Data Science for Smart Electrode System Development

27 May 2020

Master of Applied Data Science student, Jiaqi Cheng, has developed a computational tool that supports the development of Smart Electrodes.

  • Jiaqi news - joule log heating knots modelling

In 2016, during the Joule log heating experiments, our research team, led by Dr Bill Heffernan, found that the segmented electrodes designed to measure the distribution of electric current flow in log faces allowed determination of various conducting and non-conducting parts at the logs’ ends, such as log contours, heartwood, and sapwood. Knowing the size of these parts is of great value for log trading, wood processing, and heat treatment.

However, the resolution of our electrodes was limited to 30 segments, which prevented accurate estimation of the log parameters key to certain applications. The analogy could be made with low-quality images from old digital cameras due to low pixelation: our electrodes equipped with 30 segments could not provide a high-resolution image of current distribution. Hence, in July 2019, we started the development of a 1,023 segment electrode, the Smart Electrode System (SES), which will allow us to identify small log features such as knots (Figure 1). The number of segments was optimised by sensitivity analysis for the size of SES to ensure the most accurate output.

According to our earlier studies, the segments in contact with knots would drive lower electrical current than those in contact with surrounding sapwood. However, without a working prototype of SES, an experimental visualisation and prediction of the current distribution, as affected by knots, was impractical. Therefore, in December 2019, we invited a Master’s student in Applied Data Science from the University of Canterbury, Jiaqi Cheng, to do his degree’s project at the EPECentre.  

Jiaqi has developed a computational tool, which enabled us to model electrode current distribution on 100 pseudo logs (Figure 2). His main challenge was to ensure that the artificial dataset of the various log parameters was unbiased and behaved similarly to the original experimental dataset. For this, Jiaqi employed multiple data science methods and tools, which allowed him to mimic the correlation between wood parameters, such as basic density (BD), moisture content (MC), and log weight.

The EPECentre team was delighted to work with Jiaqi and wishes him the very best in graduating and the future.