A phased deployment strategy that builds capability, proves payback, and scales across the fleet
Build capability, scale across fleet, then optimize the portfolio
0–12 months
Stand up the platform, prove first predictive use cases, and achieve payback by month 12–18. Foundation APIs, core modules, and pilot plants operational.
12–24 months
Multi-plant rollout across all technology types. Fleet-wide benchmarking, Computerized Maintenance Management System (CMMS) integration, and automated reporting. Standardized deployment templates.
24–36+ months
Autonomization of routine decisions. Cross-fleet optimization, financial portfolio management, and continuous Machine Learning (ML) improvement from fleet-wide feedback loops.
Each plant progresses through three capability tiers
3–6 months per plant
6–12 months per plant
18–36 months
By end of 2026, the fleet runs a single internal "Industrial Data + AI Platform" where:
How the platform layers build on each other
piwebapi (.NET): fast REST API for operational data – visualization, summaries, extraction
pifastapi (Python): analytics API for statistics, ML, batch processing
pipolars: high-performance Polars SDK + industrial transforms (used by pifastapi)
swapp (core): auth, navigation, workspaces, saved views, API Gateway (routes to piwebapi/pifastapi)
explorer, trend, stats – core reliability workflows
alert, insight, report – predictive intelligence, performance tracking & business reporting
scadanerve: sensors-to-agents pipeline (OPC UA/Modbus/MQTT – historian – feature store – agents)
Detailed deliverables and success metrics for 2026
Status: Production-ready data access layer powering all SWAPP modules and analytics workflows.
Goal: Launch the platform core with first application modules. Establish Single Sign-On (SSO) and Role-Based Access Control (RBAC).
Goal: Deliver the "daily driver" trending and analytics modules as the primary analysis interface.
Goal: Launch predictive modules and connect detection to business workflows.
Goal: Scale the platform and launch business reporting across the fleet.
Strategic independence from proprietary historians with an open, agent-driven data pipeline
OPC UA / Modbus / MQTT streaming from plant sensors with edge preprocessing and buffering
Open-source time-series store with AF-like asset model, replacing proprietary historian dependency
Autonomous anomaly detection and alert explanation agents running per asset class
Scadanerve and PI running side-by-side, validating data parity before full transition
Real-time and batch feature engineering pipeline feeding ML models directly from the open stack
Multi-plant Scadanerve federation with centralized governance and cross-plant data products
Scadanerve starts as a real plant pilot slice in 2027, proving the full sensors – agents – open time-series path
What "AI-first" means in SWAPP – implemented in three tiers
SWAPP at DT3 (Predictive), moving toward DT4 (Optimization)
Key decision points tied to outcomes
explorer.swapp adopted by pilot plants; platform core operational with SSO and API Gateway
trend.swapp and stats.swapp adopted as daily drivers; 50+ recurring users; 20% less dependency on licensed tools
alert.swapp and insight.swapp in use; first predictive alerts dispatched to CMMS; financial impact of avoided outage documented
Multi-plant scaling proven; report.swapp live with Power BI; total avoided-cost tracked in dashboards
Explore use cases and ROI projections for each technology