Bogor – A real-time sensor-based digital twin has demonstrated remarkable capability in improving predictive maintenance for vehicle gearbox systems by continuously monitoring component health and detecting potential failures before they occur. The study was conducted by Stiven Alfonso Siregar, I Nengah Putra Aprianto, Sasono Rahardjo, and Yunia Amelia Chairunisa from the Faculty of Defense Engineering and Technology, Indonesia Defense University, and was published in the International Journal of Scientific Multidisciplinary Research in July 2026. The research introduces an intelligent maintenance approach that enables maintenance decisions to be based on the actual condition of vehicle components rather than fixed maintenance schedules.
Maintaining the reliability of vehicle transmission systems is essential for ensuring operational readiness, particularly in applications where vehicle availability is critical. Conventional maintenance strategies generally rely on scheduled inspections based on operating time or mileage. While widely used, these methods often result in unnecessary maintenance or delayed repairs because they do not reflect the real condition of the gearbox. Such limitations increase maintenance costs, extend downtime, and reduce overall operational efficiency.
To overcome these challenges, the research team developed a digital twin capable of creating a virtual representation of a vehicle gearbox using real-time operational data. The system integrates vibration sensors, Internet of Things (IoT) communication, cloud-based monitoring, and artificial intelligence algorithms to continuously evaluate gearbox health, detect abnormalities, and estimate the remaining useful life of critical components. By relying on actual operating conditions, the proposed system enables maintenance activities to be performed precisely when needed rather than according to predetermined schedules.
The study employed a quantitative case study approach using a military vehicle gearbox as the primary research object. Three ADXL345 accelerometer sensors were installed at the gearbox input shaft, gear mesh, and output shaft to capture vibration data under different operating conditions. The collected data were transmitted in real time through an ESP32 microcontroller using the MQTT communication protocol to a cloud database. Machine learning techniques were then applied, with the Random Forest algorithm classifying gearbox conditions and the Gated Recurrent Unit (GRU) neural network predicting component health scores and estimating the Remaining Useful Life (RUL) of the gearbox.
Experimental results demonstrated that the digital twin successfully monitored gearbox performance under three different operating scenarios: when the vehicle was turned off and stationary, turned on but stationary, and operating while moving. The vibration signals collected from the three sensors changed according to operating conditions, allowing the system to automatically distinguish between normal, moderate, high-risk, and critical conditions. As operating loads increased, vibration levels also increased, enabling early identification of abnormal mechanical behavior before severe failures occurred.
One of the most significant findings was the outstanding performance of the machine learning models. The Random Forest algorithm achieved a gearbox condition classification accuracy of 98.20 percent, while the GRU neural network produced an R² value of 0.9248, indicating excellent capability in predicting gearbox health scores from real-time vibration data. These results demonstrate that combining intelligent algorithms with continuous sensor monitoring can provide highly accurate diagnostics for gearbox condition assessment.
Beyond predictive accuracy, the research successfully developed a comprehensive five-layer digital twin architecture consisting of the physical layer, data acquisition layer, communication layer, virtual model layer, and application or decision layer. Through this architecture, vibration data are transmitted automatically to a web-based monitoring dashboard displaying gearbox condition, health score, risk level, anomaly detection, and Remaining Useful Life estimation. The architecture diagram presented on page 11 illustrates how sensor data flow seamlessly from the physical gearbox to the intelligent decision-support system.
According to Stiven Alfonso Siregar and colleagues from Indonesia Defense University, the developed digital twin represents a significant step toward predictive maintenance by transforming maintenance practices from reactive and schedule-based approaches into condition-based maintenance. Instead of waiting for failures to occur, maintenance personnel can monitor gearbox health continuously and receive early warnings whenever abnormal vibration patterns emerge. This capability allows repairs to be scheduled proactively, minimizing unexpected failures and reducing maintenance downtime.
The study also highlights broader implications beyond military vehicle applications. The proposed digital twin framework can be adapted for commercial vehicles, heavy machinery, industrial manufacturing equipment, and public transportation systems that rely on gearbox or transmission components. Integrating real-time sensors, IoT communication, and artificial intelligence into maintenance systems has the potential to reduce operational costs, improve equipment reliability, and support smarter asset management. As industries continue adopting digital transformation strategies, predictive maintenance technologies such as this digital twin model are expected to play an increasingly important role in improving operational efficiency and long-term sustainability.
Author Profile
Stiven Alfonso Siregar – Faculty of Defense Engineering and Technology, Indonesia Defense University
I Nengah Putra Aprianto – Faculty of Defense Engineering and Technology, Indonesia Defense University
Sasono Rahardjo – Faculty of Defense Engineering and Technology, Indonesia Defense University
Yunia Amelia Chairunisa – Faculty of Defense Engineering and Technology, Indonesia Defense University
Research Source
Article Title: Development of a Real-Time Sensor-Based Digital Twin for Predictive Maintenance of Vehicle Gearbox Systems
Journal: International Journal of Scientific Multidisciplinary Research (IJSMR), Vol. 4, No. 7, 2026
DOI: https://doi.org/10.55927/ijsmr.v4i7.93
Journal Link: https://journalijsmr.my.id/index.php/ijsmr
0 Komentar