Know what was measured.
Methane and total biogas are distinct targets. Dataset provenance, units, reactor identity and timestamps stay attached to each study.
RESEARCH & EVIDENCE
Make the reasoning
part of the result.
The scientific foundation is physics-guided AI for anaerobic digestion: temporal prediction, soft-sensor objectives, process constraints and reproducible evaluation. A product interface is not, by itself, validation.
Respect the timeline.
No random time-series split. Preprocessing and feature construction must not learn from future validation or test observations.
Report what wins.
A sequence model is not automatically better. Compare persistence, XGBoost and LSTM, retain limitations and do not hide negative results.
Follow the evidence.
These links lead to existing readers and consoles, not newly generated findings. Retained epoch displays are not continuously running training.
EvidenceData & models consoleRetained evidence / existing APIEvidenceModel training & comparisonsRetained training evidence; not a training jobEvidenceIndustrial dataset studyTarget and source distinctions retainedEvidenceMethod & equationsPublic methodology readerEvidenceEvaluator responseRecorded suggestions and responsesEvidenceEvidence portalPublic explanatory chapters
PUBLIC AND PROTECTED, KEPT DISTINCT
Research has context.
And access boundaries.
Public status and explanatory readers remain separate from manuscripts and reviewer materials. Protected access is not weakened by this redesign.