Now showing 1 - 3 of 3
  • Publication
    A Case-Study on the Impact of Dynamic Time Warping in Time Series Regression
    It is well understood that Dynamic Time Warping (DTW) is effective in revealing similarities between time series that do not align perfectly. In this paper, we illustrate this on spectroscopy time-series data. We show that DTW is effective in improving accuracy on a regression task when only a single wavelength is considered. When combined with k-Nearest Neighbour, DTW has the added advantage that it can reveal similarities and differences between samples at the level of the time-series. However, in the problem, we consider here data is available across a spectrum of wavelengths. If aggregate statistics (means, variances) are used across many wavelengths the benefits of DTW are no longer apparent. We present this as another example of a situation where big data trumps sophisticated models in Machine Learning.
  • Publication
    Detecting Voids in 3D Printing Using Melt Pool Time Series Data
    Powder Bed Fusion (PBF) has emerged as an important process in the additive manufacture of metals. However, PBF is sensitive to process parameters and careful management is required to ensure the high quality of parts produced. In PBF, a laser or electron beam is used to fuse powder to the part. It is recognised that the temperature of the melt pool is an important signal representing the health of the process. In this paper, Machine Learning (ML) methods on time-series data are used to monitor melt pool temperature to detect anomalies. In line with other ML research on time-series classification, Dynamic Time Warping and k-Nearest Neighbour classifiers are used. The presented process is effective in detecting voids in PBF. A strategy is then proposed to speed up classification time, an important consideration given the volume of data involved.
      389Scopus© Citations 16
  • Publication
    Scoring Performance on the Y-Balance Test
    The Y-Balance Test (YBT) is a dynamic balance assessment commonly used in sports medicine. In this research we explore how data from a wearable sensor can provide further insights from YBT performance. We do this in a Case-Based Reasoning (CBR) framework where the assessment of similarity on the wearable sensor data is the key challenge. The assessment of similarity on time-series data is not a new topic in CBR research; however the focus here is on working as close to the raw time-series as possible so that no information is lost. We report results on two aspects, the assessment of YBT performance and the insights that can be drawn from comparisons between pre- and post- injury performance.
      156Scopus© Citations 4