Failure Detection System (FDS) in Railway Infrastructure

What is a Failure Detection System?

A Failure Detection System (FDS) is a monitoring platform that continuously analyses the health of railway signalling and infrastructure equipment. Unlike event loggers that simply record what happened, an FDS actively identifies developing faults, predicts failures, and alerts maintenance teams — often before any service impact occurs.

FDS vs Traditional Maintenance

| Approach | Method | Drawback | |----------|--------|----------| | Reactive | Fix after failure | Service disruption, delays | | Preventive | Scheduled maintenance (time-based) | Over/under maintenance, wasteful | | Condition-Based (FDS) | Monitor and maintain when needed | Requires monitoring infrastructure | | Predictive (Advanced FDS) | AI/ML predicts failure timing | Requires historical data and models |

What Does an FDS Monitor?

Point Machines

• Operating current waveform analysis
• Operation time trending
• Detection circuit health
• Number of operations (cycle counting)
• Motor insulation resistance

Track Circuits

• Rail-to-rail voltage levels
• Receiver current levels
• Tuning unit health
• Insulation resistance (bonded track circuits)
• Right-side and wrong-side failure detection

Signals

• Lamp current monitoring (filament or LED)
• LED degradation tracking
• Power supply voltage
• Signal aspect sequencing verification

Power Supplies

• Input voltage and frequency
• Battery voltage and capacity
• Charger output current
• Changeover switch status
• UPS health and load

Level Crossings

• Barrier motor current
• Barrier position sensors
• Road traffic light status
• Audible warning device
• CCTV system health

FDS Architecture

┌─────────────────────────────────────────────────────────────────┐
│                       FIELD LEVEL                                │
│                                                                  │
│  ┌──────────┐ ┌──────────┐ ┌──────────┐ ┌──────────────────┐   │
│  │ Current   │ │ Voltage  │ │ Digital  │ │ Environmental    │   │
│  │ Sensors   │ │ Sensors  │ │ I/O      │ │ Sensors (Temp,   │   │
│  │ (CT)      │ │ (VT)     │ │ Module   │ │  Humidity)       │   │
│  └─────┬────┘ └─────┬────┘ └─────┬────┘ └────────┬─────────┘   │
│        └─────────────┴───────────┴────────────────┘              │
│                              │                                    │
├──────────────────────────────┼────────────────────────────────────┤
│                       LOCAL PROCESSING                            │
│                              ▼                                    │
│  ┌──────────────────────────────────────────────────────────┐    │
│  │              FDS Local Unit / RTU                         │    │
│  │  - ADC acquisition (multi-channel)                        │    │
│  │  - Local waveform analysis                                │    │
│  │  - Threshold checking                                     │    │
│  │  - Data compression & buffering                           │    │
│  │  - Communication management                               │    │
│  └──────────────────────────┬───────────────────────────────┘    │
│                              │                                    │
├──────────────────────────────┼────────────────────────────────────┤
│                       COMMUNICATION                               │
│                              │                                    │
│                 [Ethernet / 4G LTE / Fibre]                      │
│                              │                                    │
├──────────────────────────────┼────────────────────────────────────┤
│                       CENTRAL SYSTEM                              │
│                              ▼                                    │
│  ┌──────────────────────────────────────────────────────────┐    │
│  │              FDS Central Server                           │    │
│  │  ┌──────────┐ ┌──────────┐ ┌──────────┐ ┌────────────┐  │    │
│  │  │ Data     │ │ Analysis │ │ Alert    │ │ Dashboard  │  │    │
│  │  │ Store    │ │ Engine   │ │ Manager  │ │ & Reports  │  │    │
│  │  │ (TSDB)   │ │ (Rules/  │ │ (Email,  │ │ (Grafana)  │  │    │
│  │  │          │ │  ML)     │ │  SMS)    │ │            │  │    │
│  │  └──────────┘ └──────────┘ └──────────┘ └────────────┘  │    │
│  └──────────────────────────────────────────────────────────┘    │
└─────────────────────────────────────────────────────────────────┘

Detection Techniques

1. Threshold-Based Detection

The simplest approach — compare measurements against predefined limits:

Point Machine P12:
  Normal operating current: 1.8A - 2.5A
  Warning threshold: > 3.0A or < 1.5A
  Alarm threshold: > 3.5A or > 5.0s duration

  Current reading: 3.2A → WARNING generated

Effective for detecting gross faults but may miss gradual degradation.

2. Trend Analysis

Track measurements over time to detect drift:

Track Circuit TC-07 receiver voltage (mV):
  Week 1:  850 mV  ████████████████████
  Week 2:  830 mV  ███████████████████
  Week 3:  790 mV  ██████████████████
  Week 4:  720 mV  ████████████████     ← Downward trend detected
  Week 5:  680 mV  ███████████████      ← Alert: approaching minimum threshold
  Minimum: 600 mV  ──────────────── (failure threshold)

3. Waveform Pattern Analysis

Compare current waveforms against known-good reference patterns:

Point Machine Current — Normal vs Faulty:

Normal:  ╭──╮
        ╱    ╲──────╮
       ╱              ╲───    ← Clean transitions, consistent shape

Faulty:  ╭╮╭──╮
        ╱ ╲╱    ╲─────╮
       ╱                ╲───  ← Spikes indicate mechanical binding

4. Machine Learning Models

Advanced FDS systems use ML for:

• Anomaly detection — Identify unusual patterns without predefined rules
• Remaining useful life prediction — Estimate when equipment will fail
• Classification — Categorize fault types (electrical, mechanical, environmental)
• Correlation — Identify relationships between multiple sensor readings

Implementation Example: Point Machine FDS

Sensor Setup

Point Machine
     │
     ├── Current Transformer (CT) ──→ ADC Channel 1 (motor current)
     ├── Detection Circuit Tap ──────→ ADC Channel 2 (detection voltage)
     ├── Supply Voltage Monitor ─────→ ADC Channel 3
     ├── Operation Counter ──────────→ Digital Input 1
     └── Position Feedback ──────────→ Digital Inputs 2-3 (N/R)

Data Collection

For each point operation, the FDS records:

1. Current waveform — Sampled at 1 kHz for the entire operation duration
2. Operation time — Start to completion (detection achieved)
3. Peak current — Maximum current drawn
4. Detection time — Time to achieve detection after reaching end position
5. Supply voltage — During operation (voltage drop indicates wiring issues)

Analysis Logic

# Simplified point machine health check
def analyze_point_operation(waveform, reference):
    peak_current = max(waveform.current)
    duration = waveform.end_time - waveform.start_time
    correlation = correlate(waveform, reference)

    health_score = 100  # Start at 100%

    if peak_current > CURRENT_WARNING:
        health_score -= 20
        create_alert("High peak current", severity="WARNING")

    if duration > DURATION_WARNING:
        health_score -= 15
        create_alert("Slow operation", severity="WARNING")

    if correlation < CORRELATION_THRESHOLD:
        health_score -= 30
        create_alert("Abnormal waveform shape", severity="ALARM")

    return health_score

FDS Dashboard

A well-designed FDS dashboard provides:

Overview Screen

• Geographic map with equipment health overlay (green/amber/red)
• Summary statistics (total assets, healthy, warning, alarm)
• Recent alerts and trends

Equipment Detail Screen

• Current health score
• Historical trend graphs
• Last N operation waveforms
• Maintenance history
• Predicted next maintenance window

Alert Management

• Prioritized alert list (critical → warning → info)
• Acknowledgment workflow
• Escalation rules
• Maintenance work order integration

Benefits of FDS

1. Reduced failures in service — Detect 60-80% of developing faults before service impact
2. Optimized maintenance — Replace time-based with condition-based maintenance
3. Lower costs — Fewer emergency call-outs, planned material procurement
4. Improved safety — Early detection of safety-critical degradation
5. Better asset knowledge — Comprehensive data on equipment health and lifecycle
6. Regulatory compliance — Documented evidence of equipment monitoring

Challenges

• Sensor reliability — Monitoring equipment must itself be reliable
• Data volume — High-frequency sampling generates large datasets
• Communication bandwidth — Especially for remote locations
• Integration — Connecting with existing legacy signalling equipment
• Expertise — Requires signal engineering + data science skills

Conclusion

Failure Detection Systems represent the future of railway infrastructure maintenance. By combining real-time monitoring, trend analysis, and increasingly machine learning, FDS platforms enable railways to shift from reactive fire-fighting to proactive condition-based maintenance — improving safety, reliability, and cost efficiency.

Related: IoT-Based Predictive Maintenance for Railway Signalling, Event Logger & Data Logger, and Point Machine Working Principle.