A factory attaches vibration sensors to a stamping press and logs a reading every second for an entire year. Engineers have never recorded an actual equipment failure in that time, so there are no labeled examples of what a breakdown looks like. They want the AI system to learn the machine's normal vibration pattern on its own and flag any reading that deviates sharply from that pattern, so a technician can investigate before a breakdown occurs. Which type of AI workload fits this requirement?

The workload that fits is anomaly detection. The engineers have continuous numeric readings but no labeled failure events, and they want the system to learn what normal operation looks like and flag readings that fall outside that pattern. That is exactly what anomaly detection does: it establishes a baseline from historical data and scores new observations by how much they deviate from it, without needing examples of past failures. Regression forecasting of future sensor values is incorrect because it predicts a specific future numeric value, such as tomorrow's vibration reading, but predicting a value is a different task from deciding whether an already-observed value is abnormal; it does not inherently produce a deviation flag. Image classification is incorrect because it operates on pictures and assigns a category label to a whole image, and nothing in this scenario involves visual data. Clustering of the sensors themselves is incorrect because it groups similar items, such as sensors with similar behavior, into segments; it organizes the population of sensors rather than continuously comparing each new reading against a learned baseline to catch outliers as they happen.