How EKAS Works
EKAS is a multi-capability manufacturing AI platform built on a governed data layer, a nine-stage security pipeline, and a full provenance enforcement standard. This page describes how the system operates.
Core Components
Data Integration Layer
EKAS connects to ERP systems (Oracle E-Business Suite, SAP S/4HANA, Epicor Kinetic, Infor LN), MES platforms (Plex, Shopfloor, FactoryLogix), SCADA systems (Rockwell, Siemens), and quality databases. Data is ingested via batch extract, API integration, or ODBC connection. No modifications to source systems required.
Governed Metrics Catalog
91 manufacturing metrics across the ISA-95 hierarchy, each defined in versioned SQL with a cryptographic hash. When metric logic changes, the version increments. Historic calculations remain tied to the definition that generated them.
AI Inference Pipeline
Questions asked in plain language are processed through intent classification, entity resolution, evidence collection, answer synthesis, and controlled refusal checks. Enterprise cloud LLM inference with zero-training-data guarantee.
Security and Audit Pipeline
Nine-stage security enforcement: authentication, tenant isolation, role-based access control, query validation, SQL injection defense, row-level security, audit logging, response sanitization, and EvidencePacket attachment.
From Question to Provenance-Backed Answer
User asks question in plain language
No SQL required. No dashboard configuration. Example: "What was OEE for Cell 3 on second shift last Tuesday?"
Intent classification and entity resolution
Platform identifies the workcenter, shift, date, and metric requested. Resolves "Cell 3" to the correct ISA-95 equipment hierarchy node.
Evidence collection before answer synthesis
The governed SQL definition for OEE is executed. Raw production data is queried. Record count, data source, and calculation timestamp are captured.
EvidencePacket construction
All provenance metadata — SQL hash, metric version, time window, record count, data source — is assembled into the EvidencePacket before the answer text is generated.
Answer synthesis with controlled refusal
If the question is answerable with the available data, a response is generated. If not — insufficient data, ambiguous question, out-of-scope request — EKAS refuses with an explanation.
Response returned with full audit trail
The user receives the answer, the EvidencePacket, and a clickable reference to the source data. Every interaction is logged for audit review.
Which Standards EKAS Enforces
ISO 22400-2 OEE Methodology
Ratio-of-sums OEE calculation enforced throughout. No average-of-averages approximations. All OEE results traceable to shift-level input data.
ISA-95 Equipment Hierarchy
Every metric is computed within the ISA-95 structure: Enterprise → Site → Workcenter → Machine → Part. Aggregation rules follow the standard.
IATF 16949 Audit Trail Requirements
Full query logging, metric definition versioning, and data lineage tracking designed to satisfy automotive Tier supplier audit requirements.
Request a Technical Deep Dive
Schedule a technical demonstration with the engineering team.