Aksoy Aslı, Haki Ömer
Engineering Faculty, Industrial Engineering Department, Bursa Uludag University, Gorukle Kampus, 16059, Bursa, Turkey.
Bosch San. Ve Tic. A.S., Organize San. Bol. Yesil Cd. No:27, 16140, Bursa, Turkey.
Sci Rep. 2025 May 7;15(1):15890. doi: 10.1038/s41598-025-99488-6.
The high-pressure pulsation test (HPPT) bench is used to test the functionality and resilience of components under high pressure and pulsation. In highly automated machining systems, it is vital to reduce the number of unplanned machine downtimes due to equipment failure, as these can lead to significant losses in resources. The objective of this study is to enhance the efficiency of HPPT benches by addressing specimen, bench, and test environment- based problems and to develop a failure condition tracking tool (FCTT) by using machine learning (ML) algorithms. The findings of this study provide a basis for the development of the company's data-driven smart predictive maintenance applications while providing an increase in the operational efficiency of HPPT benches. The data set used in the study was obtained from the HPPT benches of an automotive parts manufacturing company. Decision tree (DT), gradient boosting tree (GBT), Naïve Bayes (NB), and random forest (RF) algorithms are used to determine the best model. The comparative analysis of ML algorithms revealed that the GBT algorithm exhibits superior predictive capabilities regarding HPPT bench failure predictions. The FCTT is developed using the results of the GBT algorithm and integrated into the company's HPPT bench maintenance system. The results of this study are described as a fundamental step in the company's smart maintenance programme. Implementing FCTT has resulted in a 20% increase in HPPT utilization, a reduction in maintenance costs, and a positive contribution to the company's overall competitiveness and profitability. The utilization of FCTT has enabled the prediction of HPPT failures, the optimization of maintenance schedules, the minimization of downtime, and the improvement of maintenance practices. Furthermore, using ML technologies provides valuable insights into the performance and maintenance trends of the HPPT bench, enabling data-driven decision-making and strategic planning for the company's HPPT bench maintenance operations.
高压脉动测试（HPPT）台用于测试部件在高压和脉动情况下的功能及韧性。在高度自动化的加工系统中，减少因设备故障导致的意外停机次数至关重要，因为这些停机可能会造成大量资源损失。本研究的目的是通过解决基于试样、测试台和测试环境的问题来提高HPPT台的效率，并利用机器学习（ML）算法开发一种故障状态跟踪工具（FCTT）。本研究结果为公司基于数据驱动的智能预测性维护应用的开发提供了基础，同时提高了HPPT台的运行效率。该研究中使用的数据集来自一家汽车零部件制造公司的HPPT台。决策树（DT）、梯度提升树（GBT）、朴素贝叶斯（NB）和随机森林（RF）算法用于确定最佳模型。ML算法的对比分析表明，GBT算法在HPPT台故障预测方面具有卓越的预测能力。FCTT是利用GBT算法的结果开发的，并集成到公司的HPPT台维护系统中。本研究结果被描述为公司智能维护计划中的一个基本步骤。实施FCTT使HPPT的利用率提高了20%，降低了维护成本，并对公司的整体竞争力和盈利能力做出了积极贡献。FCTT的使用能够预测HPPT故障、优化维护计划、将停机时间降至最低并改进维护实践。此外，使用ML技术为HPPT台的性能和维护趋势提供了有价值的见解，从而为公司的HPPT台维护运营实现数据驱动的决策制定和战略规划。
