Solid particle erosion studies on polyphenylene sulfide composites and prediction on erosion data using artificial neural networks

• Published in: Wear Vol. 266; no. 1; pp. 184 - 193
• Main Authors: Suresh, Arjula, Harsha, A.P., Ghosh, M.K.
• Format: Journal Article
• Language: English
• Published: Amsterdam Elsevier B.V 05.01.2009; Elsevier
• Subjects: Applied sciences; Artificial neural networks; Composites; Exact sciences and technology; Friction, wear, lubrication; Machine components; Mechanical engineering. Machine design; Polyphenylene sulfide; Solid particle erosion; Polyphenylene sulfide; Composites; Solid particle erosion; Artificial neural networks; Short fiber; Sand; Surface metrology; Silica; Impact wear; Composite material; Phenylene sulfide polymer; Wear rate; Database; Ductile material; Erosion; Scanning electron microscopy; Brittle material; Thermomechanical properties; Neural network; Particle collision; Steady state; Tilt angle; Glass fiber; Mechanical shock; Tribology
• ISSN: 0043-1648; 1873-2577
• DOI: 10.1016/j.wear.2008.06.008

Solid particle erosion behavior of polyphenylene sulfide, reinforced by short glass fibers with varying fiber content (0–40 wt%) has been studied. Steady-state erosion rates have been evaluated at different impact angles (15–90°) and impact velocities (25–66 m/s) using silica sand particles (200 ± 50 μm) as an erodent. PPS and its composites exhibited maximum erosion rate at 30° impact angle indicating ductile erosion behavior. Though PPS is a brittle thermoplastic, incubation period was found for neat resin and its composites at normal impact ( α = 90°). The erosion rates of PPS composites increased with increasing amount of glass fiber. Morphology of eroded surfaces was examined using scanning electron microscopy (SEM) and possible wear mechanisms were discussed. Also, artificial neural networks (ANNs) technique has been used to predict the erosion rate based on the experimentally measured database of PPS composites. The results show that the predicted data are well acceptable when comparing them to measured values. A well-trained ANN is expected to be very helpful for prediction of wear data for systematic parameter studies.