Towards Data Based Corrosion Analysis of Concrete with Supervised Machine Learning
2018
Half-Cell-Potential Mapping (HP) is the most popular non-destructive testing (NDT) method for the detection of active corrosion in reinforced concrete. HP is influenced by parameters such as moisture and chloride gradients in the component. The sensitivity to the spatially small, but dangerous pitting is low. In this study we show how additional measurement information can be used with multi-sensor data fusion to improve the detection performance and to automate data evaluation. The fusion is based on supervised machine learning (SML). SML are methods that recognize relationships in (sensor) data based on given labels. We use SML to distinguish "defective" and "intact" labeled areas in our dataset. It consists of 18 measurement - each contains HP, ground radar, microwave moisture and Wenner resistance data. Exact labels for changing environmental conditions were available in a laboratory study on a reinforced concrete slab, which deteriorated controlled and accelerated. The deterioration progress was monitored continuously and corrosion was generated targeted at a predefined location. The detection results are quantified and statistically evaluated. The SML results shows a significant improvement over the best single method (HP).
Keywords:
- Correction
- Cite
- Save
- Machine Reading By IdeaReader
0
References
1
Citations
NaN
KQI