Component Integration for Compact Fluid Power Systems

Project / Leader

2E - Prof. Chris Paredis (Georgia Tech)

Title

Component Integration for Compact Fluid Power Systems.

Statement of Project Goals

The goal of the project is to reduce significantly the time and effort required to formulate and solve systems engineering problems for compact and efficient fluid-power systems.  To achieve this, analysis knowledge about fluid power components from multiple disciplinary perspectives and multiple levels of abstraction will be captured and organized in a modular, object-oriented knowledge repository using a standardized language (SysML), and synthesis knowledge about fluid power systems will be captured in the form of model transformations.  A systems engineering method and software framework will be developed in which the synthesis and analysis knowledge from the repository is used to explore efficiently and comprehensively large spaces of system architectures, with the goal to improve the compactness and efficiency of fluid-power systems while balancing other system objectives such as effectiveness, cost, and reliability.

Project's Role in Support of the Strategic Plan

The project provides a method and software framework to support the comprehensive and efficient exploration of integrated system architectures.  This will enable the integration of the fluid-power subsystem with structural subsystems (compact integration and distribution barrier) and enable the comparison between different system architectures for achieving desired system-level tradeoffs (system integration inefficiency barrier).  The framework could also enable the evaluation of the impact of introducing new component technologies (component efficiency barrier) or higher pressures (high pressure operation barrier) on system-level performance.

Descripiton and Explanation of Research Approach

With the advent of electronic control, fluid power systems have become increasingly integrated and multi-disciplinary in nature, and the number of potential system architectures has exploded.  With new demands on compactness, efficiency, and effectiveness, system engineers need to explore new system architectures that provide adequate tradeoffs across these conflicting objectives.  The main barrier that needs to be overcome is one of complexity: a very large amount and variety of knowledge is necessary to synthesize and analyze promising system architectures.  Unless this knowledge is managed well, the cost of acquiring, validating and applying this knowledge will limit significantly our ability to increase the functionality and performance of future fluid power systems.  To overcome this barrier, a systems engineering framework is required consisting of model repositories, algorithms for instantiating and linking these models, and algorithms for selecting appropriate models at each step of the design process.

The corresponding research question is:

How should one represent, store, retrieve and use knowledge efficiently and effectively in support of the design of fluid power systems?

The need for a systems engineering framework for fluid-power systems has been recognized before with initial work by Krus et al. at Linköping University, Tilley et al. at the University of Bath, and da Silva et al.  at the Federal University of Santa Catarina (Brazil), with more recent work by Pedersen at Aalborg University and Schlemmer et al. at the Technical University of Aachen.  In this related work, the focus has been on traditional optimization approaches with a model of the objective at a single level of abstraction, sampled by the optimizer as a black box model, implemented in an imperative (rather than declarative) programming language.  In addition, the work has focused almost exclusively on the modeling of the fluid power aspects of the system, with only a few efforts allowing for seamless integration with other disciplines (e.g., structural mechanical, thermal, electrical, controls). Finally, the past work either focused on optimization of the sizing parameters of a specific architecture, or used expert systems to guide the selection of a feasible architecture. The efficient exploration and optimization of system architectures has not been addressed.

In this project, the approach for realizing a systems engineering framework for fluid power systems is based on the formal, declarative representation of knowledge.  By capturing the knowledge formally, it can be more easily reused, allowing the cost of capturing and modeling the knowledge to be amortized over many re-uses.  In addition, by representing the models in a declarative form (i.e., an implementation-independent formalization of the mathematical relationships), the models can be transformed, combined, and symbolically manipulated to create and solve system-level models that are larger and more comprehensive than could be practically achieved otherwise.  Advanced solvers, such as Mixed-Integer Non-Linear Programming (MINLP) solvers or Equation-based Object-Oriented solvers for Differential Algebraic Equations can then symbolically manipulate the declarative equations to solve large system-level models much more efficiently than can be achieved with the current state of the art, i.e., imperative Matlab models with iterative optimizers.  The combination of well-structured and formal modeling languages with model repositories and advanced solvers enables the design of fluid power systems with a level of thoroughness and efficiency that was previously unachievable in both broadness of exploration and depth of analysis.

We are developing the systems engineering framework illustrated in Figure 1.  It consists of three layers that can be considered separately or in an integrated fashion:  The top layer addresses the generation and topological analysis of different fluid-power circuit configurations, the second layer sizes the components within a given circuit configuration based on algebraic models, and the third layer optimizes the components (under uncertainty) based on dynamic simulations.  Such a layered approach allows one to use resources efficiently by only performing detailed analyses if the performance predictions obtained in a previous layer are sufficiently promising.  The framework relies on formal representations in the Systems Modeling Language (OMG SysML^TM) to represent the problem definition, the libraries of fluid power components, and the analysis models that characterize these components from different perspectives and at different levels of abstraction (both as algebraic and as differential-algebraic models).  By capturing this information and knowledge formally, it can be transformed in an automated fashion using model transformations.

Figure 1: An overview of the proposed three-layer systems engineering framework.

In the previous project years, most of the blocks illustrated in Figure 1 have been addressed and illustrated individually.  In the coming years, these individual capabilities will be integrated with each other into the complete framework illustrated above.