Bond graphs and their application to mechatronics

Whether creating a new product or improving an existing one, the engineer in charge of creating a piloted system generally follows a two-phase approach:

• A first step is to study the passive (uncontrolled) behavior of the system;

• the second stage essentially concerns the definition of specifications for the determination and implementation of control laws to drive the active system.

The tasks to be carried out depend on the objectives to be achieved, whether it's building a suitable, validated model, designing an actuation and measurement system that meets performance and reliability constraints, choosing robust controls that are insensitive to inaccuracies or parameter drifts, defining a fast, high-performance control/supervision system, or implementing a prototype with its technological and cost constraints.

The first essential phase in this process is modeling.

The modeling of a system cannot be considered without first analyzing the level of complexity required for the study in progress, and the use that will be made of it. The engineer may therefore have to choose between different types of model during the course of his study.

For example, a finite element model may be suitable for the static dimensioning of a mechanical system, while an identified black box model may suffice for the calculation of a simple control law to drive it.

These different models are complementary, specific to a particular type of study, but each requires a new effort to build. It is not possible to envisage their use throughout the entire design process.

What's more, each field of physics has its own traditions and habits in terms of notations, methodologies and conventions, making it difficult for specialists from different fields to communicate with each other. These specialists have to collaborate more and more because of the multi-disciplinary nature of new products, combining mechanics, hydraulics, electronics...

The bond graph tool can provide a solution to some of the difficulties encountered, thanks to its specific features:

• it is a language for representing power transfers within a system, which assumes an energetic approach to problems, and is therefore universal;

• it is graphic and unified for all areas of physics, since it is based on the notion of analogy.

The model obtained can be a good knowledge model, because the very principle of its construction makes it a "grey box" as opposed to "black box" models obtained by identification: the structure...