Early Career Researchers from UPGro Gro for GooD reflect on their time with the programme with Nancy Gladstone (University of Oxford)
What was your research focus in the Gro for Good project?
I joined as a member of the engineering team and my work focused on the use of artificial intelligence to remotely monitoring the condition of the handpumps in Kwale County. By retrofitting a small sensor inside the handle of the handpump, we are able to measure the smallest vibrations while the handpump is being operated and then use those to train machine learning algorithms that can predict the deterioration in the condition of the handpump.
What fieldwork did you do in Kwale County and what data did you collect?
I visited the handpumps of Kwale County to catalogue the condition and collect vibration data from more than 50 different handpumps over the last four years. In addition, we identified 15 locations for the long-term, continuously monitored sites in three clusters – grouped according to the depth of the borehole – in Shimba hills (deep), Ukunda scheme (medium), and near the coast (shallow). At these sites, I installed “loggers” (accelerometer sensors) inside the handpump handles and a diver (water level) at the bottom of the rising main. This resulted in approximately 1-2GB of time-series vibration data collected at each handpump every month. Our team from Engineering worked closely with Rural Focus Ltd and FundiFix to locate and maintain the sites over the course of this project.
How have you shared your work in Kenya?
This work was presented to the Water Authorities and stakeholders at the annual project workshops in Kwale. During our four years of data collection, we trained three members of our research team in Kwale on time-series data collection and modeling. In addition, our findings were shared with local entrepreneurs from FundiFix.
What will you do next?
I am continuing my research on the use of artificial intelligence and machine learning for monitoring systems in limited-resource settings as a post-doc at Oxford University. In addition to further developing the algorithms for the handpumps, I have started to transfer learnings from this project to the monitoring of other systems in low-resource settings, such as using wearables to monitor deterioration in ICU patients in Vietnam.
- Greeff, H., Manandhar, A., Thomson, P., Hope, R., and Clifton, D.A., Rural Infrastructure Health Monitoring System: Using AI to Increase Rural Water Supply Reliability. 2018 Conference on Neural Information Processing Systems, AI for Social Good Workshop, 2018, 1-4.
- Greeff, H., Manandhar, A., Thomson, P., Hope, R., and Clifton, D.A., Distributed Inference Condition Monitoring System for Rural Infrastructure in the Developing World, IEEE Sensors Journal 19(5): 1820-1828, 2018.
- Greeff, H., Manandhar, A., Thomson, P., Hope, R., and Clifton, D.A.: Method and System for Monitoring a Remote System, GB 1819717.8, 2018, (Distributed Inference).