Smart Normalization Defining Normal Equipment Behavior
ChronX Smart Normalization defines realistic operating ranges and behavior patterns for industrial equipment by analyzing signal values, operating ranges, and rate of change.
OVERVIEW
Defining Normal Behavior Before Machine Learning
After industrial data has been cleaned, interpolated, and aligned, the next step is to define what normal equipment operation looks like. Machine learning models must understand normal behavior before they can detect abnormal behavior.
Industrial equipment does not operate at a single fixed value. Equipment operates under different loads, speeds, temperatures, and operating conditions. Therefore, normal behavior must be defined as a range of operation and behavior pattern rather than a single value.
ChronX Smart Normalization analyzes historical operational data to define realistic operating ranges and behavior patterns for each signal and equipment.
OPERATING RANGE BEHAVIOR
Understanding Operating Range Behavior
Industrial equipment operates within certain ranges depending on operating conditions, load, and environment. Instead of defining a single normal value, ChronX defines normal operating ranges.
For example:
• Temperature may operate between 60 and 90 degrees
• Pressure may operate between 4 and 6 bar
• Vibration may operate between 1 and 2 mm/s
These ranges change depending on operating conditions. Smart Normalization identifies realistic operating ranges from historical data.
RANGE-BASED NORMALIZATION (IQR METHOD)
Range-Based Normalization
Range-Based Normalization focuses on the range where most data normally operates and ignores extreme outliers.
ChronX uses statistical methods such as Interquartile Range (IQR) to determine the normal operating range of signals. This method focuses on the middle portion of the data distribution and ignores extreme values.
This ensures that rare spikes, sensor errors, or abnormal events do not redefine normal behavior.
Range-based normalization:
• Defines realistic operating ranges
• Ignores extreme outliers
• Improves anomaly detection
• Prevents incorrect model learning
CHANGE-BASED NORMALIZATION (GRADIENT METHOD)
Change-Based Normalization
Change-Based Normalization analyzes how fast signals change over time instead of only analyzing signal values.
Many equipment failures start with instability or rapid changes rather than values exceeding limits. By analyzing the rate of change of signals, ChronX can detect early warning signs of instability.
Change-based normalization:
• Detects rapid changes
• Detects instability patterns
• Detects gradual degradation
• Detects abnormal signal behavior
• Provides early warning signals
BEHAVIOR BASELINE CREATION
Creating Behavior Baselines
After normalization, ChronX creates a behavior baseline that represents normal equipment operation.
This baseline includes:
• Normal operating ranges
• Normal rate of change
• Signal relationships
• Operating patterns
• Load-based behavior patterns
This baseline becomes the reference for anomaly detection and failure prediction.
NORMALIZATION OUTPUT DATASET
Normalized Dataset Output
After Smart Normalization, ChronX produces normalized datasets where signals are scaled, operating ranges are defined, and behavior baselines are created.
These normalized datasets are used for:
• Behavior learning
• Machine learning model training
• Anomaly detection
• Failure prediction
• Operational optimization