Applied AI Blueprint: Vehicle Maintenance

The Reality

Tire condition on rolling stock (wheel tread profiles) and road fleets (rubber tire tread depth, sidewall damage, uneven wear patterns) is a safety-critical parameter governed by strict regulatory thresholds. Manual tire inspection is slow, subjective, and inconsistent — different inspectors measure differently, under different lighting conditions, at different points on the tire circumference. For railways, wheel tread profile measurements are critical for safety. A flange that's too thin risks derailment. A flat spot from emergency braking creates vibration that damages the bogie. For bus and truck fleets, tire inspection happens during vehicle walkarounds — a visual check that catches only obvious damage, not gradual tread depth reduction that approaches the legal limit.

The Proven AI Capability

At the world's second-largest train manufacturer, Sensfix measures brake pad thickness with sub-millimeter precision from standard camera imagery as each train passes through the depot entry point. The same dimensional measurement pipeline — camera captures a component profile, AI extracts measurements, rule engine compares against tolerance — applies directly to wheel tread profile measurement and tire tread depth analysis.

How It Works

Cameras positioned at the depot entry (or fleet parking lane) capture tire/wheel images as each vehicle passes. The AI extracts tread depth measurements across multiple points, identifies uneven wear patterns (indicating alignment issues), detects sidewall damage (cuts, bulges, cracking), and tracks wear progression per vehicle over time. The Rule Engine flags vehicles approaching minimum thresholds and predicts replacement timing based on wear rate trends.

Applicable Modules

The Reality

Modern rolling stock relies on complex electrical systems — traction motors, auxiliary power inverters, battery banks, wiring harnesses, circuit breakers, and passenger information system displays. Electrical faults account for a significant percentage of unplanned vehicle withdrawals from service. Diagnosing electrical issues traditionally requires specialized technicians with multimeters, oscilloscopes, and thermal cameras — expensive, slow, and scheduled rather than continuous. Early signs of electrical degradation — hot spots on connections, discolored wiring insulation, arc damage marks on contactors — are visible but subtle. They escape notice during time-pressured depot inspections focused on mechanical and safety checks.

The Proven AI Capability

ServiceScanAI's 42+ defect detection models identify surface anomalies from standard imagery — including discoloration, heat damage marks, and material degradation that indicate electrical stress. At the European WWTP, thermal deviation monitoring detects overheating on motor windings. At the 5G manufacturing facility, vibration signatures from electric motors distinguish between mechanical and electrical fault origins (imbalance vs. winding asymmetry).

How It Works

During depot inspections, technicians use the Sensfix mobile app to photograph electrical cabinets, junction boxes, and wiring runs. The AI flags: discolored or swollen connections (indicating overheating), arc marks on contactor surfaces, insulation cracking or melting, corrosion on battery terminals, and display failures on passenger information systems. Thermal cameras on the same mobile device detect hot spots that invisible-light cameras cannot see. The Rule Engine correlates visual findings with operational data (e.g., a hot connection on a traction motor contactor + reduced acceleration performance = priority intervention).

Applicable Modules

The Reality

Rolling stock and fleet vehicles operate in corrosive environments — salt spray on coastal routes, de-icing chemicals on winter roads, moisture ingress from power washing, and galvanic corrosion at dissimilar metal joints. Corrosion is progressive — by the time it's visually obvious, structural integrity may already be compromised. Manual inspection catches only surface-level corrosion that's already advanced. For railways, underframe corrosion is particularly insidious because it occurs in hard-to-see locations beneath the vehicle. For bus and truck fleets, wheel well corrosion, frame rail deterioration, and body panel perforation follow predictable patterns but are tracked only during periodic roadworthiness tests.

The Proven AI Capability

Defect detection models trained on corrosion patterns at industrial facilities distinguish between surface oxidation (cosmetic), active corrosion (progressing), and structural corrosion (requiring intervention). At a major port, infrastructure corrosion from salt spray is detected on quay walls and bollards via camera monitoring.

How It Works

Depot-entry cameras capture underframe images as vehicles pass over inspection pits. Mobile inspections capture body panels, wheel wells, and structural joints. AI classifies corrosion severity (Grade 1-4 scale), measures affected area, and tracks progression per vehicle over time. Historical trending triggers intervention before structural thresholds are reached — not after.

Applicable Modules

The Reality

Pantographs — the spring-loaded arms that connect electric trains to overhead catenary wires — are among the most wear-intensive and safety-critical components on electric rolling stock. Carbon contact strip wear, horn damage, frame cracking, and spring degradation all require regular monitoring. Traditional inspection requires elevated platforms or specialized lifting equipment to examine the roof — expensive, time-consuming, and creating fall-from-height safety risks.

The Proven AI Capability

Overhead cameras (the same architecture as depot-entry scanning) positioned above the track at the depot entrance capture pantograph images as each train passes through. AI measures carbon strip thickness, detects frame cracks, identifies horn damage, and monitors spring compression. The same sub-millimeter measurement precision proven on brake pads applies to carbon strip wear profiling.

Applicable Modules

The Reality

Most depot maintenance follows fixed schedules — every X kilometers or every Y days, regardless of actual vehicle condition. This means healthy vehicles get serviced too often (wasting labor and parts) while stressed vehicles may be serviced too late. Condition-based maintenance is the industry goal, but it requires continuous data streams that manual inspection can't provide.

The Proven AI Capability

The SAAI platform doesn't just detect defects — it tracks them over time per vehicle. Brake pad wear rate, corrosion progression rate, tire tread depletion rate, and fluid consumption rate per vehicle create individual vehicle health profiles. The Rule Engine uses these profiles to predict when each component will reach its intervention threshold and generates work orders in advance — scheduling the right vehicle for the right maintenance at the right time. At a Bay Area automaker facility, this approach reduced spare parts overuse and loss by 80% — because every maintenance action was tied to a specific vehicle, specific component, and specific condition assessment.

Applicable Modules