VEXPLOR
VEXPLOR INSIGHT

When the AI is wrong, the factory stops.So it gives no answer without evidence.

Insight is the manufacturing AI analytics module that answers with evidence attached. It follows definitions and relations to find evidence in company records and gives no answer without it. In the demo environment it diagnoses root causes from five lines of evidence.

EVIDENCE SEARCH [PRODUCT SPEC] · 5-EVIDENCE DIAGNOSIS · SHAPLEY · WHAT-IF [DEMO ENVIRONMENT]
VEXPLOR INSIGHT · EVIDENCE SEARCH (ASK)
VEXPLOR live service screen — Ask: one item code follows definitions and the graph to show 19 related records and the relations followed as evidence (Logic Studio demo tenant, sample data, fictitious company names)
FIG — LIVE SERVICE · ANSWERS FOUND BY FOLLOWING DEFINITIONS AND THE GRAPHEvidence search in Logic Studio · demo tenant
VEXPLOR INSIGHT · ROOT CAUSE DIAGNOSIS
VEXPLOR Insight running in the demo environment — root cause diagnosis: five lines of evidence (vibration, forming-force residual, defect-rate trend, peer press comparison, ontology causal link) point to the same cause (die wear), with the defect contribution broken down. The causal link values are preset and labeled as such (simulated data)
FIG — DIAGNOSIS CHECKING EACH OF FIVE LINES OF EVIDENCE AGAINST THE CANDIDATE CAUSEDemo environment · Simulated data
QUERY → RETRIEVE [ DEFINITIONS · RELATION GRAPH ] → GROUNDING CHECK [ no evidence, no answer ] → ANSWER + EVIDENCEFIG — EVIDENCE-FIRST PIPELINE · No evidence, no answer [Product spec]
01PROBLEMPLAUSIBLE IS NOT ENOUGH

One quality incident,
three hours of root-cause analysis.

The data is there, but the answer comes late. General-purpose chatbots produce plausible wrong answers — and on a factory floor, the price of a wrong answer is downtime.

  • Search with evidenceFollows definitions and relations to find evidence in company records and gives no answer without it — a live Logic Studio function.PRODUCT SPECDEFINITION · RELATION GRAPH — NO EVIDENCE, NO ANSWER
  • Root-cause diagnosis from five lines of evidenceChecks one by one whether vibration, forming-force residual, defect-rate trend, peer equipment and the equipment relation definition (preset) fit the candidate cause, and shows how many agree.DEMO ENVIRONMENT5 EVIDENCE CHECKS · NORMAL = NO SIGNAL
  • Contribution and what-ifSplits defect contribution by cause with Shapley values on the process model and previews changed conditions in a branched simulation.DEMO ENVIRONMENTSHAPLEY CONTRIBUTION · WHAT-IF BRANCH
PRODUCT BENCHMARK3 hours → 30 secQuality root-cause analysis
PRODUCT BENCHMARK2 hours → 1 minEquipment impact analysis
PRODUCT BENCHMARK4 hours → 1 minRecall scope identification

Applied across six manufacturing-AI project sites.

Predictive maintenance, quality prediction and agents are built as on-site projects.

Cases (KR)Request a Diagnosis