Fuse Inspection, Work-Order, and Sensor Data

Use this recipe when a facility or manufacturing team wants one reviewed view of equipment condition, maintenance activity, and sensor behavior.

The output can support:

• facility operations review;
• predictive maintenance evidence preparation;
• work-order prioritization;
• AI Agent context;
• BI reports for reliability or service teams.

Outcome

At the end of this workflow, you should have:

• inspection dataset;
• work-order dataset;
• sensor history dataset;
• shared asset identity mapping;
• fusion task;
• reviewed conflicts;
• output dataset ready for downstream use.

Step 1: Prepare inspection data

Create or select a dataset with:

• inspection ID;
• asset ID or equipment reference;
• inspection time;
• finding category;
• severity;
• notes;
• inspector or team;
• attachment reference when needed.

Preview rows and confirm that the asset reference can be matched to the operational asset model.

Step 2: Prepare work-order data

Create or select a dataset with:

• work-order ID;
• asset ID or equipment reference;
• created time;
• closed time;
• status;
• priority;
• work type;
• failure code;
• action notes;
• owner team.

If the source is CMMS or EAM, use DFS Lite REST, JDBC, CSV, or Fabric connector workflows depending on the deployment.

Step 3: Prepare sensor history

Create or select a dataset with:

• asset ID;
• timestamp;
• metric name;
• value;
• unit;
• quality flag;
• source connector;
• ingestion time.

Use the predictive maintenance signal history recipe for detailed signal preparation.

Step 4: Align asset identity

Before fusion, confirm that all datasets can refer to the same asset.

Check:

• asset ID;
• equipment name;
• alias;
• location;
• source-system code;
• parent system;
• site.

Create mapping corrections before running fusion when identity is ambiguous.

Step 5: Choose fusion mode

Use Rule Matching when shared keys are available.

Examples:

• same asset ID;
• same work-order ID;
• timestamp within a defined window;
• same source-system equipment code.

Use Semantic Matching when descriptions, aliases, or notes need comparison.

Use LLM Assisted when language-based help is useful and all uncertain output will be routed to review.

Step 6: Configure the fusion task

Go to:

Data Integration > Data Fusion

1. Create fusion task.
2. Select inspection, work-order, and sensor datasets.
3. Choose fusion mode.
4. Select a method when reusable logic is needed.
5. Set conflict threshold.
6. Set output dataset name.
7. Save.
8. Run the task.

Step 7: Review conflicts

Open Review Queue for conflicts, source disagreement, low confidence, or manual flags.

Review:

• source record;
• matched record;
• asset identity;
• timestamp window;
• severity or status;
• confidence;
• reason;
• downstream effect.

Approve, reject, or manually resolve each item with a short note.

Step 8: Validate the output dataset

Open the fusion output dataset in Dataset Center.

Check:

• preview rows;
• profile;
• row count;
• required columns;
• lineage;
• steward;
• quality issues.

Validate the dataset after review is complete.

Step 9: Use the output

Typical downstream uses:

• BI report showing open maintenance risk by asset;
• AI Agent workflow that summarizes recent evidence;
• predictive maintenance workflow that uses work history and signal features together;
• Inspector workflow that shows relevant maintenance context during field review.

Troubleshooting

Symptom	Check
Many records are unmatched	Asset identity mapping, timestamp window, and source-system aliases.
Review queue is large	Conflict threshold, matching mode, and method logic.
Output rows duplicate	Matching key, join logic, and duplicate source records.
Sensor values are missing	Sync history, quality dashboard, and mapping rules.
Work-order context is stale	Connector refresh cadence and source query filter.

Next step

Use Create an AI Agent-Ready Dataset when this output should be used by an Agent workflow.