Research Information System
Schmidt K. V. (Executive)
TUBITAK Project, 2011 - 2014
Traditionally, manufacturing systems are realized as dedicated
manufacturing lines (DML) or flexible manufacturing systems (FMS) in order to either achieve a high
product quality at high volumes and low cost in a dedicated manufacturing plant
or to be able to produce a variety of different products on the same
manufacturing plant. However, in order to address the aggressive competition
and the rapid changes in the product development and the production technology
in contemporary manufacturing systems, reconfigurable manufacturing systems
(RMS) were introduced as a new paradigm in manufacturing. The aim of RMSs is to
support the quick launch of new product models, the rapid adjustment in the production
capacity, the fast integration of new functions and technologies and the easy
adaptation to variable product quantities for niche marketing. To this end,
RMSs are composed of specific reconfigurable machine tools (RMTs).
From the control perspective, RMSs require rapid and frequent changes of
local RMT controllers and of controllers that govern the coordination among the
RMTs of an RMS. In addition, the controller design for RMSs must allow a
modular realization, the easy redesign or integration of new controller
components in case of configuration changes, the efficient use of system
redundancies and the re-usability of controllers for standard components. Furthermore,
it is important to guarantee the diagnosis of system failures and the
autonomous controller reconfiguration for the purpose of failure recovery.
Currently, there are two main approaches towards the control and failure
diagnosis of RMSs. On the one hand, agent-based manufacturing considers RMSs as
a conglomerate of autonomous physical and logical agents that have to perform
interaction in order to achieve local and global goals. The advantage of
agent-based manufacturing is the possibility of a fully distributed realization
of the reconfiguration control. However,
this approach can lead to a high complexity of the required interactions, and
it is difficult to assess the overall system performance and correctness. On
the other hand, discrete event system (DES) models are used to represent the
elementary operations and interactions of RMSs. Such models are suitable for
the design of controllers and failure diagnosers for the sequential behavior
and coordination in RMSs and allow a global system view. However, all existing
approaches are restricted to certain scenarios such as the failure recovery or
the re-usability of controllers. Moreover, most approaches do not include the
coordination control, do not consider a modular controller realization and are
not scalable to large-scale RMSs. In summary, there is currently no approach
that integrates the reconfiguration control, the failure diagnosis and the
failure recovery for RMSs in a general framework that allows the coordination
controller design, a modular controller realization and that is scalable to
RMSs of realistic size.
The subject of this project is the development of a continuous workflow
that addresses all relevant issues of the reconfiguration control and failure
diagnosis for RMSs. The first objective is the development of a comprehensive
hierarchical DES modeling framework. This framework will enable the easy
representation of modular system components, the design of local and
coordinating reconfiguration controllers and the modeling of and recovery from
system failures. The second objective of this project is the development of analysis
tools for relevant properties of RMSs within the proposed modeling framework.
Based on these analysis tools, the third project objective is the development
of reconfiguration control and diagnosis design techniques that are scalable to
large-scale systems that are suitable for failure recovery, that require a
minimal amount of controller re-computation and that support a modular
realization. The final objective is the application of the modeling, analysis
and design methodology to a large-scale RMS that is acquired in the scope of
this project. In this context, it is intended to create a simulation platform
that allows the direct execution of the designed controllers and failure
diagnosers on a PC that is connected to the physical RMS in a hardware-in-the-loop
simulation. Then, it is possible to conduct a behavioral analysis of RMSs. All
objectives include the implementation of the designed algorithms in the form of
a software library.
Different from all of the previous work, the proposed
approach allows a comprehensive representation of RMSs and a rapid realization
of the designed controllers and failure diagnosis. All relevant design aspects
such as modularity, scalability, diagnosability and re-usability will be
achieved in the same formal framework. As an important feature, the proposed
framework will allow the correctness by design by employing formal DES design
techniques. In addition, the software support allows the convenient application
of the developed methodology. Furthermore, the developed simulation platform
makes it possible to directly connect the simulated controllers and failure
diagnosers to a physical RMS for the purpose of verification and behavioral
analysis.
As the modeling framework and the controller design methodology are novel,
the results of the research will be publishable in international conferences
and indexed journals. In addition, the outcome of the project has the potential
to attract the industrial community since it offers a generic modeling
technique and design workflow with algorithmic support and direct applicability
to physical RMSs. Hence, the project will both contribute to the academic
literature and the wealth and prosperity of the country.
The project is expected to be completed in three
years by the works of the primary investigator, one PhD student and two master
students. At the end of the project, these students will have experience in the
system analysis, controller design, failure diagnosis and implementation of
manufacturing systems with a focus on RMSs.
In the course of the project we will collaborate
with the Chair of Automatic Control at the University of Erlangen-Nuremberg,