Process historians store years of sensor trends — but without a maintenance layer interpreting that data, anomalies go unnoticed until a breakdown confirms what the trend line had been showing for weeks. The data model for historian AI integration in work order triage bridges that gap: structured time-series analysis feeds directly into OxMaint, so every statistically significant deviation from baseline triggers a properly contextualized work order before the equipment fails. This page covers the data schema, anomaly detection logic, and the CMMS handoff that turns a historian signal into a scheduled maintenance task. Teams using this integration reduce unplanned downtime events attributable to trend-based failures by an average of 55%. To connect your historian to OxMaint, start a free trial or book a 30-minute historian integration session with a reliability engineer.
Historian Integration · Predictive Maintenance
Data Model for Historian AI Integration in Work Order Triage
Time-series trends in your historian are the earliest warning signs your equipment sends. OxMaint reads them, classifies them, and acts on them automatically.
The Core Data Schema
Every historian data point that enters OxMaint's triage engine is mapped to a five-field normalized object before anomaly scoring begins.
asset_id
string
OxMaint asset record ID linking this tag to the physical equipment, location, and maintenance history.
tag_name
string
Historian tag identifier. Mapped to a human-readable measurement label during integration setup.
timestamp
ISO 8601
UTC timestamp of the historian sample. Used as the reference point for trend window calculation.
value
float
Engineering unit value of the measurement. Normalized to SI units during ingestion where applicable.
anomaly_score
float 0–1
AI-computed deviation score. Values above the configured threshold (default 0.75) trigger work order creation.
trend_direction
enum
Indicates whether the value is rising, falling, or oscillating relative to the 30-day rolling baseline.
55%
Average reduction in trend-based failure events after historian AI integration
14 days
Average lead time from historian anomaly detection to physical failure — time now used for planned repair
5 min
Latency from historian data ingestion to work order creation in OxMaint triage engine
Supported Historian Platforms
| Historian Platform |
Connection Method |
Data Query Support |
Backfill Available |
| OSIsoft PI (AVEVA) | PI Web API / REST | Raw, interpolated, summary | Yes — up to 2 years |
| GE Proficy Historian | REST API / OPC-HDA | Raw, compressed | Yes |
| Honeywell Uniformance PHD | OPC-HDA, REST | Raw, interpolated | Yes |
| Aspentech IP.21 | REST API | Raw, aggregated | Yes |
| Ignition Historian | JDBC / REST | Raw, aggregated | Yes |
| Custom / SQL-based | ODBC / REST API | Raw query | Yes — with schema mapping |
Your Historian Is Already Predicting Failures. Let OxMaint Act on Them.
Connect your PI, Proficy, or custom historian to OxMaint in under a week. Predictive work orders start generating from day one of data ingestion.
Anomaly Detection Methods
01
Statistical Baseline
Rolling 30-day mean and standard deviation. Values beyond 3-sigma from baseline trigger predictive work orders. Suitable for stable continuous processes.
02
Rate-of-Change
Detects abnormally rapid increases or decreases even when absolute values remain within normal range. Critical for bearing temperature and vibration trends.
03
Pattern Matching
Compares current trend shapes against historical failure signatures stored in OxMaint's library. Identifies pre-failure patterns weeks before threshold breach.
Expert Review
Reviewed by a Process Reliability & Condition Monitoring Specialist
The value of historian integration isn't in the data — it's in what happens next. Most organizations have years of trend data sitting in a PI or Proficy instance that nobody in maintenance can query. The data model that matters most is the one that normalizes historian tags into an asset-linked schema before anomaly scoring, so the resulting work order arrives with equipment context rather than just a raw tag name and a number. That context is what allows a technician to act on a predictive alert instead of dismissing it as a false alarm.
Frequently Asked Questions
How does OxMaint avoid generating false positive work orders from historian data?
OxMaint uses a three-layer validation process before creating a historian-triggered work order: the anomaly score must exceed the configured threshold, the condition must persist for a minimum duration window (configurable from 5 to 60 minutes), and the asset must not have a recent work order for the same tag. This eliminates transient spikes and known maintenance windows from triggering unnecessary work orders.
Start a free trial to configure these thresholds for your process.
Can OxMaint use historical failure data to train its anomaly models?
Yes. During onboarding, OxMaint can backfill up to two years of historian data and cross-reference known failure events from your maintenance records to build asset-specific baseline models. This means anomaly detection starts with context about your actual failure patterns rather than industry defaults.
Book a demo to see how backfill training improves day-one detection accuracy.
What happens when a historian tag loses signal or goes offline?
OxMaint monitors data freshness for every integrated historian tag. If a tag stops reporting within its expected polling interval, OxMaint creates a data quality alert and can optionally create a work order for the sensor or historian connection itself. This prevents silent failures where a missing reading is mistaken for a healthy flat-line. Tag health dashboards show last-received timestamp and data quality scores across your entire historian integration.
Is historian data retained within OxMaint or only read for triage?
OxMaint retains only the anomaly event records and their associated context — not the full raw time-series stream, which remains in your historian. This keeps OxMaint's data footprint minimal while preserving the full historian as the system of record for raw process data. Anomaly event records are retained for the life of the work order and linked to the asset maintenance history permanently.
14 Days of Warning Is Already in Your Historian
Connect OxMaint and start acting on the trends your historian has been tracking all along — before they become breakdowns.
