OutlierX360

outlierX360 is an ML-powered outlier detection and diagnostic solution that continuously monitors plant sensor data to identify abnormal process behavior, subtle process drifts, and potential equipment issues before conventional alarms are triggered.

Using existing plant data, outlierX360 continuously evaluates critical process variables and distinguishes between instrument drift, genuine process or equipment faults, known disturbances, and expected baseline shifts. By providing severity context, reason codes, correlated tags, and expected values for detected events, it helps engineering teams focus on genuine abnormalities and take proactive action.

Salient Features

Multi-Layer Anomaly Detection

Evaluate every monitored tag using eight independent detection checks covering abnormal levels, changing operating patterns, extreme readings, sudden jumps, peer deviations, slow drift, new sustained levels, and drift from initial performance.

Intelligent Fault Diagnosis

Classify detected abnormalities as process issues, instrument or tag issues, equipment faults, or known disturbances, helping engineers distinguish genuine operational problems from misleading signals.

Early Detection & Process Drift Monitoring

Identify developing failures, subtle process drifts, leaks, and changing operating behavior earlier, providing teams with additional time to investigate and plan preventive action.

Continuous & Configurable Monitoring

Monitor critical variables 24×7 using historian connectivity or perform on-demand analysis using spreadsheet data, with configurable tag limits, directions, criticality, and known process disturbances.

Customizable Limits, Priorities, etc.

a. Engineers can manipulate configured KPI limits based on plant priorities.

Sensitivity Analysis

Optimax provides feature to run multiple cases simultaneously with different limit ranges to enhance the actionable insights.

Role-Based Access and Security

Provides access control to ensure only authorized personnel can view or edit monitoring configurations.

Result Analysis Tool

Offers diagnostic tools to help engineers quickly investigate and resolve performance issues with different configurations

Client Benefits

10× Faster Fault Identification

Differentiate equipment faults from sensor drift and other process conditions to accelerate investigation and fault identification.

<2% False Positives

Isolate genuine abnormalities and reduce unnecessary alerts, helping minimize alarm fatigue.

80% Less Time Chasing Data

Replace repetitive manual trend investigation with a ranked list of diagnosed findings and supporting context.

Enable Proactive Maintenance Planning

Detect abnormal behavior earlier and provide evidence-based findings that help teams investigate equipment and process issues before they escalate.

Reduce Production Loss Risk

Provide an earlier detection window for developing issues, giving operations and engineering teams more time to take preventive action and protect production.

Dashboard & Snapshots

Configure all tags related to specific Equipment:

Solution configuration [Defining the Objective: Maximize Production, Minimize Utility]:

Solution configuration [Defining cases of interest – Multiple cases can also be added]:

Result Tag Configuration: Identify the potential actionable insights.

Sensitivity Analysis: Supporting engineers to evaluate appropriate configurations

Case Studies

Development of Soft Sensor to predict the C5 contents in Debutaniser column Overhead:

Background

In modern petrochemical operations, maintaining product specifications within tight limits is critical for both quality assurance and process efficiency. One of the key challenges faced by a leading chemical plant was the need for real-time prediction and monitoring of C5 content in the overhead stream of the debutanizer column — a crucial step to ensure product specification compliance and optimize operations.

Challenge

Traditional laboratory testing methods for measuring C5 content introduced significant time delays, making real-time adjustments difficult. The plant required a fast, accurate, and reliable solution to continuously monitor C5 levels, allowing operators to take timely corrective actions and avoid off-spec products.

Solution with AnalyticX

Using AnalyticX, the engineering team was able to develop a soft sensor — a machine learning-based predictive model — specifically designed to estimate the C5 content in the debutanizer overhead stream in real time.

Key steps included:

Seamless ingestion of historical operational data into AnalyticX without any coding requirements.
Applying statistical analysis to identify key process variables influencing C5 concentrations.
Rapid development and training of a machine learning model tailored to predict C5 content with high accuracy.
Deployment of the predictive model into live operations for real-time product spec monitoring.

Results

Enhanced Product Monitoring:
Operators now have a real-time view of the predicted C5 content, significantly reducing reliance on delayed lab measurements.
Improved Process Control:
By closely tracking C5 levels, the plant was able to make timely adjustments to operational parameters, ensuring consistent product quality.
Reduced Off-Spec Production:
Real-time insights helped minimize off-spec batches, leading to material savings and reduced reprocessing costs.

Key Takeaways

AnalyticX enabled the rapid development of a soft sensor customized for critical KPI monitoring.

In Stock