Understanding Fault Detection

Devices

Portable balancer & Vibration analyzer Balanset-1A

$2,347.12 Izvorna cijena bila je: $2,347.12.$1,901.46Trenutna cijena je: $1,901.46.

Parts

Vibration sensor

$106.96 Izvorna cijena bila je: $106.96.$71.30Trenutna cijena je: $71.30.

Parts

Optical Sensor (Laser Tachometer)

$147.36 Izvorna cijena bila je: $147.36.$83.19Trenutna cijena je: $83.19.

Devices

Balanset-4

$8,084.78

Parts

Magnetic Stand Insize-60-kgf

$54.67

Parts

Reflective tape

$11.88

Devices

Dynamic balancer “Balanset-1A” OEM

$2,061.90 Izvorna cijena bila je: $2,061.90.$1,663.78Trenutna cijena je: $1,663.78.

Definition: What is Fault Detection?

Fault detection is the process of identifying that a defect or abnormal condition exists in equipment through analysis of monitored parameters such as vibration, temperature, performance metrics, or other indicators. Fault detection answers the binary question “Is there a problem?” before proceeding to fault diagnosis (identifying the specific problem) and prognosis (predicting remaining life). It is the first and most fundamental step in condition-based maintenance, distinguishing normal operation from deteriorating or faulty conditions.

Effective fault detection provides early warning—detecting problems months before functional failure—enabling the lead time necessary for planned maintenance, parts procurement, and scheduled downtime that are the core value propositions of predictive maintenance programs.

Detection Methods

1. Threshold Exceedance

Simplest and most common:

• Compare measurement to predefined threshold
• If measurement > threshold → fault detected
• Example: Overall vibration > 7.1 mm/s triggers alert
• Advantages: Simple, automated, clear criteria
• Limitations: Requires proper threshold setting, lag time to exceed threshold

2. Trend Deviation

Detects changes from normal pattern:

• Increasing trend indicates developing fault
• Detect before absolute threshold exceeded
• Rate of change alarming (rapid increases)
• Advantages: Earlier detection, machine-specific
• Requirements: Historical trending data needed

3. Spectral Anomaly Detection

Identifying abnormal frequency components:

• New peaks appearing in spectrum (bearing frequencies, harmonics)
• Existing peaks growing in amplitude
• Pattern changes (sidebands developing)
• Advantages: Specific fault type indication
• Requirements: Spectral analysis capability, baseline spectra

4. Statistical Methods

• Values outside normal statistical distribution
• Outlier detection (> mean + 3σ)
• Control chart violations
• Advantages: Accounts for normal variability
• Requirements: Adequate statistical sample size

5. Pattern Recognition

• Machine learning algorithms
• Neural networks trained on normal vs. faulty signatures
• Automated anomaly detection
• Advantages: Can detect subtle patterns
• Requirements: Training data, computational resources

Detection Performance Metrics

Sensitivity (True Positive Rate)

• Percentage of actual faults detected
• Target: > 90-95% of real problems detected
• Higher sensitivity = fewer missed faults
• Measure: (True Positives) / (True Positives + False Negatives)

Specificity (True Negative Rate)

• Percentage of healthy equipment correctly identified as healthy
• Target: > 90-95% of healthy equipment not falsely alarmed
• Higher specificity = fewer false alarms
• Measure: (True Negatives) / (True Negatives + False Positives)

False Alarm Rate

• Percentage of alarms that were false (no actual fault)
• Target: < 5-10% false alarms
• High false alarm rate causes alarm fatigue
• Balance with sensitivity (trade-off)

Detection Lead Time

• Time from fault detection to functional failure
• Longer lead time = more value (time for planning)
• Typical: Weeks to months for vibration-detected bearing faults
• Method-dependent: Envelope analysis detects earlier than overall levels

Challenges in Fault Detection

Early vs. False Detection Balance

• Very early detection increases false alarms
• Waiting for clear signals reduces lead time
• Optimize through multi-stage alarming
• Use multiple parameters for confirmation

Intermittent Faults

• Problems that come and go
• May be below threshold during periodic measurements
• Requires continuous monitoring or peak hold

Multiple Simultaneous Faults

• Several problems developing concurrently
• May mask each other in vibration
• Requires comprehensive analysis
• Multiple detection methods help

Multi-Parameter Fault Detection

Vibration + Temperature

• Both increasing: Confirms bearing problem
• Vibration only: Mechanical issue (unbalance, misalignment)
• Temperature only: Lubrication or friction issue
• Combined confirmation reduces false detections

Multiple Vibration Parameters

• Overall level increase + bearing frequency emergence
• Confirms bearing fault specifically
• More confident detection than single parameter

Automation vs. Manual Detection

Automated Detection

• Advantages: Fast, consistent, 24/7 capability
• Methods: Threshold checking, statistical algorithms, machine learning
• Limitations: Can miss subtle problems, may generate false alarms

Manual (Expert) Detection

• Advantages: Human judgment, context awareness, pattern recognition
• Methods: Spectrum review, waveform inspection, multi-parameter correlation
• Limitations: Time-consuming, not scalable, expertise required

Hybrid Approach (Best Practice)

• Automated detection for screening
• Expert review of exceptions
• Combines efficiency with accuracy
• Standard in mature programs

Fault detection is the foundational capability that enables predictive maintenance, identifying developing problems early enough to enable planned interventions. Effective fault detection—combining appropriate detection methods, properly set thresholds, and balance between sensitivity and specificity—provides the early warnings that maximize equipment utilization while minimizing maintenance costs and failure risks.

---

← Back to Main Index