Maximo Visual Inspection 9.1.17, Geospatial Linear Assets, and the 3D Inspection Workflow: A 2026 APM Pattern

Visual inspection has historically been the most under-digitized part of asset maintenance. The technician walks the asset, sees the corrosion, fills out a paper form, returns to the truck, types it into Maximo. The cycle is slow, the data quality is variable, and the AI never sees the image because the image never makes it off the phone.

The 2026 release train — MVI 9.1.17 (shipped May 27, 2026) plus the MAS 9.2 spatial and APM changes — closes that loop. Visual inspection is now a first-class condition-monitoring channel: AI-assisted defect classification runs in real time on the device or on the edge, geofencing drives the inspection workflow, 3D asset views give the technician spatial context, and linear-asset inspection routes anomalies into Health scoring and work execution. This article walks through the practical pattern.

The MVI 9.1.17 baseline

The Maximo Visual Inspection component shipped 9.1.17 on May 27, 2026, along with the MVI Edge subcomponent. The release is a security and stability update — it resolves several CVEs, including CVE-2026-33750 (a brace-expansion library issue in the expand() function, CVSS 7.5, that could cause process hang and memory exhaustion). The affected versions were 9.0.0–9.0.19 and 9.1.0–9.1.11, and the fix is in 9.1.17 (for the 9.1 line) and 9.0.20 (for the 9.0 line).

The practical implication: if you are on MVI 9.1.11 or earlier in the 9.1 line, or 9.0.19 or earlier in the 9.0 line, you are exposed to a remotely triggerable denial-of-service vulnerability. The fix is a straightforward component upgrade, no data migration required, and the upgrade can be done as part of the normal feature-channel cadence.

The 9.1.17 release also bundles general code and security fixes across MVI Server and MVI Edge, so the upgrade is worth scheduling even if the CVE were not present.

The four patterns that the 9.2 release train enables

1. AI-assisted defect classification on the device

The 9.2 release tightens the integration between the Maximo Mobile camera pipeline and the MVI inference engine. The pattern:

• The technician opens the inspection form in Mobile, takes a photo of the asset, and the image is sent to MVI for inference.
• MVI returns a defect class and a confidence score. The defect class is matched to the failure hierarchy (the MVI model is trained on a customer's failure taxonomy).
• The technician confirms or overrides the AI's classification. The override is recorded as a feedback signal that can be used to retrain the model.
• The classification lands in the inspection record and propagates to the work order failure code.

The deployment options for the inference engine:

• MVI on the cluster. The MVI model runs in the MAS cluster, in the same namespace as the rest of MVI. The technician's device sends the image to the cluster, the cluster returns the classification. Lowest latency for a connected device, requires connectivity at inspection time.
• MVI Edge on the device. MVI Edge is a smaller model that runs on a laptop, a ruggedized tablet, or an industrial gateway. The technician's device sends the image to the local edge, the edge returns the classification, and the result is cached for later sync. This is the right pattern for sites with intermittent connectivity (a remote substation, a mine pit, a rail corridor).
• MVI Edge on a gateway. A site-level gateway (NVIDIA Jetson, Intel NUC, or a rugged industrial PC) runs the MVI Edge model, and the technicians on the site connect to it. This is the right pattern for a plant with multiple technicians and limited network reach to the central MAS cluster.

The 2026 reality is that the inference is fast enough — typically under 2 seconds per image on MVI Edge — that the technician does not experience it as a delay. The model retraining loop (technician overrides feeding back into a periodic retraining) is the part that closes the AI quality loop over time.

2. Geofenced asset inspection

The 9.2 release turned geofencing into a first-class feature for inspection workflows. The pattern:

• A geofence is defined around an asset (a substation perimeter, a transformer yard, a tank farm) or a linear asset (a pipeline segment, a rail line, a conveyor route).
• The Mobile app, with the technician's consent, monitors the device's location against the geofence. When the device enters the geofence, the Mobile app surfaces the relevant inspection forms ("you have 3 inspections due at this substation").
• The technician completes the inspections, the data is recorded locally on the device, and the results are synced to Manage when the device exits the geofence or regains connectivity.

The geofenced inspection pattern is the one that closes the gap between "I am supposed to do inspections every 30 days" and "I am physically at the asset." The compliance metric for the inspection program shifts from "WO created and closed" to "technician geofenced at asset + inspection record with photo evidence + AI-classified defect status." That is a more defensible compliance story in a regulated industry.

3. 3D asset visualization

The 9.2 release added 3D asset visualization as a first-class spatial capability. The pattern:

• An asset or asset location is associated with a 3D model (a point cloud, a CAD model, a photogrammetry mesh, or a simplified geometry).
• The 3D model is rendered in the Maximo Spatial or Mobile view, with the asset's hierarchy overlaid (child assets as labeled points, attached documents as icons, sensor data as color-coded markers).
• The technician can navigate the 3D view, click on a child asset to see its record, click on a sensor to see its current reading, click on a defect icon to see the inspection history.

The 3D pattern is the right answer for complex assets where a 2D plan view or a text list does not give the technician enough spatial context — a substation, a refinery unit, a process plant, a large engine room, an offshore platform. The 3D model is typically produced once (a laser scan or photogrammetry capture) and reused for every inspection, every work order, and every reliability review. The model becomes a shared spatial reference for the entire asset team.

4. Linear-asset inspection with anomaly routing

The 9.2 release tightened the linear asset (pipeline, rail, conveyor, road, transmission line) inspection pattern. The model:

• A linear asset is modeled as a series of segments, each with a starting and ending chainage, a centerline geometry, and a set of child features.
• The technician walks or drives the linear asset, recording observations against the chainage ("at chainage 12.4 km, observed corrosion on the north-side pipe support").
• The observation is geo-referenced (GPS or chainage) and the chainage-to-location mapping is recorded.
• The observation is routed to the asset's health scoring (the failure class contributes to the asset's condition score) and to the work order system if the observation is actionable.

The anomaly routing pattern is the one that closes the loop between linear-asset inspection and the rest of the reliability stack. The observation is no longer a free-form note in an inspection form. It is a structured record with a failure class, a geo-reference, a chainage, a photo, and an AI-classified defect status, and it lands in Health, Predict, and Work Order Tracking the same way a sensor-detected anomaly would.

The configuration pattern that holds it together

The configuration surfaces for the 2026 visual inspection stack are:

• MVI Server and MVI Edge — the inference engine. Configured as part of the MAS install, with model selection (prebuilt vs. custom) and deployment topology (cluster vs. edge).
• Inspection forms — the question/answer forms, the photo capture workflow, the AI-classification override workflow, the failure class mapping. Lives in Manage, syncs to Mobile.
• Geofence definitions — the polygon definitions, the geofence-to-asset mapping, the device consent policy.
• 3D asset models — the model files (typically OBJ, glTF, or similar), the asset-to-model mapping, the spatial reference (the model's coordinate system and its alignment to real-world coordinates).
• Linear asset definitions — the segment definitions, the chainage mapping, the centerline geometry.
• Health scoring configuration — the asset scoring group, the contributors, the failure class-to-contributor mapping.

The right way to ship this is staged. The minimum viable visual inspection stack is MVI + the inspection form + the camera-to-classification pipeline. The next stage is the geofence-driven inspection workflow. Then the 3D model. Then the linear-asset pattern. Each stage is a separate change, with its own training, its own adoption metric, and its own rollback path.

The Health and Predict handoff

The reason the 2026 visual inspection stack matters for reliability engineers specifically is the handoff to Maximo Health and Maximo Predict.

• Health uses the inspection data as a contributor to the asset's health score. The AI-classified defect status becomes a weighted contributor in the scoring group, alongside the existing contributors (sensor data, maintenance cost, age, remaining useful life).
• Predict uses the inspection data to update the asset's failure prediction model. The historical inspection records become training data for the Predict model, and the model's prediction is updated with each new inspection.

The handoff is what closes the loop between "we took a photo" and "the asset's predicted failure window has moved." A utility with a transformer fleet that is doing AI-classified visual inspections quarterly will see the Predict model's confidence improve with each inspection cycle, and the prediction window will tighten from "12–18 months" to "8–11 months" over 18 months of data. That is the metric that justifies the program.

The 2026 release train is the first time the visual inspection stack has been a real reliability input, not a documentation output. The teams that recognize this and build the inspection program with Health and Predict in mind — not as a standalone image-capture workflow — are the ones that will get the reliability return.