Sensors and Sensor Networks: Development and Testing of Practical Algorithms for Online Interpretation of Sensor Data through Wavelet Decomposition

项目摘要：通过小波分解在线解释传感器数据的实用算法的开发和测试提案# 0330145传感器技术及其应用的最近发展使得收集丰富的数据变得容易。不幸的是，通过传感器收集的数据通常远远超过我们目前处理它们的能力。解决这一瓶颈问题的一个办法是开发有效的、可在线实施的数据解释统计算法。此外，传感器的持续小型化显著地增加了传感器制造技术的复杂性，传感器制造技术本身严重依赖于传感器。因此，传感器技术的两端（制造和数据解释）都将从这项研究中获益。 获得的资金将使我们能够1）开发理论分析，以评估在线多分辨率算法在解释传感器数据方面的有效性，并构建有效的单变量和多变量在线算法，2）在从纳米级制造过程中收集的传感器数据上测试在线算法，以及3）使用在线传感器数据分析算法开发新的Wavelet-EWMA反馈控制策略。我们的目标#1的策略是首先建立一个数学模型的多分辨率数据分析方案，以评估性能指标，如误报率，平均运行长度。 我们的测试策略是精心设计的。我们将从测试台收集各种传感器数据，这些数据与一些突出的操作问题有关。 对于每一个数据集，我们不仅会测试算法的性能，还会使用测试结果来回答上述问题。如果成功，本研究开发的数学分析将为基于小波的在线多尺度算法的性能评估提供坚实的基础。这一科学基础将导致开发在线和实际可行的传感器数据分析算法。通过与半导体行业的合作，将通过拟议研究发现的知识转移到技术创新中。

Project Summary: Development and Testing of Practical Algorithms for Online Interpretation of Sensor Data through Wavelet Decomposition Proposal # 0330145 Recent growth in sensor technology and its applications have made it easy to gather abundance of data. Unfortunately, the data collection rate via sensors usually far exceeds our current ability to process them. A remedy to this bottleneck situation is the development of effective and online implementable statistical algorithms for data interpretation. Furthermore, continued miniaturization of the sensors have significantly increased the sophistication of the sensor fabrication technology which itself is heavily dependent on sensors. Thus, both ends of the sensor technology (fabrication and data interpretation) stand to gain from this research. The funding granted would allow us to 1) develop theoretical analysis needed to assess the effectiveness of online multiresolution algorithms in interpreting sensor data, and construct efficient univariate and multivariate online algorithms, 2) test online algorithms on sensor data collected from a nanoscale fabrication process, and 3) develop a new Wavelet-EWMA feedback control strategy using online sensor data analysis algorithms. Our strategy for objective #1 is to first build a mathematical model for the multiresolution data analysis scheme in order to assess performance measures such as false alarm rate, and average run length. Our testing strategy is an elaborate one. We will collect a variety of sensor data from the test bed that relate to a number of outstanding operational issues. With each of these data sets, we will not only test the performances of the algorithms, but also use the test results to provide answers to the above issues. If successful, the mathematical analysis developed by this research would provide a strong foundation for performance assessment of wavelet based online multiscale algorithms. This science base would lead to the development of online and practically viable algorithms for sensor data analysis. Transfer of knowledge discovered through the proposed research into technological innovations will be enhanced through our collaborations with the semiconductor industry.