At the end of the seventies of the last century it was noticed that a change of the character of the noise and vibration may be related to changes in the machine’s technical condition. The researchers started to record vibration signals, with particular emphasis on vibration signals generated by working technical facilities.

Each element of a working machine generates a unique vibration signal. The vibration signal will be different for a running electric motor, generated by a gear or an unbalanced shaft. The described phenomenon may illustrate the diagram shown in figure 1.

A vibration signal recorded by a specialized device contains information about all elements working within reach of the sensor. The nature of this signal depends on factors such as kinematics, speed or power.

In the early stages of vibro-diagnostics development it was assumed that the signal measured on the bearing housing will mainly include components specific to the operation of the shaft where the bearing was seated. Similarly, placing the sensor on the housing of the parallel gear would give us the information related mainly to the process of the tooth meshing. The indicators of wear of individual elements in this approach are simple analysis of the vibration signal, like its energy or peak-to-peak amplitude.

Implementation of diagnostic tasks in the manner described above can be done in two ways. The first method is regular measurement of vibration at the designated points with specialized mobile device and regular statistical analysis of measured signals. The second way is the real-time condition monitoring system.

Online Maximizes the Safety

Although periodic measurements of the signals on the selected machines are very useful in the early detection of failures, they do not give full protection, particularly for critical machines that require continuous diagnostic monitoring. Online condition monitoring systems are used to minimize the risk of a failure and to provide maximized operational safety.

Monitoring systems are complex devices, consisting of many components such as sensors, data acquisition modules, servers or workstations.

Sensors can measure such quantities as vibration, temperature, pressure, stress and many others. This signal is then recorded by data acquisition modules and sent to a central server system, which collects data from all data acquisition units.

The signals from the measuring channels are subject to advanced post-processing and the system takes control if the alarm thresholds are exceeded. In the case of alarm threshold breach, the information is sent to the operator. The measurement data and the detected alarms are recorded in the database, which gives the history of the machine’s work. In order to connect to the system, a typical PC running a dedicated programme is normally used at the operator’s stations. With the help of this computer it is possible to view the recorded signals from the machine and configure the system.

Such systems were not applicable in most industries in the initial period of development. This was due to the relatively high costs and the limited possibilities of analysis. The main customer for diagnostic and monitoring systems was the power industry, and the first machines supported by the continuous condition monitoring were steam turbines. With the development of measurement techniques and information technology, the prices of diagnostic and monitoring systems had fallen enough to allow for their expansion into other industries. Systems originally installed on turbines were therefore implemented to monitor fans, pumps and other less expensive devices. However, it was still not possible to diagnose machines with more complex kinematics.

The general growth of vibration, as initially assumed, allowed the detection of damage development of one of the elements of the machine. However, it did not allow for unambiguous identification of the damage. The development of digital signal processing techniques was very helpful, as they enabled the implementation of frequency analysis to monitoring systems. This enabled the representation of a vibration signal in a spectrum view containing information about the energy of individual components (Fig. 2).

Fig.1. The spectrum view.

Fig.2. Spectrum view of the signal with variable speed before applying the order analysis (top) and after the application (bottom).

Fig.3. Vibrations generated by the various components of the kinematic chain.

The new functionality resulted in the implementation of the vibrodiagnostic systems to a number of industries on a large scale. Nowadays it is possible to perform much more precise diagnostics of rolling bearings, due to the implementation of the order analysis, allowing the detection of even very subtle changes in the dynamic state of a bearing. The solutions used nowadays help maintenance departments to monitor the whole technological processes.

Also for Varying Speeds

As mentioned previously, the signal generated by the machine vibration is affected not only by its structure, but also by the operational conditions. A large group of the machines working in the industry are machines operating in variable conditions. An excellent example of such machines is wind turbines, for which parameters such as speed and power can change within a few minutes by up to several tens of percent.

For such machines it is impossible to apply frequency analysis, because the spectral lines (Fig. 3) are becoming fuzzy, and their unambiguous identification is impossible.

The solution to this problem was the introduction of the order analysis to the vibration analysis techniques. One of its fundamental assumptions was the continuous measurement of speed synchronized with the vibration signal registration. With this method it is possible to perform diagnostics regardless of the changeable operating conditions.

Falling prices of the diagnostic services and the growing machine safety and reliability standards cause the monitoring systems to expand onto further industries. Often there is a need for vibrodiagnostic supervision over machines with a very complex structure at the lowest possible cost, which results directly in a relatively small number of installed sensors. This creates a pressure on engineers to introduce increasingly complex and sophisticated methods of vibration analysis. Currently the newest scientific achievements from the area of digital signal processing are being implemented in the industrial applications. This allows detecting and drawing conclusions about the phenomena not detectable by classical methods.

The measurement equipment, computing capabilities and latest achievements of science that are available to us nowadays make the implementation of vibration-based condition monitoring possible in almost every industry. At the same time, falling prices of diagnostic services encourage the implementation of comprehensive solutions, not just by large corporations but also small manufacturing facilities.

The author is Ph.D. at AGH University of Science and Technology in Krakow.

Adam Jablonski

Senior Research & Development Analyst at EC Systems,

ajablonski@ec-ystems.pl