A Novel Efficient Big Data Processing Scheme for Feature Extraction in Electrical Discharge Machining

Electrical discharge machining (EDM) can machine hard conductive workpieces that are difficult to machine using traditional machining techniques. For monitoring the EDM process using virtual metrology (VM), probes with a very high sampling rate are needed to acquire the voltage and current signals of electrodes, thereby generating a huge amount of sensor data and raising a big data processing issue in extracting features from raw sensor data. This paper proposes a novel efficient big data processing scheme for feature extraction in EDM, called BEDPS, based on Spark and HDFS. A parallel gap-based wave detection mechanism using Spark is designed to efficiently detect effective machining waves from big machining raw data of EDM without the slow internode data communications in HDFS. A pre-loaded memory-based feature calculation mechanism is also developed to calculate the key features from machining-wave data in Spark in a parallel-processing manner. Testing results show that the proposed BEDPS is much more efficient in terms of total execution time in machining wave detection and feature calculation and more scalable in the data size that can be processed, compared to the existing system.