mIQ Anomaly
With real-time anomaly detection and LLM-powered data science, mIQ Anomaly optimizes your production, maintenance, quality, and energy processes, reduces losses, and maximizes performance.
Anomaly Detection Use Cases
-
General Anomaly Detection: Continuously monitors process parameters to distinguish between normal and abnormal conditions; can also detect unknown (novel) anomaly types. “Catch deviations instantly, prevent loss before it occurs.”
-
Predictive Maintenance: Predicts failures by analyzing IoT, PLC, and operator data, optimizing the maintenance plan. “See unplanned downtime in advance, turn it into planned maintenance.”
-
Predictive Quality: Predicts quality issues using historical and real-time data, automatically adjusts process parameters. “Prevent scrap before it's produced.”
-
Energy Anomalies: Monitors energy consumption profiles, flags unexpected increases, leaks, and off-shift consumption. “Reduce energy costs and lower your carbon footprint.”
LLM-Powered Data Science Lifecycle (Optional)
-
Data Source Review & Project Proposal: Analyzes existing data and proposes feasible anomaly detection projects.
-
Data Preparation & Feature Engineering: Manages missing and outlier values, prepares the dataset for model training with feature suggestions.
-
Algorithm Proposal & Code Generation (Python): Selects appropriate algorithms, generates training code; users can intervene if they wish.
-
Model Testing & Reporting + XAI: Creates test reports; presents results with explainable AI views that show which variables influenced the outcomes. “Make ML expertise accessible to everyone.”
Data Management and Integration
-
Extensive Data Access: Rich data sources including products, stations, work orders, quality/OEE, maintenance/downtime, planning, energy, etc.
-
ETL + Function Call: Data extraction using filters; adding calculated values returned from the REST function to the data set.
-
Data Quality & Preprocessing: Fill missing values with FF/FB/average; eliminate outliers; quick discovery with visual statistics.
-
Source Limitation: Safely limit the data set size based on the RAM/Disk ratio. “Your data is ready, your model is ready in minutes.”
Model Distribution and Management
-
Microservice-Based Backend: Model services with a scalable, flexible architecture.
-
One-Click Deployment & Monitoring: Deploy models that pass the success threshold, monitor their performance; stop/restart.
-
Input/Output Mapping: Select model inputs (tag/REST) in production; specify the tag/channel where results will be written.
-
MES Alarm/Rule Integration: The alarm/rule engine automatically initiates actions based on model output.
-
Project/Model History: Data sources, versions, training/testing processes are recorded. “End-to-end traceability from model to impact.”
Expected Business Outcomes (Sample KPIs)
Reduction in unplanned downtime, increase in OEE, decrease in maintenance/energy/scrap costs, reduction in detection–intervention time. (Factory-specific targets are set after installation.)