What Data for This Category Includes
Daily documents for general mechanical equipment assembly include assembly process cards, BOMs, torque records, clearance adjustment records, commissioning reports, factory inspection sheets, and packing lists. Data sources include PLM/ERP, part inspection data, assembly tool collection, test benches, MES, and customer technical agreements. Updates are adjusted per device or per order. Most document fields contain quantifiable indicators with units, such as torque values, concentricity tolerances, pre-tightening torques, and content is added in real time alongside assembly process nodes.
Constraints Imposed on Workflow Orchestration by These Characteristics
Data sources for this category are scattered, and update frequency is high. Workflow orchestration must support real-time pulling and integration of multiple data sources. Documents contain a large number of quantifiable fields with units. Workflows must use variables to store exclusive parameters for a single device to prevent mixing data across orders. Because assembly processes rely on sequential execution of process nodes, nodeId must be used to mark different assembly links to ensure alignment with process procedure steps. Additionally, the time-consuming step of verifying multi-level BOMs and physical material availability requires embedding knowledge base retrieval and matching links in the workflow to pre-check consistency between assembly materials and process requirements.
Configuration Settings
This configuration is based on FastGPT v4.17.0.
| Configuration Item | Recommended Approach | Rationale |
|---|---|---|
| Data source trigger rule | Trigger per assembly process node | Matches the assembly process sequence for each device per order |
variables | Store BOM number, torque threshold, and clearance tolerance for a single device | Prevent mixing data across orders, adapt to exclusive parameters for a single device |
nodeId | Assign unique identifiers per assembly process, such as "01-base assembly" or "02-drive component installation" | Distinguish process procedures and recorded data for different processes |
| Knowledge base recall limit | Top 8 entries | Covers retrieval needs for multiple document types including BOMs, process cards, and inspection sheets |
| Timeout threshold | 120 seconds | Adapts to average time spent pulling multiple data sources and retrieving from the knowledge base |
| Single retrieval token limit | 8000 tokens | Adapts to average length of assembly documents, avoids context overflow |
Three Common Mistakes
- Phenomenon: The process content returned by knowledge base retrieval in the workflow does not match the current assembly process, and the number of returned results exceeds the preset range. Cause: The retrieval scope is not limited via
nodeId, resulting in recall of all documents. - Phenomenon: Assembly record fields returned in AI conversations are empty, and cannot match the torque or clearance data for the current order. Cause: The exclusive data ID for the single device is not bound in
variables, resulting in the called data source not being associated with the current order information. - Phenomenon: A timeout error triggers when the workflow executes the commissioning report verification link. Cause: A reasonable timeout threshold is not set, or data sources are not triggered per assembly process node, resulting in repeated pulling of redundant data.
How to Verify Correct Configuration
- Manually trigger the assembly workflow for a single device, and check if exclusive order parameters such as BOM number and torque threshold are correctly loaded in
variables. - Switch between process nodes marked with different
nodeIdvalues, and verify that the knowledge base only recalls process procedures and recorded documents for the corresponding process. - Submit a process question with units, such as "What are the torque requirements for base assembly", and confirm that the AI returned answer includes matching quantifiable indicators.
- Simulate abnormal input, such as entering a torque value exceeding the preset threshold, and check if the workflow triggers an abnormal verification link.
The values provided here are common starting points, and should be evaluated against the reader's own samples.