TY - JOUR
T1 - Statistical analysis of cavitation erosion impacts in a vibratory apparatus with copulas
AU - Taillon, Gabriel
AU - Onishi, K.
AU - Mineshima, T.
AU - Miyagawa, Kazuyoshi
PY - 2019/3/28
Y1 - 2019/3/28
N2 - A method of analysis of cavitation peaks (impact events) using copulas is developed. Impact events, otherwise known as peaks, are defined as maximum in the pressure amplitude applied to a material surface. These impact events were measured using a high speed pressure sensor in a cavitation apparatus based on the ASTM G32 standard. A total of 46180 impacts were measured over 100 realizations of 4ms long recording. First, the impact duration and amplitude's joint marginals are modeled as gamma distribution (part of the exponential family), determined by a Kolmogorov-Smirnov test (KS test). Then, copulas enable the study of the dependence structure of the measured impacts characteristics. The measured parameters are shown to not be independent but instead have a complex, asymmetric dependence structure. There are almost no impacts that have a combination of a high amplitude (>12MPa) and low duration (<5μs). The Tawn copula best fitted the data, as determined by a maximum likelihood method. An extension of the KS test to two dimensions demonstrated that the copula is a better fit compared with a joint distribution of independent marginals.
AB - A method of analysis of cavitation peaks (impact events) using copulas is developed. Impact events, otherwise known as peaks, are defined as maximum in the pressure amplitude applied to a material surface. These impact events were measured using a high speed pressure sensor in a cavitation apparatus based on the ASTM G32 standard. A total of 46180 impacts were measured over 100 realizations of 4ms long recording. First, the impact duration and amplitude's joint marginals are modeled as gamma distribution (part of the exponential family), determined by a Kolmogorov-Smirnov test (KS test). Then, copulas enable the study of the dependence structure of the measured impacts characteristics. The measured parameters are shown to not be independent but instead have a complex, asymmetric dependence structure. There are almost no impacts that have a combination of a high amplitude (>12MPa) and low duration (<5μs). The Tawn copula best fitted the data, as determined by a maximum likelihood method. An extension of the KS test to two dimensions demonstrated that the copula is a better fit compared with a joint distribution of independent marginals.
UR - http://www.scopus.com/inward/record.url?scp=85063933812&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85063933812&partnerID=8YFLogxK
U2 - 10.1088/1755-1315/240/6/062035
DO - 10.1088/1755-1315/240/6/062035
M3 - Conference article
AN - SCOPUS:85063933812
SN - 1755-1307
VL - 240
JO - IOP Conference Series: Earth and Environmental Science
JF - IOP Conference Series: Earth and Environmental Science
IS - 6
M1 - 062035
T2 - 29th IAHR Symposium on Hydraulic Machinery and Systems, IAHR 2018
Y2 - 16 September 2018 through 21 September 2018
ER -