Using artificial intelligence to improve the monitoring and evaluation of safety-critical communications
Rail companies must regularly monitor frontline staff’s adherence to safety-critical communication (SCC) criteria. This ensures the safe running of the railway.
However, monitoring SCCs manually takes a lot of time and labour. This means that only a small amountof SCCs are assessed.
So, industry currently has an incomplete picture of issues and trends to take action.
This research confirmed that off-the-shelf AI technology can be customised to monitor and evaluate SCCs.
Using a combination of automatic speech recognition and natural language processing models, the proof of concept (PoC) successfully picked up adherence to selected SCC criteria.
The research also created a high-level economic assessment. This found that the extent of further costs and benefits can vary greatly. This is due to the possible range of AI development and implementation costs.
An AI system could greatly increase the coverage of SCC monitoring.
The PoC system currently processes around 30 forms in 2 hours. This could feasibly be increased to around 30 forms in 30 minutes. This substantially increases the total SCCs that could be monitored.
Insights gathered from AI SCC monitoring can collect feedback and broad level trends much more effectively. Improved monitoring and training is likely to improve adherence to SCC criteria. This would contribute to reducing safety incidents where poor SCCs are a factor.