Scribd
Upload
22 views44 pages

2 - CBM - Prof A K M

Uploaded by

mis2022sbp
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
22 views44 pages

2 - CBM - Prof A K M

Uploaded by

mis2022sbp
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 44

VIBRATION SIGNAL BASED MACHINE FAULT DIAGNOSIS

A K Mukhopadhyay
Prof. (Retd); IIT(ISM)
1.0 Introduction: Increasing demands for ensuring safety and reliability of machines, fault
diagnosis (FD) analysis has received considerable attention in modern in modern
industries. Vibration, Acoustics, Oil and Thermal imaging monitoring techniques are
commonly employed to monitor the condition of machines in time. Among these,
vibration signals provide better indicators of damages. In recent times Machine
Learning (ML) models have been introduced for data collection, signal processing and
feature extraction of vibration signals. Potential faults and abnormalities are
identified in advance before any catastrophic failure. The process of fault diagnosis
can be divided into four categories: fault detection, fault identification, fault severity
assessment, fault growth and remaining useful life assessment.

2.0 Maintenance Practices in Industries: Four types of maintenance practices are


followed
(a) Routine Maintenance
(b) Breakdown Maintenance
(c) Planned Maintenance
(d) Condition Based Maintenance (CBM) through vibration analysis, oil analysis,
noise analysis and thermography analysis.

Routine maintenance is carried out on daily basis prior to start of a new shift. It includes
cleaning of the machine with water spray and air jets, lubricant condition in bearings, gear
boxes, presence of unusual noise emitting from the machine etc. Breakdown
maintenance is taken up when a machine undergoes a severe machine component
failure, cannot be put into operation till the fault repaired. In Planned Maintenance the
machine is stopped for thorough inspection irrespective of failure after certain hour of
operation or a targeted production is met. CBM is scheduled based on the information
received from vibration measurements, oil analysis and acoustic measurements.
Breakdown maintenance not recommended as it is too late to put the machine into
operation. Planned Maintenance is too early to stop a machine. CBM ensures
uninterrupted operation of the machine in good condition. CBM is required as and when
the machine requires to be stopped.

3.0 Vibration Features


Vibration measuring sensors are of following types:

# Velocity Transducer

# Accelerometers

#Non-contact Probes
Fault and dominant frequency chart
4.0 ACOUSTIC ANALYSIS: This area deals with measurement of noise
signals. From the signal, noise level and dominant frequencies are determined.
The noise analyzer uses a programmed software to evaluate sound level,
frequency, amplitude and wavelength. Statistical analysis approaches in the
assists in fault diagnosis.

5.0 MACHINE LEARNING (ML): Machine Learning is a subdomain of AI and


is data driven technology and focuses on developing systems that learn and
improve performance based on large pool of data ingested.ML gives
computers the capacity to learn without being explicitly programmed. ML
make data from the input data by modelling the algorithm from data.
Algorithms (Rules) find the pattern and relationships from the acquired data
with the faults. Let one of the rules states that (i) Vibration Level is maximum
in horizontal direction (ii) Dominant Frequency 4XRPM (iii) 2 signals are 1800
out of phase, then the Fault is UNBALANCE.
6.0 SIGNAL PATTERNS
(a) Orbit Motion: It represents the path of the shaft center. Two orthogonal
probes required to observe the motion of the shaft. Motion can be
observed in the real time by feeding the output of the orthogonal probes
to the X and Y of a dual channel oscilloscope.

(b)Signal Averaging: Averaging is carried out to improve the S/N ratio in the
signal buried under noise to improve accuracy in diagnosis. Time domain
and frequency domain signals are averaged 1, 4, 8, 16, 64, 256, 1024, 4096,
16384 times and so on up to maximum measurement time of interest.
(c) FFT Signals: Vibration signal is broken into individual frequencies to
examine individual faults in the machine, if any.
(d) Transfer Function (TF): The output signal shows some degree of
uncertainty in its display compared to its input signal. TF is the ratio of the
output signal to input signal non-linearity.
(e) Bandwidth: The bandwidth accommodates maximum amount of
frequency domain/time domain signals as per convenience for analysis.
(f) Autocorrelation: It indicates difference in response of output signals as
a function of time.
(g) Cross correlation: It indicates similarity in response of output signals as
a function of displacement.
(h) Coherence: It indicates how closely two different signals are related to
each other. It value will lie between 0 and 1
(i) Resolution: It is the ratio of bandwidth to number of lines of resolution
of a signal required for convenient analysis. For example, say the bandwidth
is 100 Hz and number of lines of resolution is 1600 selected, then the
smallest frequency of 0.0625 Hz is detectable.

7.0 CONCLUSION: The text presents vibration methods for fault diagnosis
of industrial machines. It discusses vibration data collection techniques,
instruments used, signal processing techniques, feature extraction
methods and introducing ML methods.
*********************

You might also like