Product Vision

Chapter 11 — Product Vision

What RAPID AI Looks Like

RAPID AI is an AI-based prescriptive maintenance platform built on a specific premise: every recommendation must be traceable to physics and engineering rules. It is not a black box that outputs risk scores from statistical correlation. It is an engineering intelligence system that reasons about why machines fail and what to do about it.

The target user is a reliability engineer or maintenance planner — someone who understands machines but needs a system that can monitor hundreds of assets simultaneously and surface the ones that need attention, with engineering justification for every alert.

The Three Tracks as Product Tiers

The philosophical framework from Chapter 2 translates directly into product capabilities:

Track 1 Product: Real-Time Diagnostic Intelligence

What it does: Monitors sensor data streams, validates signal quality, detects fault patterns using 119 physics-based rules, classifies trends, computes entropy-based stability, fuses evidence into system health scores, and prescribes maintenance actions with engineering justification.

User experience: The reliability engineer opens the Plant Overview dashboard at 7 a.m. and sees 400 monitored assets color-coded by health state. Three assets show red (alarm), seven show amber (warning), fifteen show yellow (watch). The engineer clicks on pump P-101A (red) and sees:

Health Stage: Critical (SSI = 0.84) Diagnosis: Bearing outer race spalling — BPFO frequency trending upward with envelope energy rise and temperature delta increasing at 2C/week Evidence: AFB rule match (confidence 0.91), accelerating trend (slope 0.07), declining stability (SI = 0.38) RUL: 12 days (Weibull-adjusted, 30-day failure probability: 78%) Action: ACT005 — Bearing replacement (planned). Priority: 87 (Immediate). Schedule within 24 hours. Physics: Energy cannot flow freely through the DE bearing due to progressive outer race surface damage (spalling). Strain energy is accumulating (displacement trending upward) and friction losses are increasing (velocity rising). The entropy analysis confirms loss of spectral order — energy is spreading from discrete bearing defect frequencies into broadband noise, indicating advanced damage.

Every number in that assessment traces back through the IMS to a specific sensor measurement, rule evaluation, and physical principle. The engineer can drill down into any element to see the evidence chain.

Track 2 Product: Dynamic RCM Workbooks

What it does: Maintains living RCM workbooks for every monitored asset, with failure probabilities and strategy selections that update automatically from sensor data. Provides a structured decision framework for maintenance planning that goes far beyond “replace when broken.”

User experience: The maintenance planner opens the RCM Workbook for pump system SYS-CW-001. The workbook shows 22 failure modes across the pump, motor, coupling, and seal. Each failure mode has a current RPN that reflects both the static design assessment and the real-time sensor evidence:

FM-003 (Cavitation): RPN was 36 (moderate) at commissioning. Current suction pressure trending low -> probability adjusted from 3 to 4 -> RPN now 48 (approaching high). Strategy automatically upgraded from “routine monitoring” to “increased frequency monitoring + process review.”
FM-008 (Bearing outer race): RPN was 32 at commissioning. Envelope spectrum now showing BPFO -> probability adjusted to 5 -> RPN now 40. Strategy: CBM with planned bearing replacement.
FM-014 (Stator insulation): RPN is 60 (high). Last Megger test showed IR declining. Next scheduled test moved up from 6 months to 3 months.

The planner can see the complete maintenance plan for the next 90 days, with tasks prioritized by dynamic RPN and scheduled around production windows.

Track 3 Product: Imperfection Analysis and Design-Out Engineering

What it does: Identifies structural weaknesses in equipment design, installation, and operation that cause chronic failure. When Module G detects a design contradiction, it generates an engineering change recommendation with estimated ROI.

User experience: The reliability manager reviews the quarterly imperfection report. RAPID AI has identified three chronic failure patterns:

Chronic #1: Cooling tower fan bearing CT-FAN-301. Bearing replacement every 18 months despite correct maintenance. Root cause: Design contradiction CT02 — clearance set tight for low vibration at operating temperature, but during cold starts the clearance is too tight, crushing preload. Recommendation: Specify bearing with temperature-compensated clearance. Estimated annual savings: $45,000 in bearing replacements + $120,000 in avoided unplanned downtime.

Chronic #2: Pump P-205 seal failure every 14 months. Root cause: Shaft overhang ratio violation (IMP_001, overhang/diameter = 1.8, limit 1.5). Excessive bending deflection at seal face. Recommendation: Machine shaft to reduce overhang, or install stiffening bearing. Estimated ROI: $200,000 over 5 years.

Chronic #3: Gearbox GB-401 tooth pitting recurring after every overhaul. Root cause: Foundation looseness (FND03) amplifying mesh forces. The gearbox design is sound but the installation amplifies loads. Recommendation: Foundation repair and grouting. Estimated cost: $15,000 one-time vs. $40,000/year in recurring overhaul costs.

Each recommendation includes the full evidence chain, FRETTLSM factor classification, IAR analysis, and economic justification.

5-Product Suite Architecture (Production)

The production frontend delivers five distinct products within a single SvelteKit application. Each product has its own route, accent color, and identity — expressed through CSS tokens and data-product wrappers, sharing a single auth session and sidebar layout.

Product	Route	Purpose
COMMAND	`/`	Fleet dashboard — attention queue, health dot grid, FleetPulse breathing-glow animation. Surfaces highest-priority equipment.
FLEET	`/fleet`	Hierarchy browser — canvas/table toggle, SchemaBuilderPanel. @xyflow/svelte flow diagrams. Click a node → Equipment Hub.
OPERATIONS	`/operations`	Alarms, work orders, PM schedules, spare parts inventory. Operational planning hub.
DIAGNOSTICIAN	`/diagnose` + Ctrl+K modal	Signal analysis runner, AI RCA with evidence chains, pipeline explorer. Also accessible via keyboard shortcut.
EQUIPMENT HUB	`/equipment/[id]`	Machine detail destination (6 tabs). Reached from Fleet hierarchy, not from sidebar directly.

Sidebar: Command · Fleet · Operations · Diagnostician · Admin · Settings (6 items).

This 5-product architecture replaced a previous multi-app suite (Explorer, Invent, Monitor, Suite, Workshop) — consolidated into one deployment for simpler operations and shared state.

Dashboard Design

The dashboard presents six primary views:

Asset Health Card

The primary per-asset view. Displays:

Health stage with color coding (green/yellow/amber/red)
Severity score (0-1 scale with numeric display)
Confidence level (with qualitative label)
Active fault diagnosis with matched rule ID and physics basis
Trend visualization (sparkline showing last 30 readings with trend class annotation)
AESF state indicator (S0-S4)
RUL estimate with confidence band

Data source: GET /dashboard/assets/{assetId}/health-card

A persistent top-level alert for critical and unstable assets. Appears when any asset reaches SSI >= 0.60 or priority score >= 70. Shows:

Asset ID and location
RPN and consequence category
Recommended escalation path
Time since alarm onset

Maintenance Action Tile

Displays the recommended maintenance task for assets requiring intervention:

Action title and catalog ID (ACT001-ACT015)
Action window (Immediate / 24h / 7d / Next shutdown)
Responsible team
Spare requirements and availability status
Link to the specific IMS row that generated the recommendation

Highlights assets with safety or environmental consequence classifications:

Compliance status against regulatory thresholds
Overdue safety-critical inspections
Safety-consequence failure modes with current RPNs
Escalation status and responsible parties

Performance Loss Tile

Quantifies economic impact of detected imperfections:

Efficiency degradation percentage
Estimated energy waste (kWh/month)
Production impact (units/day)
Cumulative cost of imperfection since detection
Comparison to design-out investment required

Plant Overview

The executive view. Aggregates all assets at a site into a single prioritized display:

Count of assets by health state (stacked bar: healthy/watch/warning/alarm)
Top 10 priority action items ranked by priority score
Plant-wide availability trend (30-day rolling)
Overall risk profile (aggregate RPN distribution)
Maintenance backlog by team and urgency

Data source: GET /dashboard/plant-overview?site={site}

Design Principle

Each widget renders from structured API output — the dashboard never computes diagnostic logic client-side. This preserves the IP protection boundary and ensures every number on screen traces back through the pipeline to the physics that produced it.

The LLM-Augmented Diagnostic Copilot

The future product roadmap includes a conversational interface where reliability engineers can ask questions in natural language and receive engineering-grounded answers:

Example interaction:

Engineer: “Why does pump P-205 keep failing at the seal?”

Copilot: “Pump P-205 has experienced 4 mechanical seal failures in the last 5 years (FM-006, recurring). RAPID AI’s imperfection analysis identifies the root cause as shaft overhang ratio violation (IMP_001): the overhang-to-diameter ratio is 1.8, exceeding the 1.5 engineering limit. This creates excessive shaft deflection at the seal face, causing uneven contact pressure and accelerated seal wear.

The FRETTLSM classification is F003 (Pipe strain / shaft geometry) as Initiator, with L007 (Under-lubrication) as Accelerator at the seal face.

Recommended resolution: Machine the shaft to reduce overhang, or install a stiffening bearing between pump and seal. Estimated one-time cost: $25,000. Estimated annual savings: $40,000 in seal replacements plus avoided unplanned downtime.”

The copilot is not a general-purpose chatbot. It is an orchestration layer that queries the RAPID AI diagnostic engine, retrieves specific IMS rows and rule evaluations, and synthesizes the results into engineering language. Every statement the copilot makes is backed by a specific data point from the pipeline — not generated from statistical language patterns.

Product Architecture Summary

                    RAPID AI Product Architecture

Users:  Reliability Engineer  |  Maintenance Planner  |  Plant Manager
              |                        |                       |
              v                        v                       v
        +-----------+          +-----------+          +-----------+
        | Health    |          | RCM       |          | Plant     |
        | Cards     |          | Workbooks |          | Overview  |
        +-----------+          +-----------+          +-----------+
              |                        |                       |
              +------------------------+-----------------------+
                                       |
                               SvelteKit Dashboard
                                       |
                               REST API (FastAPI)
                                       |
              +------------------------+-----------------------+
              |                        |                       |
     Signal Intelligence     Stability Intelligence    Engineering Intelligence
     (GUARD + SENSE)         (FUSE)                    (ACT + INSIGHTS)
              |                        |                       |
              +------------------------+-----------------------+
                                       |
                          PostgreSQL + pgvector
                          (IMS + Rules + Seed Data)

Market Position

RAPID AI occupies a unique position in the industrial maintenance software market:

Capability	Generic PdM	RAPID AI
Anomaly detection	Statistical pattern matching	Physics-based rules with engineering explanation
Root cause analysis	”Anomaly in bearing"	"Bearing spalling due to lubrication starvation (AFB03) initiated by contaminated grease (L004), accelerated by thermal cycling (TT001)“
Maintenance recommendation	”Schedule inspection"	"ACT005: Replace bearing in planned window. Check grease supply for contamination source. Verify thermal cycling exposure.”
Design improvement	Not available	”Contradiction CT02: Clearance specification conflicts with thermal growth. Redesign to temperature-compensated clearance.”
Audit trail	Model version	Full evidence chain from sensor measurement through physics rule to dashboard message

The competitive advantage is explainability. In an industry where a maintenance decision can cost $500,000 (whether by acting or by not acting), the engineer needs to understand why the system is making a recommendation. Black-box confidence scores do not provide that understanding. Physics-based evidence chains do.

User Journeys

The product design begins with three primary user personas and their daily workflows.

Journey A: The Reliability Engineer — “A Pump Is Showing Increasing Vibration”

The reliability engineer logs into RAPID AI at 7 a.m. The Plant Overview shows three red assets. She clicks on P-101 (Cooling Water Pump) and sees the Asset Health Card:

Health Score drops from 62 to 38 over two weeks. Trend class: accelerating.
She clicks “View Diagnosis” — the system has already run the diagnostic pipeline overnight. Result: AFB03 (bearing lubrication starvation), confidence 0.88. Evidence: BPFO frequency rising, envelope energy elevated, bearing temperature trending upward at 2C/week.
She clicks “Evidence Trail” and sees the complete chain: Module A validated the signal (DG003 passed, bandwidth OK), Module B matched AFB03 with directional ratio A/H = 1.4 (isotropic), Module B.2 classified the trend as accelerating (slope 0.07, NLI 0.62), Module C fused evidence to SSI = 0.84 (alarm state).
She clicks “RCM Decision” — Module E recommends ACT005 (bearing replacement, planned), priority 87, schedule within 24 hours. The RPN is 48 (severity 4, probability 4, detectability 3).
She reviews the maintenance plan: replace DE bearing, inspect lubrication supply, verify grease type and quantity, check for contamination at grease nipple. Spare parts: bearing kit, grease, labyrinth seal.
She exports the work order to the CMMS, attaching the full evidence chain as documentation.

Total time from dashboard login to work order creation: 8 minutes. Without RAPID AI, the same investigation would require manual vibration data analysis, spectral interpretation, historical comparison, and report writing — approximately 2-4 hours per asset.

Journey B: The Maintenance Planner — “What Needs Attention This Week?”

The maintenance planner opens the Plant Overview each Monday morning:

Priority Action List shows 22 tasks ranked by priority score. Top 5 are color-coded red (priority > 70).
He filters by maintenance window: “next shutdown” (12 items), “next available” (6 items), “within 24 hours” (4 items).
For each task, he sees the resource requirements: team assignment, estimated duration, spare parts needed, availability status from the parts inventory.
He builds the weekly maintenance schedule, dragging tasks into available time slots. The system warns when two tasks compete for the same crane or the same isolation boundary.
He exports the weekly schedule to the CMMS and sends notifications to the assigned crews.

Journey C: The Plant Manager — “Where Is My Risk?”

The plant manager reviews RAPID AI quarterly for capital planning:

Risk Profile shows the distribution of assets by health state across the entire site. This quarter: 82% healthy, 12% watch, 4% warning, 2% alarm.
Chronic Failure Report (Track 3 output) identifies three recurring failure patterns that cannot be solved by maintenance alone. Each includes a design-out recommendation with estimated ROI.
Maintenance Effectiveness trend shows whether the ratio of planned to unplanned maintenance is improving. Target: 80% planned. Current: 68%. Trend: improving at 2% per quarter.
Budget Impact shows the estimated cost avoidance from RAPID AI interventions this quarter: $1.2M in avoided unplanned downtime, $340K in optimized maintenance scheduling, $180K in extended component life.

Implementation Roadmap

The product vision maps to a phased delivery strategy:

Phase 1: Foundation (Months 1-3)

Delivered: Data ingestion, asset hierarchy, signal processing, basic fault detection. The system can ingest sensor data, validate signal quality, and detect the most common fault patterns (bearing defects, misalignment, imbalance) using Module A and Module B rules.

User value: Automated fault detection with physics-based explanation. Replaces manual spectral analysis for common faults.

Phase 2: Intelligence (Months 4-6)

Delivered: Trend analysis (Module B.2), entropy monitoring (Module B.3), system fusion (Module C), health staging (Module D), and maintenance planning (Module E). The system can track degradation trajectories, fuse multi-parameter evidence, stage health conditions, and prescribe maintenance actions.

User value: Complete diagnostic pipeline from sensor to action. The reliability engineer receives actionable maintenance recommendations with full evidence chains.

Phase 3: Strategy (Months 7-9)

Delivered: RCM framework, dynamic RPN, RUL estimation (Module F), contradiction detection (Module G), imperfection analysis. The system can maintain living RCM workbooks, estimate remaining useful life, and identify design contradictions.

User value: Strategic maintenance optimization. Maintenance resources are allocated by risk rather than by schedule.

Phase 4: Innovation (Months 10-12)

Delivered: FRETTLSM causal analysis, design-out engineering, engineering copilot, learning and governance. The system can perform root cause analysis using the 88-factor taxonomy, recommend design changes, and support natural language queries.

User value: Closed-loop failure elimination. Chronic failures are identified and designed out permanently.

Integration Architecture

RAPID AI is designed to integrate with existing plant infrastructure rather than replace it:

System	Integration	Direction
Historian (OSIsoft PI, Aveva, Honeywell)	Sensor data ingestion via REST or OPC-UA	Inbound
CMMS (SAP PM, Maximo, Infor EAM)	Work order creation, spare parts lookup	Outbound
SCADA/DCS	Real-time process variables (flow, pressure, temperature)	Inbound
Online monitoring (Bently Nevada, SKF, Emerson)	Vibration waveform and spectral data	Inbound
ERP (SAP, Oracle)	Cost data for ROI calculations	Inbound
Document management	Engineering drawings, maintenance records	Bidirectional

The integration layer uses standard protocols (REST API, OPC-UA, MQTT) and can operate in three deployment topologies:

Cloud-hosted — RAPID AI runs in the cloud, sensor data is forwarded from the plant network
Hybrid — Diagnostic engine runs on-premise for data sovereignty, dashboard and reporting in the cloud
Fully on-premise — Complete deployment within the customer’s network, no external connectivity required

The deployment topology is a customer choice driven by data governance requirements, network architecture, and regulatory constraints. The diagnostic intelligence is identical in all three topologies.

Next: Chapter 12 — Appendices Previous: Chapter 10 — Implementation

Standards Alignment

Standard	Relevance to This Chapter
ISO 55000/55001 — Asset management	The three-track product model (CBM, RCM, Design-Out) implements ISO 55000’s value-driven asset management lifecycle, from operational monitoring through strategic maintenance optimization to capital improvement.
ISO 13374 — Condition monitoring and diagnostics of machines	The product vision describes user experiences that directly expose ISO 13374 processing outputs (health assessment, prognostics, advisory generation) through the dashboard interface.
SAE JA1011/JA1012 — RCM evaluation criteria	Track 2 (Dynamic RCM Workbooks) implements living RCM workbooks that satisfy SAE JA1011’s seven questions with continuously updated risk scores from sensor data.
MIMOSA OSA-CBM — Open System Architecture for CBM	The product’s integration with existing plant historians, CMMS, and SCADA systems follows OSA-CBM’s open architecture principles for condition-based maintenance platforms.

Changelog

Version	Date	Author	Changes
2.1.0	2026-03-17	Rick D	Added standards alignment, living doc metadata, changelog
2.0.0	2026-03-17	Rick D	Enriched with production codebase content
1.0.0	2026-03-17	Rick D	Initial chapter creation