Start with four HIGH alerts

Import the IoT scenario and separate the evidence and master data needed to order four machine inspections.

Question for this chapter

What evidence is needed to choose the first inspection among four alerts?

Why this matters now

Starting with pipeline names can leave you with a successful run but no explanation of why its output changes maintenance work. First, define the roles of the source signal, alert, summary score, and machine relationships.

Try it

iot.zip Download(47 KB)

In the left sidebar, select Collections. On the Collections page, choose the upper-right more (⋯) menu → Import (가져오기) and upload the ZIP. If validation reports no conflicts, select Start import and wait for completion.

Success looks like this

The collection tree shows raw (Raw IoT Collection) and processed (Processed IoT Collection). The current scenario registers its data and executable assets under processed; raw is empty. Import is complete when machine_sensors (Machine Sensor Readings) contains 216 rows.

Korean Portal showing sensor data and anomaly-detection assets in the processed IoT collection — The Portal capture is in Korean. Confirm the source sensor dataset and the alert and health outputs populated later.

Interpret the result

Evidence	Question it answers	Asset to inspect
Source sensor	What value and status signal arrived?	`machine_sensors`
Alert	Which row was classified as anomalous, and why?	`anomaly_detections`
Summary score	How can machines be compared?	`equipment_health`
Relationship context	Who owns this incident, where, and when?	`iot_machine` ontology

Deep dive — assets imported by the scenario

Asset type	Count	Contents
Collections	2	`raw`, `processed`
Datasets	8	sensor readings, anomaly detections, machine health, five master tables
Codes	4	normalization, detection, health rollup, ontology materialization
Pipelines	3	anomaly detection, health rollup, ontology materialization
Ontology	6 entities / 6 relations	all five top categories (Agent, Object, Event, Location, Time)
Knowledge / dashboard	1 each	`maintenance_manual`, `equipment_health`

The five master tables are machines, locations, technicians, shifts, and maintenance_log. They load directly from Parquet alongside the sensor readings, while anomaly_detections and equipment_health are populated after you run pipelines in later chapters.

Next decision

You have separated the roles of the evidence. Next, find the quality_flag values that produced the alert candidates — and the rows that cannot be used — among the 216 sensor rows.