For product design optimization processes, computer simulations form a critical part. This allows engineers to test a number of configurations and select the best among a slew of alternatives. Nonetheless, simulations are very expensive and are time-intensive, be with state-of-the-art resources at Argonne National Laboratory of the U.S. Department of Energy.
In a bid to accelerate the design process of products, a research team at Energy Systems Division, Argonne National Laboratory created a new design optimization application called ActivO. The new tool can reduce the time needed to find the best design to a great extent.
The design process involves employing a novel machine technique. The techniques helps users to focus on how to target computational resources most efficiently.
“In fact, ActivO performs the simulation in a smart manner, and quickly identifies the components of the design space that needs to be focused on,” said of the researchers. A process that used to take two to three months to give the optimum design that can be completed within about a week.
Meanwhile, the ActivO approach successfully demonstrated for use in optimizing combustion engines, and presented in an article published by the American Society of Mechanical Engineers.
According to the lead author of the study, ActivO is a hybrid algorithm that uses the capacities of two types of machine learning surrogate models for superior performance.
In fact, the design of machine learning models is such so as to work in cooperation. Advantageously, one of the models allows to explore the design space adaptively, rather than running randomly sampled simulations. Essentially, this guides to the region most likely to contain the global optimum.