A Visual Method to Detect Broken Reciprocating Compressor Valves Under Varying Load Conditions

Reciprocating compressors are heavily used in modern industry, for instance for gas transportation and storage. In many cases, compressors run at high capacity and without backup. Hence unexpected shutdowns lead to large losses in productivity. Furthermore, there is an economic trend towards saving labor costs by reducing the frequency of on-site inspection. Such considerations mean that compressors are run by remote control stations and monitored by automated technical systems. In this case, the system must be able to retrieve and evaluate relevant information automatically to detect faulty behavior. The proposed method evaluates time-frequency representations (spectrograms) of vibration measurements at the valve covers. Based on previous publications, we know that a cracked or broken valve influences the amplitudes of the power spectrum in certain frequency bands. Furthermore, it is obvious that the load control system changes the timing of the valve events. Of course, both factors are reflected in the spectrogram. Keeping that knowledge in mind, we have a look at the point-wise difference of a faultless reference spectrogram and a test spectrogram. Depending on the fault state of the valve and the load levels, it shows specifically shaped structures. The positions of the structures within the spectrogram are varying unpredictable with the valve type and the load. Hence, an automated detection would be hard to realize. Additionally, measurement noise would make the detection even more difficult. Both problems can be solved by applying two-dimensional autocorrelation to the point-wise spectrogram difference: the significant structures are centered and the noise effects are reduced. Thus makes it easier to recognize and evaluate the state of the test spectrogram instantly.