This project tackles a critical issue in the packaging and delivery industry, focusing on structural behaviors of corrugated paper boxes. Although these boxes are widely used because of their lightweight, recyclable, and customizable properties, they are prone to buckling during storage and shipment, resulting in product damage, lost revenue, wasted resources, and customer complaints.
Currently, the industry relies on the Dr. Box Calculator Pro, which employs Finite Element Analysis (FEA) to predict box performance under various conditions. While effective, FEA is time-consuming and demands significant computational resources. To address these limitations, we propose integrating Deep Neural Networks (DNN) into the Dr. Box Calculator Pro to replace part of the FEA calculation. DNN's ability to model complex non-linear relationships through multiple layers offers advantages such as automatic feature learning, scalability, and end-to-end learning, making them well-suited for big data applications.
Our solution aims to provide instant predictions of buckling strength, reduce computational costs, and efficiently handle diverse box types and conditions. The expected outcome is enhanced prediction accuracy, faster analysis, and overall improved efficiency in packaging and logistics operations. The primary challenge is achieving accurate and efficient predictions of buckling strength, with specifications including high predictive accuracy ($\geq$ 90\%), fast analysis rate ($\leq$ 5 minutes per simulation), the ability to handle various box configurations, and system reliability above 99.9\% in practice.
The project plan is methodically structured. In the initial two weeks, the project will commence with requirement analysis, followed by two weeks of data preparation and cleansing. The subsequent four weeks will focus on developing and initially training the DNN model. The final five weeks will involve optimizing the DNN model, integrating it with the Dr. Box Calculator Pro, conducting system testing, and validating it against traditional FEA methods.
The integration of DNNs is anticipated to revolutionize the packaging industry by offering faster, more accurate, and scalable predictions of buckling strength. This advancement will reduce operational costs and enhance customer satisfaction by minimizing product damage during transit. The motivation behind this project is to apply modern machine learning techniques to solve long-standing industry problems, improving the efficiency and effectiveness of packaging and logistics operations.
