Glossario IA
Il dizionario completo dell'Intelligenza Artificiale
Consumption anomaly
Significant and unplanned deviation between measured energy consumption and expected consumption, often indicating a technical failure or operational malfunction.
Fault tree
Deductive logical model that represents combinations of elementary events that can lead to energy system failure, used to analyze root causes identified by AI.
Backpropagation diagnosis
Technique using gradients of a neural network to trace a detected anomaly at the output back to the most influential sensors or input variables, facilitating identification of the failure source.
Criticality factor
Weighted score calculated by AI for each detected anomaly, combining probability of failure, potential impact on production, and repair cost to prioritize interventions.
Isolation Forest
Anomaly detection algorithm that isolates observations by building random decision trees, where anomalies are quicker to isolate and require fewer splits in the tree.
Prediction time horizon
Future period over which a predictive maintenance model estimates the probability of failure occurrence, crucial for optimally planning interventions.
Residual load profile
Difference between actual energy consumption and consumption predicted by a reference model, whose pattern analysis reveals the nature and location of failures.
Dynamic alert threshold
Anomaly detection limit that automatically adjusts based on operational conditions (e.g., season, production) to reduce false positives in a variable energy system.
Vibrational signature
A set of unique frequency and time characteristics of machine vibrations, analyzed by AI to identify mechanical defects such as imbalance or faulty bearings.
Mean Time Between Failures (MTBF)
Reliability indicator calculated by AI from failure history, used to calibrate prediction models and evaluate the effectiveness of maintenance strategies.