Automatic recognition of primitive changes in manufacturing process signals
Pattern Recognition
Detection of abrupt changes: theory and application
Detection of abrupt changes: theory and application
Fast subsequence matching in time-series databases
SIGMOD '94 Proceedings of the 1994 ACM SIGMOD international conference on Management of data
Neural networks for pattern recognition
Neural networks for pattern recognition
Fast time-series searching with scaling and shifting
PODS '99 Proceedings of the eighteenth ACM SIGMOD-SIGACT-SIGART symposium on Principles of database systems
Event detection from time series data
KDD '99 Proceedings of the fifth ACM SIGKDD international conference on Knowledge discovery and data mining
On the approximation of curves by line segments using dynamic programming
Communications of the ACM
Data Mining and Knowledge Discovery for Process Monitoring and Control
Data Mining and Knowledge Discovery for Process Monitoring and Control
Minimum Message Length Segmentation
PAKDD '98 Proceedings of the Second Pacific-Asia Conference on Research and Development in Knowledge Discovery and Data Mining
Time Series Segmentation for Context Recognition in Mobile Devices
ICDM '01 Proceedings of the 2001 IEEE International Conference on Data Mining
Optimal Time Segmentation for Signal Modeling and Compression
ICASSP '97 Proceedings of the 1997 IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP '97)-Volume 3 - Volume 3
A MAP solution to off-line segmentation of signals
ICASSP '94 Proceedings of the Acoustics, Speech, and Signal Processing,1994. on IEEE International Conference - Volume 04
Hi-index | 0.00 |
The data storage capacities of modern process automation systems have grown rapidly. Nowadays, the systems are able to frequently carry out even hundreds of measurements in parallel and store them in databases. However, these data are still rarely used in the analysis of processes. In this article, preparation of the raw data for further analysis is considered using feature extraction from signals by piecewise linear modeling. Prior to modeling, a preprocessing phase that removes some artifacts from the data is suggested. Because optimal models are computationally infeasible, fast heuristic algorithms must be utilized. Outlines for the optimal and some fast heuristic algorithms with modifications required by the preprocessing are given. In order to illustrate utilization of the features, a process diagnostics framework is presented. Among a large number of signals, the procedure finds the ones that best explain the observed short-term fluctuations in one signal. In the experiments, the piecewise linear modeling algorithms are compared using a massive data set from an operational paper machine. The use of piecewise linear representations in the analysis of changes in one real process measurement signal is demonstrated.