St. Louis, Missouri
June 18, 2000
June 18, 2000
June 21, 2000
5.180.1 - 5.180.11
Creating Artificial Neural Network Modules For Use In Rapid Application Development Garrett S. Harris a, Bruce E. Segee a, Vincent M. Allen b a University of Maine at Orono / b Modicon Corporation
N eural networks and fuzzy logic have emerged as useful tools for the calibration of arrays of thin film gas sensors. Properly choosing network parameters is essential to achieve acceptable network performance. Often, choosing said parameters involves a time consuming search of many possible candidate networks. When the neural network code is incorporated with in an application with other code, such as for data capture, presentation, and hardware control, interdependencies often form between code segments. Enhancements, modifications, and fixes to the code lead to an extensive and time-consuming rewrite of many parts of the software. Thus, the need arises for neural network software modules that can be easily incorporated in application software but whose interface is well defined and whose implementation is entirely separate from the functionality it provides. By providing debugged and proven software modules encapsulating neural network functionality that can be simply inserted into any application, the entire software system can be modularized. These modules can be reused easily, and changing the neural network operating parameters no longer involves a complete software rewrite or even a recompile. By following the guidelines of rapid application development techniques, and using emerging technologies such as ActiveX and OLE, these modules can be easily developed.
Artificial neural networks (ANNs) have emerged as useful tools in a number of areas including but not limited to gas sensor array calibration [Bajaria, 1996]. Artificial neural networks are so named because they employ a large number of simple processing elements and are able to "learn" appropriate behavior based on training data. The training data usually consists of data gathered under known conditions. The training process consists of adjusting network parameters to reduce the difference between the actual network output and the desired output for that data. In a gas sensing application, arrays of thin film gas sensing platforms are normally used. These devices usually have responses to a wide variety of gases, target as well as interferents. While the sensors are responsive to the majority of the gases presented to them, individual sensors respond differently to different gases. By combining the sensors into an array configuration, combinations of sensors can reliably detect specific gases. Artificial neural networks can learn to approximate the actual gas concentrations based on training from data gathered under laboratory conditions for known concentrations of mixtures of gases.
Allen, V. M., & Harris, G., & Segee, B. (2000, June), Creating Artificial Neural Network Modules For Use In Rapid Application Development Paper presented at 2000 Annual Conference, St. Louis, Missouri. https://peer.asee.org/8250
ASEE holds the copyright on this document. It may be read by the public free of charge. Authors may archive their work on personal websites or in institutional repositories with the following citation: © 2000 American Society for Engineering Education. Other scholars may excerpt or quote from these materials with the same citation. When excerpting or quoting from Conference Proceedings, authors should, in addition to noting the ASEE copyright, list all the original authors and their institutions and name the host city of the conference. - Last updated April 1, 2015