Do you know how a type 316 stainless pressure vessel can fail?
The correct answer is transgranular stress-corrosion cracking caused by chloride stress corrosion !
Most pressure vessels are subject to some kind of corrosion or erosion. In most cases, corrosion can be minimized by either designing intelligently or controlling the environment carefully. In other situations corrosion is difficult to prevent. A vessel containing a corrosive substance kept in motion by a stirring paddle will deteriorate more rapidly for example.
Corrosion can attack the pressure vessel uniformly or locally. Large area corrosion wastes away materials uniformly from the vessel surface. Although this can lead to general reduction of thickness and strength in the long run, it is far less malicious comparing to localized corrosion. Precautions such as allowing extra thickness according to service life can be easily taken by designers.
Local attack on the other hand is more of an enemy to designers. Stress concentrations are produced at pits; grooves and notches start to form; and ultimately cracks cut through the weakest point of the vessel.
Corrosion and stress are actually interacting together and locally corroded metal experiences more stresses under normal pressure. And in return when the metal becomes more stressed, it corrodes faster. This combined action creates a vicious cycle that brings to sudden rupture and must be avoided at any cost.
Another example of the real hazard of stress induced by corrosion is the example of the NDK Crystal factory explosion in 2009.
NDK Crystals operated a factory producing crystals in an industrial area in Belvedere, Illinoi. They produced crystals for a variety of products, especially for electronic devices. The facility housed 8 massive cylindrical pressure vessels with 8 inch thick steel walls, standing 50 feet tall. Inside, Silica was mixed with corrosive sodium hydroxide solution at extremely high pressure and temperatures. The vessel were kept sealed for one hundred fifty days to grow large crystals of quartz. The silica and sodium hydroxide react with the ions in the walls of the steel vessels forming a layer of sodium ion silicate or acmite.
The company believed that this ion coating would protect the vessel from the corrosive effect of the chemical inside. Over the years, NDK was warned that corrosion may compromise the walls of the pressure vessels yet the company continued to operate those vessels without performing recommended inspections
In 2009, one of the vessels suddenly exploded, killed a truck driver, and caused million dollars of damage. The explosion was due to stress corrosion cracking in the vessels that had not been inspected for years.
Can we prevent such problems using FE Analysis?
Corrosion itself cannot really be simulated because it is more a phenomenological phenomenon than something that can be predicted accurately. There is too many factors to consider to make accurate prediction which make the potential of simulation quite low (Chemical reactions, velocity of fluid, decrease in thickness).
Nevertheless, some engineers have been using simulation to predict the zones were corrosion may appear. For that purpose, CFD Analysis can be used to investigate causes of cracking inside a pressure vessel. If there is little dispersion of the steam exiting a nozzle, the velocity of the steam mixing with cooler condensate can result in a violent reaction which causes localized heating and cooling of vessel shell. This kind of cycling can causes thermal stress which results in fatigue cracking (here is a case study I found)
In addition, The effect of metal loses due to corrosion on burst pressure can be studied using nonlinear FE Methods which may allows to predict the failure (here is another case study)
Different methods have been using FE Analysis to estimate stress corrosion cracking growth (Details here)
The list is certainly not complete and a specialist on the topic will have a lot more to tell about it.
What about you? Would you use FEA to predict stress-corrosion failure during the design phase or would you prefer some other method?