Short answer: automate a heat-treatment batch report by defining the approved report fields first, linking each field to a trustworthy source, freezing the calculation and template versions, and generating the PDF or workbook from the completed batch record—not from manual copy-and-paste.
A colourful graph is often the visible win. The deeper improvement is consistency: the same identity rules, statistics, event handling and layout every time, with a clear path back to the raw log.
Start with a content contract
Ask the people who create, approve and consume the report what must travel with a batch. Typical candidates include:
- batch, heat, customer, job and furnace identity where applicable;
- start/end time, recipe or setpoint stages and operator context;
- channel names, sensor roles and units;
- trend graph, selected statistics and soak information;
- alarms, interruptions, acknowledgements or comments;
- approval/sign-off fields and report-template version.
Not every report needs every field. A single-batch service may need a concise professional record, while recurring production may require searchable history and automatic generation at batch close.
Map every visible value to a source
| Report element | Preferred source | Risk to control |
|---|---|---|
| Batch identity | Validated operator entry or upstream system | Wrong or missing ID carried into every page |
| Temperature series | Protected acquisition archive | Gaps or resampled data hidden by a smooth graph |
| Statistics | Versioned calculation method | Different people using different ranges or formulas |
| Alarms/events | Timestamped event log | Only showing events someone chose to remember |
| Corrections | Authorised sensor/instrument record | Corrected values that cannot be reconstructed |
What the finished record can look like
The preview below is deliberately illustrative, but it shows the difference between a chart export and a usable batch record. Identity, time range, channel summary, an event trail, review state and page/version information travel with the graph.
| Group | Min | Max | Mean |
|---|---|---|---|
| CH 1–3 | 968.2 | 981.4 | 975.1 °C |
Analysis window begins
Operator note retained
Worked mapping: one value, one accountable source
Suppose the visible report field is “Batch HT-0842, analysis-window mean 975.1 °C.” The batch identifier might originate from a validated operator entry or an upstream production system. The analysis-window start and end might come from approved process-stage markers. The mean should be calculated from clearly identified channels, using an explicitly versioned rule for invalid values and corrections. The PDF should retain those source references or an internal record key so the number can be reconstructed.
The report generator should not accept a manually typed mean merely because it looks reasonable. It should also avoid silently recalculating an older batch with today's changed rules. When regeneration is allowed, keep the original released output and identify the template/calculation version used for the new rendering.
Design the generation trigger
The report may be generated on batch completion, after an authorised review, on demand for an older record, or by a scheduled workflow. Decide what happens when the batch ID is missing, a channel is invalid, logging contains a gap or the PDF service is unavailable. A failed report must not silently become a successful batch.
Separate batch complete, report generated and report approved as distinct states. This prevents a technical export failure from changing the historical process record and prevents an unreviewed draft from being mistaken for an approved report. If the plant does not require approval, define the release state and responsible owner explicitly.
Do not let the report create evidence that the logger did not capture
Automation can format, calculate and assemble evidence; it cannot repair missing source data by drawing a smooth line across a gap or inventing an alarm acknowledgement. Invalid, stale or missing data needs an explicit quality state. If corrections are applied, retain the raw value and correction provenance. If values are resampled for presentation, keep the original archive and disclose the display method.
Clock handling also deserves design attention. Furnace acquisition, operator events and upstream batch systems can carry different clocks or time zones. Store an unambiguous time basis and make any alignment rule reviewable, especially when a report combines sources.
Validate with known examples
- Compare automated values with an independently checked sample batch.
- Test missing fields, interrupted logs, outliers and a regenerated older report.
- Confirm page breaks, legends, units and print readability.
- Record the report template and calculation version in the output.
- Keep the raw record available so a reviewer can explain every result.
Common automation mistakes
- Starting with the PDF: polishing a template before agreeing the underlying identities and sources simply automates inconsistency.
- Free-text everything: uncontrolled furnace, recipe and batch names make search and retrieval unreliable; validate key identifiers while preserving useful notes separately.
- Statistics with an invisible window: readers need to know which period and channels produced the displayed minimum, maximum, mean or spread.
- Missing-data cosmetics: smoothing, interpolation or omitted invalid points can make the report look complete when the source is not.
- Overwriting a released report: corrections and regenerated outputs need version history, reason and accountable approval where required.
- One workstation as the archive: define indexing, authorised access, backup, retention and export before the report becomes business evidence.
- Email as the only trace: an attachment in a mailbox is a delivery copy, not the authoritative batch record.
Acceptance test the workflow, not only the layout
Use a controlled set of known records: a normal batch, a missing identifier, a channel gap, an outlier, an alarm/event, a correction, a report-service failure and an older record regenerated under a newer template. Confirm that each state is truthful, no source record is overwritten, the output remains readable on screen and print, and an authorised person can retrieve the correct version later.
Use the report builder to draft a content checklist. Pyrolog reports can then be engineered around the plant's actual fields rather than a fixed catalogue template.