Design of Machine Learning method for decision-making support and reliability improvement in the investment casting process - Random Forest image defect detection

Aug 4, 2026

Upload by Quinn

Designing reliable decision support for manufacturing motivates diagnostic and decision-support tools, especially in the investment casting industry where defect identification from microscopic images is often operator dependent. Microscopic inspection typically relies on manual judgment, creating risk for inconsistent assessments. The presented approach introduces a Random Forest classifier to predict whether a defect exists in input microscope images, turning image recognition into a decision support model. The goal is to support operators and improve the reliability and consistency of defect assessment.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages8

Reading Time20 min

AccessFree to download

Introduction

Machine learning in manufacturing and decision support

Reliability and risk in visual inspection

Motivation for microscopic examination in investment casting

Methodology

Background on the investment casting process

Design of Machine Learning method for decision-making support and reliability improvement in the investment casting process Antonia Antoniadou a,b, Konstantinos Kyprianidis a, Ioanna Aslanidou a, Anestis Kalfas c and Dimitrios Siafakas b a Mälardalen University, Vasteras, 72220, Sweden, b TPC Components AB, Hallstahammar,734 92, Sweden, c Aristotle University of Thessaloniki, 54124 Thessaloniki, Greece [antonia.antoniadou@mdu.se](antonia.antoniadou@mdu.se) Abstract The need to improve reliability and support decision-making in manufacturing has drawn attention to the application of diagnostic and decision-support tools. Particularly in the investment casting industry, datadriven methods can be the enabler for process diagnostics and decision support. Images from the microscopic examination in the investment casting process are used as data input, to detect defects in produced pieces. The microscopic examination usually relies solely upon the ability of the operator to determine whether animage from the microscope contains a defect. Therefore, an effective strategy for this decision-making process is crucial to improve the reliability of the examination. The use of the machine learning classifier Random Forest is introduced to derive predictions on the existence of a defect in the input image. This work focuses on employing machine learning tools for image recognition and the developed approach constitutes a decision support model to assist the operator and improve the reliability of their assessment. 1. Introduction During the last decade, machine learning (ML) techniques have been widely implemented indifferent production processes, aiming to enhance the quality of the products, apply process diagnostics, or support decision-making (Esmaeilian et al., 2016) Utilization of ML methods has found application in production operational management centers to facilitate decision-making processes (González Rodríguez et al., 2020), or use predictions to support decisions in inventory management (Mohamed & Saber, 2023) . The need to improve the reliability of decision-making for fault detection and diagnostic processes represents one of the strategic objectives of many industries. In manufacturing, reliability refers to machines, equipment, and systems being able to perform their intended functions with consistency and predictability. Providing reliable products is vital to the success of the industry, as traditionally reliability is evaluated by the final product quality (Safhi et al., 2019) . Numerous measures can betaken to increase manufacturing reliability, such as regular maintenance and calibration of equipment, as well as diagnosing faults in components or systems. The microscopic examination mentioned in this work is a part of the investment casting process, a process aiming to create components that can be used in turbomachinery applications, characterized by high geometrical complexity, and later subjected to demanding performance conditions. The production of such parts has multiple subprocesses and is a very sophisticated procedure with much attention to detail (Warren et al., 2021) . Most current practices in industry involve experts inspecting individually each piece produced and detecting defects manually (Jawahar et al., 2021) . Particularly in the aerospace manufacturing industry, visual inspection still dominates the testing of parts including engine blades, accounting for approximately 90% of all inspections (Aust et al., 2021) . With quality assessment being one of the essential steps of the process, relying solely on the ability of an inspector to detect faults could be of high risk (Aust et al., 2021) . Studies have shown that during the inspection of parts, the judgment of professionals can be biased by expectations coming from contexts such as prior knowledge or experience and inspectors may be unaware when their judgments are affected (MacLean & Dror, 2021) . Bias can come from different

FAQs

What is the main purpose of introducing the ML approach?

To improve the reliability of the inspector’s decision-making mechanism, reduce the risk of false assessments, minimize bias, and increase objectivity during microscopic examination.

Aug 4, 2026

Upload by Quinn