Detect abnormal equipment behavior by analyzing sensor data
summary
Amazon Lookout for Equipment builds a custom anomaly model per asset using historical sensor data and maintenance logs and reads from S3 to train and then run continuous inference. It accepts up to 300 time-series inputs, automates timestamp alignment, imputation, algorithm and hyperparameter selection, and outputs detections back to S3. Example pump model trains in about 56 minutes, flags five abnormal behaviors on labeled events on average 23h10m before failure with six false positives averaging nine minutes, and CEPSA sees a clear pattern 45 days before a major event.
AI-written summary · about
chapters (11)
- 0:00Intro
- 0:42Unplanned equipment downtime can be costly
- 1:32For high-value applications, investment has been rolling into equipment health monitoring
- 2:41Current analytics methodologies have their limitations 1. Single variable analysis
- 5:49In reality, the industrial space is challenging
- 8:06Introducing Amazon Lookout for Equipment
- 9:33How Amazon Lookout for Equipment works
- 14:12What is happening under the hood
- 22:07CEPSA equipment monitoring pain points
- 23:24CEPSA and Amazon Lookout for Equipment
- 24:49Using Lookout for Equipment in current workflows
related talks
38mAnomaly detection in the cloud: how to detect issues in data [...]
30mPredictive Maintenance and Anomaly Detection for Wind Energy
53mDeveloping new findings using machine learning in Amazon GuardDuty
19mEnterprise AI-Powered Predictive Maintenance Platform
01h 06mAmazon Machine Learning to Predict Responses
01h 01mBetter reliability and uptime with analytics and observability
26mServerless Architecture for Product Defect Detection
29mThresholds Are for Old Threats: Demystifying AI and Machine Learning to Enhance SOC Detection