Easy anomaly detection for industrial applications
29/01/2020
SensiML™ Corporation, a developer of artificial intelligence (AI) tools for building intelligent Internet of Things (IoT) end-points, has announced that its new SensiML Analytics Toolkit delivers quick and easy development of smart sensors for real-time anomaly detection through intelligent IoT end-points. This approach addresses a broad spectrum of industrial equipment, processes and industries. By simplifying the development and implementation of anomaly detection devices, SensiML enables companies to focus their scarce and valuable technical resources on application-specific functionality, not hand-coded machine learning (ML) and dataset manipulation.Anomaly detection works by regularly monitoring equipment and constantly measuring key variables such as temperature, vibration, sound, motion, flow and other time-series sensor data. AI-based analysis is used to establish ‘normal’ and ‘outlier’ behaviour. Detected anomalies can then be flagged for immediate intervention or recorded for later analysis. Anomaly detection can be part of a larger condition monitoring (CM) or predictive maintenance (PdM) programme for a company or can be used independently in a more targeted way.
The SensiML Analytics Toolkit provides a fast and simple way for industrial companies to implement anomaly detection for their specific machines and processes. For example, using SensiML-generated algorithms, technicians can transform traditional sensors into customisable, machine-specific anomaly detectors during initial machine installation or during a later retrofitting. Some of the new functionality available through the use of these ‘smart sensors’ includes:
- Baseline algorithms that can adapt with edge learning to improve insight with use;
- Triggered events that can optionally report feature vector results or corresponding raw sensor data; and
- Simple closed-loop feedback that can involve operators in system tuning as appropriate.
