Application of Hougaard stochastic model for flow-accelerated corrosion wall thinning in an orifice
- PDF / 560,939 Bytes
- 4 Pages / 595.276 x 790.866 pts Page_size
- 40 Downloads / 170 Views
ORIGINAL RESEARCH
Application of Hougaard stochastic model for flow‑accelerated corrosion wall thinning in an orifice Mahendra Prasad1 · V. Gopika1 · Arunkumar Shridharan2 Received: 21 November 2019 / Accepted: 15 June 2020 © Society for Reliability and Safety (SRESA) 2020
Abstract In this paper, Hougaard process stochastic degradation model has been used for the prediction of wall thickness reduction in pipes between two consecutive in-service inspections. Experimental data on FAC in gypsum orifice, from the literature, were used. There were nine locations downstream of the orifice where the wear of wall was measured at 1 h, 2 h, 3.5 h, 4.5 h, 6.5 h, 9.6 and 13 h into the experiment. The predicted wall thickness reduction compared well with the experimental wall thickness reduction from 9.6 h to 13 h. Keywords Hougaard process · Saddle point · Maximum likelihood estimation · Flow accelerated corrosion
1 Introduction Components used in power industry degrade over time. This is caused by different types of corrosion, fatigue due to operational cycles and material changes. FAC is a combined electrochemical and diffusion process, whereby small quantity of pipe wall material is progressively removed. This process involves coupled mechanisms of electrochemical reactions and fluid dynamics, dependent on geometry, temperature, fluid velocity, pH and dissolved oxygen. In NPPs, straight pipes in primary and secondary sides undergo non-uniform FAC-induced wall thickness reduction. This becomes more severe in bends and orifices, due to an increase in fluid turbulence. Mechanistic modelling of FAC (Poulson 2014; Lee et al. 2007; Keck and Griffith 1987; Sanchez-Caldera et al. 1988) has been carried out, assuming that there is a steady state for the rate of corrosion with a constant value of system parameters. The difference between predicted thickness of pipe from a mechanistic model and the actual thickness can be high (Poulson 2014; Lee et al. 2007). In addition, in other forms of corrosion such as pitting, large uncertainties have been found (Melchers 2005a, b). The unreliability in
* Mahendra Prasad [email protected] 1
Bhabha Atomic Research Centre (BARC), Trombay, Mumbai, India
Indian Institute of Technology Bombay, Mumbai, India
2
mechanistic modelling for corrosion degradation has led to the development of probabilistic models for FAC (Yuan et al. 2009, 2008). Probabilistic modelling is a complementary tool for mechanistic modelling for prediction. Linear regression model (Meeker and Escobar 1998), non-linear model and mixed effect model (Yuan and Pandey 2009) were used to model degradation behaviour. The models make assumptions regarding the degradation process, for example, the measurement error (or random noises) is assumed as Gaussian with zero mean and a fixed variance, and a deterministic path of degradation. The non-linear models are often mechanistic and difficult to compute. Brownian motion is a continuous stochastic process which has been used in many fields to model random characteristic evolv
Data Loading...
