> ## Documentation Index
> Fetch the complete documentation index at: https://teamrobomanipal-3a657712.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Predictive ML Model

> Technical overview of the tinyML (rule-based ML) Model

This project provides a **Rule-Based Diagnostic System** to calculate the health percentage of an industrial actuator based on real-time sensor data. It evaluates electrical and mechanical stress factors to provide a predictive maintenance score.

## Overview of Metrics

The calculator uses three primary sensor inputs to determine the degradation of the hardware:

1. **Current (A):** Measures electrical load and resistance.
2. **Voltage (V):** Monitors power stability and supply quality.
3. **Vibration (g):** Detects mechanical wear, bearing failure, or misalignment.

***

## Logic & Ruleset

The health score starts at **100.0%** and is reduced based on the following logic:

| Rule                  | Trigger                              | Penalty                  |
| :-------------------- | :----------------------------------- | :----------------------- |
| **Current Increase**  | Every 1.0A above baseline            | -3.0%                    |
| **Voltage Deviation** | Deviation from target (e.g., 220V)   | Proportional (up to 20%) |
| **Vibration**         | Every 1.0g above baseline            | -8.0%                    |
| **Trend Analysis**    | Rapid spikes in Current or Vibration | -5.0% each               |
| **Cross-Correlation** | High Current **AND** High Vibration  | -15.0% (Severe Alert)    |
| **Efficiency**        | Power consumption > 120% of expected | -8.0%                    |

***

## Implementation

```python theme={null}
def calculate_health(current, voltage, vibration, 
                     baseline_current=5.0, 
                     baseline_voltage=220.0, 
                     baseline_vibration=0.5,
                     current_trend=0.0,
                     vibration_trend=0.0):
    
    Calculates health percentage (0-100) based on sensor deviations.
    
    health = 100.0
    
    # 1. Current penalty
    current_increase = max(0, current - baseline_current)
    health -= current_increase * 3.0 
    
    # 2. Voltage deviation penalty
    voltage_deviation = abs(voltage - baseline_voltage)
    health -= (voltage_deviation / baseline_voltage) * 20.0
    
    # 3. Vibration penalty
    vibration_increase = max(0, vibration - baseline_vibration)
    health -= vibration_increase * 8.0 
    
    # 4. Trend penalties
    if current_trend > 0.05: health -= 5.0
    if vibration_trend > 0.02: health -= 5.0
    
    # 5. Cross-sensor correlation (Severe State)
    current_severe = (current - baseline_current) > (0.2 * baseline_current)
    vibration_severe = (vibration - baseline_vibration) > (0.5 * baseline_vibration)
    if current_severe and vibration_severe:
        health -= 15.0
    
    # 6. Power efficiency check
    power = voltage * current
    expected_power = baseline_voltage * baseline_current
    if power > expected_power * 1.2:
        health -= 8.0
    
    return max(0.0, min(100.0, health))


```

<a href="health_model.py" download>
  Download tinyML Python Script
</a>