TY - GEN

T1 - Identification of Objects in Oilfield Infrastructure Using Engineering Diagram and Machine Learning Methods

AU - Asmai Ismail, Muhammad Hami

AU - Tailakov, Dmitry

N1 - Публикация для корректировки.

PY - 2022

Y1 - 2022

N2 - A processing plant consists of huge number of equipment, which working together to achieve their designed function and target output. Each of the equipment (assets) need to be managed properly to ensure maximum uptime of the plant operation and avoid any unnecessary downtime. To properly manage plant assets, each of the equipment must be properly recorded in a Computerized Maintenance Management System (CMMS) for management and tracking. Normally, each equipment is referenced by their unique identifier called Tag, created during engineering design phase. Currently, in the industry the standard way of collecting equipment Tag is very manual, involving the visual screening of each page of as-built P&IDs and other drawings. Each identified Tags are then recorded in a database in a hierarchical manner (parent-child structure) and each of them will be assigned with appropriate classification according to the owner's standard. This current process is time consuming and labour intensive while AI could be implemented to solve this problem in the future. Thus, this study was initiated with the objective to implement computer vision and deep learning architectures to identify objects that appears on as-built engineering piping and instrumentation diagram (PID). The algorithm then tasked to classify objects detected according to ISO 14224 Equipment Classification, match it to the corrent Tag that appears in the same drawing and finally create parent-child structure between the objects identified. The long-term aim of this project to prove a hypothesis: a good pre-trained model of P&ID object detection can be achieved.

AB - A processing plant consists of huge number of equipment, which working together to achieve their designed function and target output. Each of the equipment (assets) need to be managed properly to ensure maximum uptime of the plant operation and avoid any unnecessary downtime. To properly manage plant assets, each of the equipment must be properly recorded in a Computerized Maintenance Management System (CMMS) for management and tracking. Normally, each equipment is referenced by their unique identifier called Tag, created during engineering design phase. Currently, in the industry the standard way of collecting equipment Tag is very manual, involving the visual screening of each page of as-built P&IDs and other drawings. Each identified Tags are then recorded in a database in a hierarchical manner (parent-child structure) and each of them will be assigned with appropriate classification according to the owner's standard. This current process is time consuming and labour intensive while AI could be implemented to solve this problem in the future. Thus, this study was initiated with the objective to implement computer vision and deep learning architectures to identify objects that appears on as-built engineering piping and instrumentation diagram (PID). The algorithm then tasked to classify objects detected according to ISO 14224 Equipment Classification, match it to the corrent Tag that appears in the same drawing and finally create parent-child structure between the objects identified. The long-term aim of this project to prove a hypothesis: a good pre-trained model of P&ID object detection can be achieved.

UR - https://www.scopus.com/record/display.uri?eid=2-s2.0-85150593818&origin=inward&txGid=8885f76a0aabb8722ec28d2ef6c7adb8

UR - https://www.mendeley.com/catalogue/5b045e72-75e3-33e7-8b2e-b9818a8fd9b8/

U2 - 10.2523/IPTC-22467-EA

DO - 10.2523/IPTC-22467-EA

M3 - Conference contribution

SN - 9781613998335

T3 - International Petroleum Technology Conference, IPTC 2022

BT - International Petroleum Technology Conference, IPTC 2022

PB - International Petroleum Technology Conference (IPTC)

ER -