Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/BhaveshBytess/PREDICTIVE-MAINTENANCE/llms.txt

Use this file to discover all available pages before exploring further.

​
Overview

The What-If Analysis endpoint allows you to test hypothetical sensor scenarios without affecting the production system. Input manual sensor values to see predicted health scores, risk levels, and detailed feature contributions.
This is a read-only simulation endpoint. It does not affect production data, trigger alerts, or modify system state. Use it to explore “what-if” scenarios and validate risk thresholds.

​
Endpoint

POST /sandbox/predict

​
Request Body

​
voltage_v
number
required
Voltage in Volts. Must be between 150V and 300V.Example: 230.0 (Indian Grid nominal)
​
current_a
number
required
Current in Amperes. Must be between 1A and 50A.Example: 15.0 (normal operating current)
​
power_factor
number
required
Power Factor (dimensionless). Must be between 0.3 and 1.0.Example: 0.92 (good efficiency)
​
vibration_g
number
required
Vibration in g-force. Must be between 0.01g and 10.0g.Example: 0.15 (normal vibration level)
​
asset_id
string
default:"asset-001"
Asset identifier to compare against. If this asset has live data, the response will include comparison metrics.Example: "asset-001"

​
Response

​
anomaly_score
number
Anomaly confidence score between 0 and 1. Higher values indicate more anomalous behavior.Example: 0.345
​
health_score
number
Overall health score from 0 to 100. Derived from 100 × (1 - anomaly_score).Example: 65
​
risk_level
string
Risk classification based on health score:
  • LOW: health_score ≥ 75
  • MODERATE: health_score ≥ 50
  • HIGH: health_score ≥ 25
  • CRITICAL: health_score < 25
Example: "MODERATE"
​
feature_contributions
array
Array of feature contributions explaining which sensor values contributed most to the anomaly score.
​
insight
string
Human-readable explanation of the prediction based on top contributing factors.Example: "Vibration (2.50g) is significantly elevated. Current draw (35.0A) indicates potential overload."
​
comparison
object
Comparison with live system state (only present if asset has recent data).

​
Example Request

curl -X POST https://api.example.com/sandbox/predict \
  -H "Content-Type: application/json" \
  -d '{
    "voltage_v": 210.0,
    "current_a": 35.0,
    "power_factor": 0.55,
    "vibration_g": 2.5,
    "asset_id": "asset-001"
  }'

​
Example Response

{
  "anomaly_score": 0.782,
  "health_score": 22,
  "risk_level": "CRITICAL",
  "feature_contributions": [
    {
      "feature": "Vibration",
      "value": 2.5,
      "contribution_percent": 38.4,
      "deviation_from_normal": 4.12,
      "status": "critical"
    },
    {
      "feature": "Power Factor",
      "value": 0.55,
      "contribution_percent": 28.7,
      "deviation_from_normal": 3.45,
      "status": "critical"
    },
    {
      "feature": "Current Spikes",
      "value": 10,
      "contribution_percent": 19.2,
      "deviation_from_normal": 2.78,
      "status": "critical"
    },
    {
      "feature": "Voltage Stability",
      "value": 20.0,
      "contribution_percent": 8.9,
      "deviation_from_normal": 1.23,
      "status": "elevated"
    },
    {
      "feature": "Voltage",
      "value": 210.0,
      "contribution_percent": 3.1,
      "deviation_from_normal": 0.87,
      "status": "normal"
    },
    {
      "feature": "Power/Vibration Ratio",
      "value": 4.545,
      "contribution_percent": 1.7,
      "deviation_from_normal": 0.45,
      "status": "normal"
    }
  ],
  "insight": "Vibration (2.50g) is significantly elevated. Current draw (35.0A) indicates potential overload.",
  "comparison": {
    "live_voltage": 228.0,
    "live_current": 15.2,
    "live_power_factor": 0.91,
    "live_vibration": 0.16,
    "voltage_diff_percent": -7.9,
    "current_diff_percent": 130.3,
    "power_factor_diff_percent": -39.6,
    "vibration_diff_percent": 1462.5,
    "live_health_score": 87,
    "live_risk_level": "LOW"
  }
}

​
Use Cases

​
Testing Fault Scenarios

Simulate known fault conditions (motor stall, bearing failure, voltage spike) to validate that the ML model correctly identifies them: 
# Motor Stall Scenario
curl -X POST https://api.example.com/sandbox/predict \
  -H "Content-Type: application/json" \
  -d '{
    "voltage_v": 210.0,
    "current_a": 35.0,
    "power_factor": 0.55,
    "vibration_g": 2.5,
    "asset_id": "asset-001"
  }'

​
Validating Thresholds

Test edge cases to determine at what point the system transitions between risk levels: 
# Test moderate vibration
curl -X POST https://api.example.com/sandbox/predict \
  -H "Content-Type: application/json" \
  -d '{
    "voltage_v": 228.0,
    "current_a": 16.0,
    "power_factor": 0.88,
    "vibration_g": 1.5,
    "asset_id": "asset-001"
  }'

​
Comparing to Live State

When providing an asset_id with live data, the response includes comparison metrics showing how your hypothetical scenario differs from current reality.

​
Preset Scenarios

Use the GET /sandbox/presets endpoint to retrieve pre-configured fault scenarios:
  • Normal: Healthy operating conditions
  • Motor Stall: High current draw with poor power factor
  • Voltage Spike: Grid voltage spike
  • Bearing Failure: Excessive vibration

​
Error Responses

​
400 Bad Request - No Trained Detector

{
  "detail": "No trained detector for asset 'asset-001'. Please calibrate the system first."
}
The system requires calibration with baseline data before running What-If analysis. Ensure the asset has been trained via the /integration/calibrate endpoint.

​
422 Unprocessable Entity - Validation Error

{
  "detail": [
    {
      "loc": ["body", "voltage_v"],
      "msg": "Input should be greater than or equal to 150",
      "type": "greater_than_equal"
    }
  ]
}
Sensor values must be within the allowed ranges specified in the request parameters.

Build docs developers (and LLMs) love

Get started for freeTalk to us