PREDICTIVE MAINTENANCE
Key Figures from our use case deployment
Business Challenge
Poor maintenance practices can also reduce the overall lifespan of equipment, leading to higher replacement and capital expenditure costs. Furthermore, inadequate maintenance can result in inconsistent product quality and higher defect rates, negatively impacting customer satisfaction and brand reputation.
The benefits of using AI in Predictive Maintenance
- Precision (proportion of true positives among all predicted positives) was 0.63 for our model
- Recall (proportion of true positives among all actual positives) was 0.83 for our model
- F1 score was 0.71 – an F1 score above 0.5 is considered good
Our AI Models at Different Stages of Predictive Maintenance
DATA COLLECTION & PROCESSING
- Data cleansing, pre-processing
- Feature selection, extraction
ANOMALY/FAULT DETECTION
- Support Vector Machines
- Neural Networks
- Random Forest
PREDICTING OUTAGE
- Convolutional neural networks
- Time series model
- Reinforcement learning
PROGNOSTICS
- Hidden Markov Models
- Recurrent Neural Networks
RECOMMENDATION FOR OPTIMIZATION
- Reinforcement Learning
- Decision Trees
-
Bayesian Networks