Quatitative evaluation of pipe wall thinnings utilizing the combination of guided inspections and artificial intelligence

2023 Fiscal Year Final Research Report

• Project/Area Number: 21K03750
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Review Section: Basic Section 18010:Mechanics of materials and materials-related
• Research Institution: The University of Tokushima
• Principal Investigator: NISHINO Hideo 徳島大学, 大学院社会産業理工学研究部(理工学域), 教授 (50316890)
• Project Period (FY): 2021-04-01 – 2024-03-31
• Keywords: 非破壊検査 / 超音波 / ガイド波 / 人工知能 / パーセプトロン / パイプライン / 減肉

Outline of Final Research Achievements

The aim of this study is to add a method that enables the quantification of wall thinning in the guided wave method, which is an extensive and highly efficient method. The method used is the supervised multilayer perceptron, which is the most widely used method in artificial intelligence.
The important original points of the supervised perceptron are (1) the feature given to the input layer is the multi-frequency guided wave reflectance, and (2) a large number of training data were constructed using an original mathematical model. When the method using these features was verified using artificial and actual machine thinning, a correct response rate of about 80% was obtained for artificial thinning and 100% for actual machine thinning at an estimated service width of ±0.5 mm.

Free Research Field

非破壊検査

Academic Significance and Societal Importance of the Research Achievements

本手法は，公共インフラや産業インフラとして多く用いられている各種導管(パイプライン) の健全性を高効率かつ非破壊で検査できる手法である。現在は拭き取り検査に留まる各種検査を全域検査に置換できる可能性を秘めており，安心安全社会の実現に向けた研究開発の一つである。安全性の向上のみならず，産業や社会公共におけるコストの低減にも寄与する重要な研究課題である。