(Maya Utami Dewi, Lukman Santoso, Agustinus Budi Santoso)
- Volume: 3,
Issue: 3,
Sitasi : 0
Abstrak:
In the era of Industry 4.0, integrating artificial intelligence (AI) and big data analytics in the industrial sector demands high-performance computing infrastructure to handle increasingly complex and voluminous datasets. This study investigates the optimization of AI performance by implementing a hybrid computing architecture, integrating CPUs, GPUs, FPGAs, and edge-cloud computing. The research aims to enhance processing speed, model accuracy, and energy efficiency, addressing the limitations of standalone computing systems. A quantitative methodology was employed, using over 1 TB of industrial data from IoT sensors and production logs. A hybrid architecture was implemented with dynamic workload scheduling to distribute tasks efficiently across computational components. Performance metrics included processing time, model accuracy, energy consumption, and cost analysis. Results demonstrated that hybrid architectures significantly improved performance: the CPU-GPU combination reduced processing times to 650 ms, increased model accuracy to 88.3%, and achieved an energy consumption of 2.1 kWh. Meanwhile, the CPU-FPGA configuration, while slightly less accurate (87.5%), proved more energy-efficient at 1.3 kWh. AI models developed using hybrid systems exhibited superior predictive accuracy, with Mean Squared Error (MSE) as low as 0.0248 and R² of 0.91. The study concludes that hybrid computing architecture is a transformative approach for optimizing AI systems in industrial applications, balancing speed, accuracy, and energy efficiency. These findings provide actionable insights for industries aiming to leverage advanced computing technologies for improved operational efficiency and sustainability. Future research should focus on advanced workload scheduling and cost-effectiveness strategies to maximize the potential of hybrid systems.