System Reference

Dashboard (Fleet Command Center)

The primary control room view providing a real-time operational overview of all monitored assets. It features a macro-summary of the fleet's current state (Healthy/Warning/Critical ratios), a triage list ranking machines by imminent risk, and a chronological Anomaly Timeline for situational awareness.

Fleet Intelligence

A strategic analytics module designed for Plant Managers. It displays fleet-wide health distributions via histogram, systemic risk levels, and aggregate performance metrics (like total anomalies per 24h/7d/30d) across hundreds of connected machines to influence maintenance scheduling.

Analysis (Engineering Module)

A deep-dive, asset-specific signal processing suite. Designed for reliability engineers, it allows the interrogation of raw time-series vibration telemetry alongside frequency-domain (FFT) charts. It extracts both statistical (Mean, Max) and kinetic (RMS) features in real-time.

Datasets (Ingestion Pipeline)

The data ingestion interface. It allows operators to manually or automatically upload raw CSV/JSON telemetry from edge sensors (high-frequency accelerometers). Data uploaded here is automatically processed through the FFT engine and scored by the ML architecture.

ML Intelligence

A transparent view into the underlying AI models (e.g., GEMMA-3B). It exposes model performance metrics, confidence scores, feature importance (explaining which specific data points—like RMS or a specific frequency spike—drove a diagnostic decision), and historical inference logs.

RMS Energy (Root Mean Square)

A fundamental kinetic feature representing the overall continuous vibration energy of a machine. It is the most common metric for tracking general machine health. A baseline-drifting RMS often precedes structural or mounting failures.

Peak Count

The number of times a vibration spike exceeds a dynamically calculated threshold over a specific time window. High peak counts strictly indicate shocking, grinding, or impact events, such as a broken gear tooth or severe bearing spalling.

Dominant Frequency (Dom. Freq)

The specific frequency (measured in Hertz) containing the highest amplitude or energy in the spectrum. Significant shifts in the dominant frequency indicate changes in operating speed, imbalance, or emerging resonance issues.

Spectral Centroid

The 'center of mass' of the frequency spectrum. It indicates where the majority of the vibration energy is concentrated. A shifting centroid often indicates a developing fault traversing from low-frequency (imbalance) to high-frequency (bearing wear).

Spectral Energy

The total energy contained within the frequency spectrum, calculated as the sum of squared magnitudes. Sudden spikes in spectral energy across all bands often indicate catastrophic failure or severe cavitation.

Health Score (0-100)

An aggregate, human-readable metric determined by the ML model indicating the operational health of a machine. 80-100 represents Nominal (healthy) operation, 50-79 represents a Warning state (schedule maintenance), and <50 represents a Critical state (immediate intervention required).

Confidence Score (%)

The statistical certainty of the AI's diagnostic prediction. Higher confidence (e.g., 95%+) means the current signal pattern strongly matches known fault or healthy signatures in the training data. Low confidence triggers manual engineering review.

Standard Deviation (Std Dev)

A statistical measure of the amount of variation or dispersion of a set of vibration values. A high standard deviation relative to the mean indicates erratic, unstable machine operation.

Vibration Time-Series Chart

A kinetic chart plotting raw sensor displacement (amplitude on the Y-axis) against time (X-axis) at high sampling rates (e.g., kHz). It is excellent for catching sudden, short-lived impact events in the time domain before they affect the frequency domain.

FFT Spectrum Chart (Fast Fourier Transform)

A complex mathematical visualization that converts Time-Series data into the Frequency Domain. It reveals the distinct 'harmonics' of a machine. For example, a healthy motor might show a single clean spike at a 50Hz fundamental, while a failing motor will show messy 'sideband' spikes at 120Hz or 240Hz.

Fleet Health Radar / Distribution

A histogram categorizing the entire factory floor into health brackets (0-20, 21-40, 41-60, 61-80, 81-100). It allows plant management to instantly monitor if the factory ecosystem is gradually skewing towards a critical state over time.

Health Trend Line (Historical)

A continuous historical plot of machine health scores over 30+ days. This graph is critical for predictive maintenance, allowing engineers to extrapolate the curve and predict exactly when a machine will cross the failure threshold.