[Paper] IOTEL: A Tool for Generating IoT-enriched Object-Centric Event Logs

Source: arXiv - 2603.07906v1

Overview

The paper introduces IOTEL, a lightweight tool that automatically fuses Internet‑of‑Things (IoT) sensor streams with existing object‑centric event logs (OCEL). By doing so, it lets process‑mining practitioners enrich their logs with real‑world device data without the overhead of custom schemas or massive raw‑data dumps, opening the door to more insightful analyses of IoT‑enhanced business processes.

Key Contributions

• Unified OCEL‑based format: Extends the standard OCEL schema to embed IoT measurements directly alongside traditional process events.
• Automated data‑mapping pipeline: Parses heterogeneous IoT streams (MQTT, CSV, JSON) and aligns them with process objects (e.g., orders, machines) using configurable matching rules.
• Scalable log generation: Selectively aggregates sensor readings (e.g., min/avg/max per time window) to keep log size manageable while preserving analytical value.
• Tooling & UI: Provides a web‑based interface for loading source logs, defining mapping policies, previewing the enriched log, and exporting it for downstream mining tools (e.g., ProM, PM4Py).
• Real‑world validation: Demonstrates the workflow on a manufacturing case study where temperature, vibration, and energy‑consumption data are merged with a production OCEL, revealing hidden bottlenecks.

Methodology

1. Input Acquisition – Users supply (i) an existing OCEL file describing the business process and (ii) one or more IoT data sources (live streams or archived files).
2. Schema Normalisation – IoT payloads are normalised into a common measurement representation (timestamp, sensor‑ID, value, unit).
3. Object‑IoT Alignment – A rule engine matches each measurement to a process object based on identifiers (e.g., RFID tag ↔ product ID) or contextual heuristics (e.g., proximity, time windows).
4. Aggregation & Filtering – To avoid log explosion, the tool aggregates measurements per object per activity (e.g., average temperature during “Assembly”). Users can also filter by sensor type or threshold.
5. OCEL Enrichment – The enriched events are written back into the OCEL JSON structure as additional attributes under the relevant object IDs, preserving compatibility with existing process‑mining libraries.
6. Export & Visualization – The final IoT‑enriched OCEL can be exported for downstream analysis; the UI also offers quick visual checks (timeline plots, heatmaps).

Results & Findings

• Log Size Reduction: Compared to naïvely appending raw sensor rows, IOTEL’s aggregation cut the enriched log size by ≈70 % while retaining >90 % of the variance needed for typical KPI calculations.
• Improved Process Insight: In the manufacturing case, correlating temperature spikes with the “Welding” activity uncovered a previously unknown overheating issue that caused 12 % of rework incidents.
• Tool Compatibility: The generated OCEL files were successfully imported into ProM and PM4Py, enabling standard object‑centric mining (e.g., conformance checking, performance analysis) without any custom parsers.
• User Acceptance: A short user study (n = 8 process analysts) reported a 4.5/5 usability score, highlighting the intuitive mapping UI and the immediate visual feedback.

Practical Implications

• Faster IoT‑Process Integration: Developers can now plug sensor data into existing process‑mining pipelines without building bespoke ETL jobs, accelerating proof‑of‑concept cycles.
• Scalable Monitoring: By aggregating at the activity level, organizations can monitor equipment health in real time while keeping storage and computation costs low.
• Cross‑Domain Analytics: The OCEL‑centric approach makes it easy to combine IoT data with other object‑oriented logs (e.g., case files, resource logs), supporting richer multi‑perspective analyses such as “how does machine vibration affect order lead time?”.
• Vendor‑Neutral Solution: Because IOTEL outputs standard OCEL, it works with any process‑mining tool that supports the format, protecting investments against vendor lock‑in.

Limitations & Future Work

• Mapping Complexity: The current rule engine handles straightforward identifier joins; more complex semantic matching (e.g., fuzzy location inference) requires manual effort.
• Real‑Time Streaming: IOTEL focuses on batch enrichment; extending it to continuous streaming scenarios (e.g., Kafka‑backed pipelines) is left for future development.
• Domain Generalisation: The validation is limited to a manufacturing setting; additional case studies in logistics, healthcare, or smart cities would strengthen claims of broad applicability.
• Advanced Analytics Integration: Future versions could embed out‑of‑the‑box analytics (e.g., anomaly detection) directly into the UI, turning the enriched log into an actionable dashboard rather than just a data source.

Authors

• Jia Wei
• Xin Su
• Chun Ouyang

Paper Information

• arXiv ID: 2603.07906v1
• Categories: cs.SE
• Published: March 9, 2026
• PDF: Download PDF