I zat Khaled

Abstract

Pressure vessels (PV) are intended for the production, storage, or transportation of products under high pressure. The energy contained in these equipment’s is often very important and can, in case of failure of the enclosure, cause human, material and environmental damages in the vicinity of the accident. Thus, optimal design, maintenance and structural health monitoring of pressure vessels are major concern for the industrial actors of this field.

The first objective of this study is therefore to contribute to the development of a methodology for real-time assessment of the integrity of PV, by implementing a novel predictive maintenance strategy based on the Internet of Things (IoT) and digital twin technologies. The second technological goal is to develop methods to optimize the design of the next generation pressure vessels and increase their operational reliability.

According to the definition of the Alliance Industry of Future: “A digital twin is an organized set of digital models representing a physical entity of the real world to respond to specific problems and uses”.  The digital twin is a virtual clone of a physical system or process. It systematically implies the existence of a "digital model" coupled with the object it copies. Depending on the system concerned and the desired use, it can be a geometric, multi-physical, functional, behavioral, and decision-making model. It must evolve over time like its real twin. It makes it possible to understand, analyze, predict, or optimize the operation and management of the real entity.

In this paper, first is presented a state of the art on the pressure vessels design according to codified rules (CODAP…), and on various existing maintenance strategies. Then the developed physical model is presented and discussed. This, considering specificities like welds, complex stress states and local defects. Welds are modeled and analyzed according to IIW recommendations based on a stress life approach, non-welded regions with complex stress states are analyzed with multiaxial fatigue criteria and damage tolerance is applied for zones with local defects.

The developed model, once calibrated with experimental results, allows also generating data (stress, strain, damage…) to feed the data science model and reconstruct the stress field in the complete equipment from a reduced number of sensors. The latter is done thanks to development and improvement of an intelligent sensor placement code, with the aim of minimizing the number of sensors needed to monitor the health of the equipment.

Session