AI for Vibration Analysis in IBM Maximo

The traditional vibration analysis workflow in Maximo is a manual, multi-step process: a technician collects spectrum data, an analyst reviews the charts in a separate tool, diagnoses potential faults, and then manually creates a work order or notification in Maximo. AI integration stitches these siloed steps together. It acts as a virtual analyst that continuously monitors incoming vibration data—whether from periodic routes or online sensors—applies fault detection models, and automatically creates aligned records in Maximo. This connects the Condition Monitoring module's data directly to the Work Order Tracking and Asset Health modules, closing the loop from detection to action.

A production implementation typically involves an event-driven pipeline: vibration data files (e.g., .csv, .dsp) land in a cloud storage bucket or are pushed via an API. An AI service processes the spectra, running models trained on historical failure patterns for specific asset types (e.g., pumps, motors, fans). For each detection—such as imbalance, misalignment, or bearing wear—the system creates a Maximo MBO via its REST API. It generates a detailed MXWO (Work Order) with the fault diagnosis, severity, recommended actions, and links to the source spectrum, or it may create a MXPR (Problem) record for further review. This automation surfaces issues in hours instead of days, allowing planners to prioritize corrective work before a minor fault leads to downtime.

Governance and rollout are critical. Initial deployments often start in monitor-only mode, where AI generates MXALERT records instead of work orders, allowing reliability engineers to validate predictions against their expertise. Prompts and model confidence scores are logged in a MXWPLOG for auditability. Over time, as trust is built, workflows can be configured to auto-create low-severity PMs or route high-confidence, critical detections directly to the work order pool. This phased approach ensures the AI augments—rather than replaces—domain expertise, integrating seamlessly into the existing Maximo change management and role-based security (MXROLE) framework your team already uses.

The integration is built on a secure, event-driven pipeline. Vibration data collectors (portable analyzers or online systems) push spectrum files (e.g., .csv, .tdms) to a cloud storage bucket (AWS S3, Azure Blob). A listener service triggers an AI inference workflow that extracts features and runs against pre-trained models for common faults—imbalance, misalignment, bearing defects, looseness. The service enriches the raw prediction with severity scoring, confidence levels, and recommended corrective actions, referencing a knowledge base of standard maintenance procedures.

The enriched fault detection payload is then posted to IBM Maximo via its REST API (mxapisci/mxassetci). The system creates or updates a ASSET record, logs a new FAILUREREPORT linked to the asset, and—critically—automatically generates a WORKORDER. This work order is pre-populated with the fault details, the recommended repair procedure, required craft skills (e.g., 'Alignment Technician'), and links to the original spectrum file stored as a DOCLINKS attachment. The work order status is set to APPR (approved) or WAPPR (waiting approval) based on a configurable business rule, often tied to the fault's severity score.

For governance, every step is logged. The AI service writes an audit trail of the inference input, model version, and output to a dedicated log store. Maximo's native AUDITLOGS capture the record creation. This traceability is crucial for reliability engineers validating AI recommendations and for continuous model improvement. Rollout typically starts with a pilot asset group, where AI-generated work orders are flagged for human-in-the-loop review before auto-approval, allowing teams to build confidence in the system's accuracy before scaling to the entire rotating equipment fleet.

A production integration typically uses a secure middleware layer (like an API gateway or event bus) to broker communication between Maximo and external AI services. Vibration spectrum files from Maximo's Document Management module or attached to Meter Readings are securely transmitted. AI model outputs—such as fault classifications (e.g., imbalance, misalignment, bearing wear) and confidence scores—are written back to Maximo as new Failure Codes, Condition Monitoring Alerts, or directly into custom attributes on the Asset or Meter record. All data flows are logged, and access is controlled via Maximo's existing Role-Based Access Control (RBAC) and integration user security groups.

Governance is critical. We recommend establishing a human-in-the-loop review step for the first 90 days, where AI-generated work order suggestions are placed in a "Proposed" status in Maximo's Work Order Tracking module for reliability engineer approval. This creates an audit trail and allows for model validation. Furthermore, the integration should be designed to expose key metrics—like model confidence, false positive rates, and work order closure rates—back into Maximo or a separate dashboard, enabling continuous monitoring of the AI's performance and business impact.

A phased rollout mitigates risk and builds organizational trust. Phase 1 (Pilot): Target a single, non-critical asset class (e.g., HVAC fans). Integrate AI detection for 2-3 common fault modes and manually review all outputs. Phase 2 (Expansion): Scale to a critical production line, automate work order creation for high-confidence predictions, and integrate with Maximo's Job Plans and Preventive Maintenance schedules. Phase 3 (Optimization): Implement feedback loops where work order resolution data from Maximo is used to retrain and improve the AI models, closing the reliability loop. This approach ensures the technology delivers value incrementally while maintaining the integrity and security of your core EAM system.