journal article·2025-10-27

Predictive maintenance for printed circuit boards using eXtreme gradient boosting

Chien-Yi Huang, Hao-Chun Hsieh, Yan-Cheng Li

Microelectronics International· Volume 42· Issue 3-4
Publisher page

Abstract

Purpose This study investigates the impact of reflow temperature on the reliability of printed circuit board assembly (PCBA) produced through surface mount technology (SMT). The authors define failure as deviations of the temperature curve from the process window (PW) of key PCB components. This study aims to develop a prognostic and health management system for failure prediction. Design/methodology/approach The study used reflow equipment in a real-world production environment. Key parameters affecting temperature curve deviations were identified, and the eXtreme Gradient Boosting (XGBoost) method was used to construct a failure prediction model. This model allows onsite monitoring and predicts failures based on various feature combinations. Upon detecting potential failures, the model alerts the on-duty engineer, providing failure time and recommending rechecks. Findings The failure prediction system achieved high accuracy (94%), precision (99%) and recall (89%). The PHM system effectively identifies temperature curve deviations, enabling timely interventions. It calculates the PCBA product serial number in the reflow furnace zone, assesses deviations from the PW and recommends rechecks, thus enhancing PCBA production reliability. Originality/value This study integrates the XGBoost method into a PHM system for failure prediction in PCBA production. Combining real-world production data with advanced machine-learning techniques offers a novel approach to addressing reliability concerns in the SMT reflow process. Integrating failure alerts into the Shop Floor system ensures prompt rechecks of high-risk PCBs, enhancing overall system efficiency.
Source: crossref · Quality: baseline

Research context

OpenAlex topics: Industrial Vision Systems and Defect Detection · Non-Destructive Testing Techniques · Integrated Circuits and Semiconductor Failure Analysis

SciVX discovery tags are terminology-based. OpenAlex topics are source-provided classifications.

Affiliations

National Taipei University of Technology Department of Industrial Engineering and Management, , Taipei City,

Wistron Corp , Taipei City,

National Taipei University of Technology , Taipei City,

References (17)

Show source-provided references
1. Improvement on QFN leadframe design of extended leads to support the mitigation of mold flash occurrence10.9734/jerr/2021/v20i517309 ↗
2. Developing a machine vision inspection system for electronics failure analysis10.1109/TCPMT.2019.2924482 ↗
3. Applying data mining methodology to establish an intelligent decision system for PCBA process10.1108/SSMT-10-2018-0036 ↗
4. Illés · 2020
5. Reflow soldering process optimization based on surface evolver solder joint shape simulation and finite element analysis of PCB assembly
6. Solder fatigue life modeling of QFN components based on design of experiments10.1016/j.microrel.2023.115297 ↗
7. An intelligent system for reflow oven temperature settings based on hybrid physics-machine learning model10.1108/SSMT-10-2021-0063 ↗
8. High pressure heater fault monitoring and early warning based on PCA-XGBoost
9. Mechanical reliability of self-aligned chip assembly after reflow soldering process10.1108/SSMT-12-2019-0042 ↗
10. Learning memory-guided normality for anomaly detection
11. Enhanced die attach process defect recognition on QFN leadframe packages10.9734/jerr/2021/v20i317286 ↗
12. Experimental study on thermal fatigue life prediction model of PBGA solder joints10.1109/PHM-Qingdao46334.2019.8942918 ↗
13. Fault prediction of marine diesel engine based on time series and support vector machine
14. A hybrid ensemble method for pulsar candidate classification10.1007/s10509-019-3602-4 ↗
15. Critical threshold limit for effective solder void size reduction by vacuum reflow process for power electronics packaging10.1109/TCPMT.2023.3293114 ↗
16. Prognostics and health management (PHM): where are we and where do we (need to) go in theory and practice10.1016/j.ress.2021.108119 ↗
17. Prognostics based on stochastic degradation process: the last exit time perspective10.1109/TR.2021.3075213 ↗