AI for sensor health and industrial process optimisation. Making industrial plants greener, safer and more efficient
Dr Daniel EverittHead of Technology - Ada Mode
Time & Location
Thursday: 12.40 to 13.00, Stage 1
About this presentation
A wide range of processes depend on accurate determination of key parameters, from temperature, to pH, to chemical concentrations and more. Industrial plants are increasingly brimming with a wide range of sensors and instrumentation, and the data they generate are increasingly relied upon for plant/process control, optimisation, and maintenance planning. It is therefore essential to maintain confidence in the integrity of these data. However, over time many sensors suffer from calibration drift, gradually decreasing the accuracy of their readings and providing a distorted view to operators. To combat this, scheduled calibration checks of sensors must be carried out, representing a significant maintenance burden.
Where checks are at insufficient frequency, poor quality data from uncalibrated sensors can lead to downstream process compliance issues, with a wide range of potential impacts. The ability to continuously quantify the extent of sensor drift would be of high value to industry, enabling operators to shift to condition-based sensor recalibrations and boosting confidence in data integrity.
Machine learning techniques have significant promise in this area, allowing for the implementation of so called ‘soft sensors’, robust predictions of a target parameter’s behaviour against which actual sensor data can be compared. As hardware sensor values drift from the soft sensor, this can be quantified as calibration drift, unlocking significant potential benefits.
Daniel completed his PhD in Chemistry in 2016 and then began his career in industry as an engineering consultant in UK nuclear generation.
This work had a particular focus on both long-term coolant chemistry optimisation and introducing more condition-based maintenance to plant outages.
Through this work, Daniel increasingly applied data visualisation, statistical modelling and machine learning techniques alongside traditional domain expertise, allowing for deeper insight into industrial problems.
Alongside colleagues, Daniel founded Ada Mode in early 2020 with a mission to improve the performance of complex industrial processes through artificial intelligence, engineering expertise and design thinking.
As Head of Technology, he works to bring together Ada Mode’s domain experts and data science team to tackle a range of challenges in civil nuclear, renewables, utilities and manufacturing.