Predictive Maintenance and RCA
_quality(80)_edited.jpg)
3' read
Problem
There is an excessive number of data points and alerts which need continuous monitoring for facility water pumps and complex machine equipment. Previous analytical alert system was 10-15 years outdated and failed capture alerts and on-time, leading to down-time and supply-interruption. Shortage of maintenance personnel requires precise optimization of resources.
Solution
-
Identified data quality issues in pre-processing and performed cross-environment data quality cleaning for optimal results.
-
To account for large amount of equipment types and data to be considered, we trained a GenAI model to recognise patterns in data leading to equipment issues.
-
Model classified and dynamically ranked problematic equipment due to historic issue log, time of year, etc.
-
Built LLM-based / AI Chatbot interface which allowed technicians to prompt the model on root-cause of issue and recommended solution.
-
Deployed on Palantir & MongoDB seamlessly integrating into existing infrastructure, centralizing management / monitoring and connecting with existing processes.
Impact
-
Predictive alerts on centralized system
-
Reduced technician cost from improved root-cause analysis​
-
Improved resource optimisation
-
GenAI dynamic model selection, resulting in simplified workflows
