Auto Correlation Simplifies Vibration Analysis, and Enhances Efficiency of Rotating Machinery Maintenance


Vibration analysis is one of the most successful techniques for monitoring the condition of rotating equipment, but unless you are a vibration specialist the information can often be difficult to decipher. How can peak value analysis and auto correlation help improve maintenance efficiency?


Misalignment, gear defects, insufficient lubrication, pump cavitation and rolling element bearing defects are all problems associated with rotating machinery that result in increased vibration. Vibration analysis is therefore one of the most important techniques for monitoring the condition of such machines as part of a predictive maintenance programme. The periodic and, where appropriate, continuous collection of vibration data enables potential problems to be identified earlier. This helps to prevent unexpected failures that can cause safety incidents and production loss. Maintenance can be scheduled at appropriate periods of downtime. The benefits of vibration analysis are widely recognised in terms of reduced maintenance costs and the increased safety and plant efficiency it helps to provide. However, with a shortage of experienced plant maintenance engineers, companies often do not have personnel with the necessary ability to correctly interpret the often-complex vibration data available.

Vibration analysis relies on data collected from vibration sensors monitoring the rotating equipment. This data can be collected manually and periodically using handheld vibration analysis devices. Alternatively, equipment critical to production is often monitored on a continuous basis (often referred to as online monitoring) to ensure that changes that may indicate a potential problem are not missed in between manual rounds. Online monitoring systems also often incorporate protection functionality that helps to bring equipment to a safe state (offline) should an issue be identified.

Signal processing

In general, the analogue signal from a vibration sensor is routed via an analogue signal processor, converted into a digital format and then further processed digitally. The output of the vibration sensor is expressed in g units, and the signal processing may include the conversion of the signal to velocity units. The analogue signal (in g or velocity units) is usually passed through a filter immediately before being converted into a digital format, providing assurance that the digital representation of the analogue signal is correct.

By far the most common form of signal processing for analysing vibration from rotating equipment is the Fourier Transform. This uses a fast Fourier transform (FFT) algorithm to enable the signal to be converted and to construct the spectrum either in acceleration or velocity units. This spectral analysis is helpful in separating the band-limited signal into periodic components related to the turning speed of the machine.

Standard spectral analysis is the traditional method used to gain insight into machinery problems that create vibration, but its complexity makes it difficult for anyone who is not a specialist vibration engineer to analyse and interpret the data. In contrast, the peak value analysis (PeakVue) methodology introduced by Emerson to help analyse vibration data has proven to be very effective, presenting the information in a way that makes it easier for personnel other than vibration specialists to interpret and identify problems.

Peak value analysis

Peak value analysis technology provides a simple, reliable indication of equipment health via a single trend - filtering out traditional vibration signals to focus exclusively on impacting faults, where metal parts come into contact with each other.

In this method, peak values are observed over sequential discrete time intervals, captured, and then analysed. The analyses are:

a. the peak values (measured in g’s).

b. spectra computed from the peak value time waveform.

c. the auto correlation coefficient computed from the peak value time waveform.

All three analysis tools enable the defect, and often its severity, to be identified.

As a measure of impacting, peak value analysis readings are much easier to interpret. A healthy machine that is correctly installed and well lubricated shouldn’t have any impacting. This establishes the zero principle: the peak value measurement on a healthy machine should be at, or close to, zero.

As common machinery faults begin to appear on rotating equipment, the peak value reading typically can be evaluated using the so-called Rule of 10’s. This applies to rolling element bearing machines operating between 1000 and 4000 rpm. It simply states that when the peak value levels reach 10, there is some problem with the machine; when they double to 20 there is a serious problem; and when they double again to 40 there is a critical problem (see Figure 1).

Rule of 10’s example

As an example of how the Rule of 10’s operates, let’s consider a typical process pump running at between 900 and 4000 rpm as it passes through the four stages of bearing failure before progressing to machine failure.

Stage 1-The defect is not visible to the human eye and there is no change in the overall vibration, but peak value analysis already provides an indication that something is happening. When the peak value rises to a value of 10, this indicates that there is a problem with the bearing.

Stage 2 - Small pits begin to appear and the bearing has less than 10% of its service life remaining. Typically, overall vibration still does not provide an indication of the developing faults, but the peak value level continues to climb. When it doubles to 20, this indicates a serious problem with the bearing.

Stage 3 - the bearing damage is now clearly visible. You may start to see a small increase in overall vibration of +/- 10 percent. Meanwhile, the progression in fault severity is obvious using peak value analysis.

Stage 4 - the overall vibration may rise by 20 percent or more. In comparison, the peak value level continues to increase sharply – perhaps as high as 40 g’s – and signals that the bearing is approaching the end of its life.

Machine failure - there will be a marked increase in the overall vibration at the point of actual failure, but too late to support planned maintenance. This is, in effect, notification that the machine is shutting down. In contrast, peak value analysis has been indicating a developing fault over the past weeks and months. Immediately prior to failure, peak value levels may surge rapidly to 50 g’s or higher.

Operators with no special training in machinery diagnostics can use peak value analysis measurements quickly and easily to determine both when a piece of rotating equipment is healthy and when an abnormal situation is present. Once an abnormal situation has been identified, detailed diagnostic information can be extracted from the peak value analysis waveform or spectrum to determine the exact nature of the defect. This method can be used to visualise distress signals on a machine that are simply not visible with other vibration measurements. Earlier indication of developing defects facilitates optimum maintenance planning and minimises the impact on production.

Auto correlation

Auto correlation is a time domain analysis, computed from the peak value time waveform that is useful for determining the periodicity or repeating patterns of a vibration signal. The auto correlated waveform can be presented in a circular format, which makes interpretation of the data much more straightforward.

On the following page are some examples of how vibration analysis data can be viewed, using standard spectrum analysis, time waveform, auto correlation, and finally auto correlation in a circular format.


Peak value methodology has proven to be a very useful tool for vibration analysis in rotating equipment applications where normal spectral analysis has proven to be less effective. Using auto correlation and circular displays, problems can be easily identified without vibration analysis experience. This helps to simplify maintenance tasks, enabling a greater number of devices to be effectively monitored. Previously difficult to identify problems will be quick and simple to diagnose at an early stage, helping repair work to be scheduled, preventing machinery failures, reducing overall maintenance costs and improving plant safety and efficiency. 



Vincent Burson