BACKGROUND: The Boeing Company is the world's largest aerospace company. Boeing manufactures commercial jetliners, military aircraft and is the nation's largest NASA contractor. Company revenues for 1996 were $22.7 billion, for 1997 they were $45.8 billion and for 1998 they reached $56.2 billion. The ability to integrate multiple complex systems is one of the foundations on which the success of the company depends.

PLATFORMS: PolyAnalyst PRO with the find laws algorithm requires a Windows NT environment. A 400 MHz Pentium machine with 256MB RAM was used as the baseline desktop configuration.

PROBLEM SOLVED: Integration of large, complex systems often involves embedding models of system components in various software algorithms. The engineering staff was searching for a tool to quickly develop low order models of nonlinear processes which could be implemented as part of an embedded real-time system. Traditional curve fitting and neural network techniques proved to be too costly in size and throughput as well as time-consuming to design. The symbolic knowledge acquisition technology in the PolyAnalyst find laws algorithm provided a unique capability to balance performance with complexity for models derived through machine learning. In evaluating complex system behavior during testing, the easy access to traditional data mining tools also enabled the analysis of data across the entire test program, rather than on a case-by-case basis.

PRODUCT FUNCTIONALITY: The Poly-Analyst suite offers a range of capabilities for data access, data set manipulation, machine learning, visualization and reporting.

STRENGTHS: PolyAnalyst offers some very powerful modeling tools that can be utilized quickly by relatively inexperienced users. This ease of use along with the ability to form minimal nonlinear models, which can include logic, is a potent combination. Both continuous and state machine processes can be handled. The overall integration of machine learning with data manipulation and other exploration tools provides an environment to truly interact with the data and perform "what-if" tests.

WEAKNESSES: Our tasks required that many models be generated to implement each design cycle. The find laws task is resource-intensive and does not lend itself well to concurrent learning. A queue system needs to be implemented so multiple jobs can be run sequentially without constant attention. The find laws algorithm is evolutionary and, much like neural networks, can be time-intensive to train. A more convenient interface to other analysis packages such as Matlab would be useful for us.

SELECTION CRITERIA: The unique find laws algorithm along with an easy-to-use interface made PolyAnalyst the only choice for our environment.

DELIVERABLES: The PolyAnalyst find laws exploration engine has produced equations that are used directly in real-time algorithms. This tool also provides the engineering staff with new capabilities in analyzing data across the lifetime of a test program.

VENDOR SUPPORT: Megaputer Intelligence support has been enthusiastic and timely. The staff has been generous in sharing their product and domain knowledge.

DOCUMENTATION: The on-line documentation and tutorial are good and intuitive. Supplementary documentation from Megaputer Intelligence is very useful and easily available via their Web site.

Register or login for access to this item and much more

All Information Management content is archived after seven days.

Community members receive:
  • All recent and archived articles
  • Conference offers and updates
  • A full menu of enewsletter options
  • Web seminars, white papers, ebooks

Don't have an account? Register for Free Unlimited Access