ARTi-Solutions®  analytics architecture

​

Inspections

Calculation of the so-called "operation modes" that require clustering using Principal Components extracted from the entire set of variables using highly sophisticated density-based clustering, such as DBSCAN. Commonly used clustering, such as k-means variance, scale up to even cluster size and flat geometry. Our solution scales up to non-flat geometry and uneven cluster sizes that are physically realistic for the industry problems. In-house development has lifted the problem of limited size requirements of the original DBSCAN algorithm and ARTi-solutions offers usability for unlimited size of records.

​

Diagnosis

A suite of different approaches for fault / anomaly detection that can be used appropriately in a complementary (hybrid) manner to build case-specific rules. The first approach offers results based on the entire set of variables and records and automatically calculated thresholds that classify whether a record is a fault or not without user intervention. In addition, single variable anomalies are detected based on (linear and non-linear) predictions of what the upcoming value should be. If the measurement differs unexpectedly from the prediction, then this measurement is flagged as "anomaly".

​

Root Cause Analysis

Based on all variable anomalies for short, medium and long time-frames between anomaly events and how a preceding event is affecting the following ones. Once the calculations are completed, the correlations between variables with respect to outliers (anomalies) is produced. Then, the highly correlated variables are grouped. A graph (Causality Graph) is produced based on temporal precedence of events.