Comparison of Pre-Trained CNNs for Audio Classification Using Transfer Learning

The paper investigates retraining options and the performance of pre-trained Convolutional Neural Networks (CNNs) for sound classification. CNNs were initially designed for image classification and recognition, and, at a second phase, they extended towards sound classification. Transfer learning is a promising paradigm, retraining already trained networks upon different datasets. We selected three ‘Image’- and two ‘Sound’-trained CNNs, namely, GoogLeNet, SqueezeNet, ShuffleNet, VGGish, and YAMNet, and applied transfer learning. We explored the influence of key retraining parameters, including the optimizer, the mini-batch size, the learning rate, and the number of epochs, on the classification accuracy and the processing time needed in terms of sound preprocessing for the preparation of the scalograms and spectrograms as well as CNN training. The UrbanSound8K, ESC-10, and Air Compressor open sound datasets were employed. Using a two-fold criterion based on classification accuracy and time needed, we selected the ‘champion’ transfer-learning parameter combinations, discussed the consistency of the classification results, and explored possible benefits from fusing the classification estimations. The Sound CNNs achieved better classification accuracy, reaching an average of 96.4% for UrbanSound8K, 91.25% for ESC-10, and 100% for the Air Compressor dataset.

Improvement of Dissolved Oxygen in Perlis River based on Various Aeration Systems

Water pollution is closely related to the Water Quality Index (WQI). One of the parameters in classifying WQI is dissolved oxygen (DO) that can be improved by introducing the surface and subsurface aerations. Herein, the Perlis River’s water quality was investigated by evaluating the DO’s improvement based on various aeration systems. The changes of DO (mg/L) and DO improvement (%) were evaluated during both low and high tide conditions. A total of 9 sets of data collection had been studied by comparing base DO (without running of aeration) and measured DO (with running of aeration) of river. The DO sensor was used to measure the changes of DO in the aeration measurement system. Results found that the DO improvement managed to achieve 74.89%, 10.18%, 35.58%, and 52.45% for water jet, air compressor, commercial venturi, and DIY venturi, respectively. Besides, different behaviour of DO’s improvement was observed during low and high tide conditions.

Torsional vibrations in a shaft train of an air compressor: modification, calculation and measurement

Purpose Due to the installation of the new, more powerful gearbox and the dismantling of the exciter machine, the vibration characteristics of the shaft train will be changed. Therefore, it is necessary to reassess the shaft train. It is to be investigated if the shaft train of the compressor meets the general requirements for bending and torsional vibrations and can be safely operated within the electrical network. The purpose of this paper is to show the necessary alignment of modification, calculation and measurement in such a project.. Design/methodology/approach After some modification work on the shaft train of an air compressor, it was necessary to do some engineering calculations regarding the bending natural and torsional natural frequencies and their mode shapes. The correctness of the calculated values was proven by vibration measurements performed at the shaft train in operation. Findings It can be concluded that the change and replacement of rotating equipment in a shaft train never should be done without any engineering calculations in advance and measurements after the component modification. Most important is that the calculation results have to be compared with the measurement results for verifying the calculation assumptions. In the case described above, one can see that theory and practice match well. In addition to this, the very low damping of torsional vibrations is proved again, which can be a significant problem in some situations. Originality/value Also, today one can find torsional vibration measurements of rotating machines, including frequency, magnitude and damping factor, very seldom. Especially for smaller machines, there are no real comparisons between calculation and measurement are usual. This paper shows that an alignment between theoretical and practical approaches is necessary to avoid operational problems for rotating machines.