Daniel A. McAdams The research objective of this award is to explore and develop tools that will expand the use of biological analogies in engineering design. The results of this research will enable design engineers to access the wealth of elegant and ingenious approaches to problem solving that are found naturally present in biological systems. There are several challenges that prevent designers from taking full advantage of the biological knowledge domain. The first and most significant obstacle is that most engineers do not have sufficient knowledge of biology to identify an analogy from which to develop an engineered solution. To address this challenge, this research has two main thrusts. The first is the creation of a thesaurus that translates the terminology of function as used in engineering to the terminology of function as used in biology. The second is the use of novel, new computational search tools to access repositories of biological knowledge in an effort to data mine biological systems for functional analogies Thus, as engineers move a design from function to form, they can search nature for form solutions.
If successful, this proposed research makes fundamental contributions to engineering design through creation of a biomimetic design repository and associated concept generator. Key contributions include better cross-domain search strategies on natural-language knowledge to identify novel biological analogies, and better mechanisms to enable understanding of biological phenomena so they can be more successfully applied to solve engineering problems. The biomimetic solutions will be immediately available to both academia and practicing engineering designers as the research is performed through integration in an NSF supported Engineering Virtual Organization. Curriculum development will provide a testbed for using this biomimetic repository for concept generation.
From simple cases such as hook and latch attachments to articulated-wing flying vehicles, the biological domain has been used to inspire many design ideas. While biological systems provide a wealth of elegant and ingenious approaches to problem solving, there are several challenges that prevent engineers from taking full advantage of the biological knowledge domain. The first and most significant obstacle is that most engineers do not have sufficient knowledge of biology to know of the most relevant analogies for any given problem. In this work, we developed knowledge and methods to allow engineers to find biological solution principles that are applicable to an engineering problem of interest. In this work, we address these obstacles by attempting to answer two key questions. First, what search strategies can we use to truly access the full insights already known about the biological domain? Second, how do we present this information to designers such that relevant biological phenomena can be best exploited in solving engineering problems? Toward this first obstacle, we have developed a search methodology that searches for instances of keywords in the abundance of biological knowledge already in natural-language format, e.g., texts, and papers. This keyword search is fundamentally based on the notion of function. In other words, during the design process, the engineer is trying to identify a principle, and create a form, that fulfills some needed function. Thus, the engineer is seeking some system or entity in nature that performs the function of interest. In this research, we have established that both natural and engineered systems have many functions in common. However, there is little commonality in the language used by biologists and engineers to describe these functions. A key result of this research that enables text-based search of archived biological knowledge is the creation of a thesaurus that contains synonyms of engineering function and biological function. For example, an engineer may be seeking a solution to a problem with the function of "regulate fluid." In this case, the function is "regulate" and (one of) the biological keywords is "respire." Using the search tool, the engineer enters "regulate fluid" and the search engine returns text passages that contain the term "respire." The engineer then reads through the passages to find the agent that performs the function "respire" and then adapt the specific mechanism that the agent employs to "respire" for the engineered system. As with many text based search methods, a large number of results are returned with only a limited number providing useful information to the searcher. In our case here, useful information is some principle or idea that helps the engineer solve the problem. Toward that goal, we have created computational text classifier schemes that identify what passages are most likely to provide inspiration for the engineering. After developing and comparing several text classification algorithms, we found that a simple Naïve Bayes classifier can be trained on a set of known good passages and then reliably identify biology passages that are likely to give designers ideas. Moreover, accurate classifications can be made based on a set of as few as 60 word stems. The Naïve Bayes classifier has the potential to increase the number of helpful results, from 29% of returned passages to 87%. In this work, we have made basic contributions to engineering knowledge that enable bioinsired design to move from a random brainstorming type approach to a formal, intentional, and repeatable method.
