This grant provides funding for the development and application of signal-processing and neural network methods that address monitoring and diagnostic issues commonly encountered in a variety of industries. In the first phase of the research, generic algorithms and methods for addressing several high-impact problems will be developed. A representative set of problems includes: (1) the recognition of recurring patterns of arbitrary shape in time-domain sensor data, (2) the extraction of robust features from the wavelet representation of sensor data, (3) the identification of blurred or smeared surface defects in harsh environments, (4) the acceleration of the training process of a multilayer neural network without adversely affecting its natural course of learning, and (5) stopping criteria for neural network training to strike a balance between computational efforts and accuracy. An important objective of the present research is to develop methods that are amenable to broad applicability, so that several crosscutting issues in a wide spectrum of industries can be addressed. In the second phase of the research, the methods developed in the first phase will be applied to problems in industries collaborating with the project team. The information and data collected from the industrial applications will be used for validating the knowledge generated through research and also for building living cases for classroom instruction.
The outcome of this research will significantly contribute to building self-guiding, self-correcting, and self-protecting products, processes, and electro-mechanical systems ubiquitous in a broad spectrum of industries including manufacturing, aerospace, automotive, defense, and health care. The monitoring and diagnostic methods developed during this project have a good potential to increase production yields by minimizing defective workpieces, preventing damage to machines, and avoiding idle times on machines. These methods will also substantially improve work safety and provide better mechanisms for human-machine interfacing.
