affected elements is achieved by mode separation, forecasting by extrapolating signal courses. The present paper gives an overview of the state of the art that is mainly based on bearing condition monitoring means. It shows new approaches to interpret rough and noised data by nonlinear filtering, the Hilbert Huang Transformation (HHT) and Empirical Mode decomposition (EMD). Designing a CM system involves numerous disciplines. Finally a solution is presented that is based on complying consistently with the aforementioned requirements in every development step such as planning, modeling, measuring, mode separation, signal analysis and interpretation. The long-term behaviour of the formed parameter must be suitable for forecasting of the remaining lifetime. During the monitoring steady boundary conditions might not be assumed. Nonetheless, the CM has definitely to be able to separate, to assign and to quantify signal modes of wear of different specific machine elements. CBM requires prediction. Thus recorded data or signals have to be wear-sensitive in an appropriate way. Signals that are mainly insusceptible, alternating or incidentally rise in sudden steps are not appropriate for extrapolation. As well are dissipating wear effects such as vibrations or resonance frequencies which depend on lubrication or the load of the machine. Conclusively persisting wear effects i.e. their signals or data have to be found and analyzed. The present paper shows a way to fulfill the CM’s and CBM’s requirements by modeling the feed drive as a purely kinematic system described by physically static and persistent characteristics only. An adequate approach to record the controller’s internal signals that refer the dependent figures to the feed drives position provide the according data. The monitored system model wise and measurement wise is treated as a closed causal system where the leading figure corresponds to the system’s input, the measured figures to the system’s answer. This is achieved by coupling the measured figures to the leading figure by measurement technique means in a way that an output occurs only when an alteration of the input occurs. Thus the data is completely independent of time. Here kinematic coupling always is linear wile dynamic behavior i.e. data recorded in reference to time only is linear when the machine operates in a steady state. Since transient states are much more susceptible to a system’s state variables i.e. wear it is important to explore them. Here, the paper’s approach always provides linear data while time based data recording provides warped data. The present paper provides a method that, owing to its independence from time, can specifically examine unsteady and transient states that occur during the machine’s normal operating conditions. Since the causality, the persistency and that the feed drive can be modeled purely linear, the presented solutions provides data that always is linear even if the feed drive moves unevenly. Warp or jitter does not happen. The developed method provides a wear depicting figure that bears a physically meaningful dimension. For mode separation this approach allows using linear transformations only such as Fourier transformation. Fourier transformation of data recorded in reference to the feed drive’s position results in data that is referred to resolution or distance. Reflecting long established characteristic bearing frequencies, here the term “Characteristic Distance and Resolution” is coined. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|