ENGINEERING INFERENCE ENGINE
WHAT DOES IT DO?
The Engineering Inference Engine allows you to build mathematical models that represent complex engineering products such as rockets, satellites or cars, and then solve these models to define the design and quantify the performance. The solution procedure is deduced by the inference engine itself, empowering a single designer or a small design team to wield more complex engineering solutions than ever before.
|The Engineering Inference Engine is forward chaining using local propagation for high speed. More||The number of variables is limited only by memory size. More||Models are object oriented. More|
|There are no traditional IF THEN rules, as such. More||New data types are built from basic ones. More||A standardized engineering/physics ontology. More|
|Computational complexity is minimized. More||Any problem that can be represented by known functions can be solved. More||The representation of knowledge within the Engineering Inference Engine is inherently more compact and reusable than conventional code. More|
APPLICATION EXAMPLE: TRIANGLE CALCULATOR
Some types of triangles are easier to solve than others. The video below shows a model of a triangle within EIE being manipulated via its Excel interface. This example illustrates that the triangle solves given many different ways of specifying the problem. In contrast, regular procedural programs would need to be rewritten for each case.
Similarly, more complex models representing spacecraft, for example, require a different solution procedure each time we decide that a different aspect of the design is important and should drive the solution. This aspect might be the maximum temperature of a component or the maximum weight a given rocket can launch, or the limited size of a bank account! The message here is that the solution procedure changes as you play with and begin to understand a model as it applies to a given situation. A procedurally-defined model must be continually rewritten, whereas a declaratively-defined model stays the same.
APPLICATION EXAMPLE: MICROWAVE THERMAL ROCKETS
The Engineering Inference Engine was originally developed to enable the Microwave Thermal Rocket to be designed in greater detail than possible using conventional methods. In 2006, point designs for the rocket were generated using a model comprised of roughly 500 variables. The Engineering Inference Engine was conceived in the fall of 2007, and in 2015 the rocket is described by a model comprised of 22,000 variables. These variables represent the rocket’s components, geometry, materials, propellant and flowpath, ascent trajectory and beam director, among other aspects. The Engineering Inference Engine solves this model in about 10 minutes on a desktop computer to produce a point design.
Euler diagram combined with directed acyclic graph showing the top-level propagation of information through the system model of the microwave thermal rocket. User-defined variables are shown with black backgrounds. Only selected variables and relations are shown. There are 22,000 variables in the system model as of November 2015.
A representation of the system model and top-level solution procedure is shown above. The particular combination of input variables implies the direction of the arrows in the diagram. As the system model has evolved over the years, the choice of independent variables has changed. Sometimes an independent variable is changed to eliminate an iteration, thereby reducing computational complexity. At other times, an independent variable is changed because the understanding of which variables are the true constraints that drive the design has changed.
APPLICATION EXAMPLE: FLUID FLOW
The full-screen video below shows the command-line version of EIE in use. Within EIE, an object has been created that represents a fluid flowing through a duct in quasi-1D. In this case, the fluid is hydrogen and it is flowing at cryogenic temperatures. EIE uses a Helmholtz equation of state to infer the phase of the fluid (liquid/gas/both) and its bulk thermodynamic and transport properties. Also assuming a hydrodynamically-developed flow within a smooth duct of circular cross-section, EIE deduces wall properties associated with heat exchange. The object represents just a single point within a duct and many such objects are connected together to represent the propulsive flowpath of the microwave thermal rocket, for example.
Note: To see the video more clearly, move your mouse to the bottom right hand side of the video and click full-screen on the slider that appears.