Topic

Smart manufacturing

In 2017, EVS embarked on a cutting-edge venture using deep learning for marble quality classification in the stone processing industry. This project was centred around the utilization of deep learning techniques to automate and enhance the marble quality classification process during production, thereby boosting processing efficiency significantly.

Problem

Process Bottlenecks and the Challenges of Manual Assessment

The procedure of cutting marble slabs involves several stages. Although automation has been introduced in some of these stages, the manual unloading and quality assessment of produced parts remain a significant challenge. Operators are required to assign quality grades to the marble, a decision that depends on factors such as its intended use, specific market standards, and client specifications.

Expert operators assess quality based on characteristics like veining, colour, defects, and overall texture. This manual process, however, introduces potential for errors due to subjective judgment and the natural decrease in attention over time. Moreover, this manual involvement forms the primary bottleneck in the workflow.

Solution

Advancing with Deep Learning-Based Image Classification

To address these challenges, EVS proposed an at the time innovative solution: integrating a deep learning-based image classification stage into the existing process. A high-resolution camera captures the picture of the marble slabs, which is then geometrically rectified. The algorithm segments the slab to precisely outline it, and then divides it into smaller patches. These patches are analysed by a convolutional neural network, trained to predict the quality of the marble, categorized into four distinct classes: First, Second, Commercial, Defective.

Over time, this system is trained based on the operator’s initial assessments until it is capable of autonomously replicating and even surpassing human precision. Our approach utilizes deep learning for marble quality classification through high-resolution imagery and advanced neural networks to enhance precision.

Added value

Enhancing Productivity, Precision, and Operator Safety

Implementing deep learning for marble quality classification not only boosts productivity but also significantly improves accuracy in assessing marble quality. The implementation of this system offers multiple benefits:

  • It significantly reduces manual labour time, leading to a marked increase in productivity.
  • It ensures a more objective, repeatable, and precise classification.
  • It allows for system customization, tailored to align with the producer’s unique standards.
  • It minimizes the operator’s task load, thus improving overall working conditions.
  • It lays the groundwork for further automation in subsequent processes like marking, unloading, and automatic palletization, boosting productivity and minimizing injury risks for the workforce.
Project partner

This project was accomplished through a collaboration between the University of Verona’s Department of Computer Science, and Donatoni Macchine Srl. It was funded by the European Social Fund.

Donatoni Macchine Srl is a company operating in the field of constructing CNC machines for processing marble, granite, and stone materials. With over 50 years of experience in the sector, it has become a leader in the construction of cutting-edge machinery for stone processing. Donatoni’s constant dedication to research and technological innovation has enabled the creation of highly versatile machinery, suitable for any type of processing, to meet customer needs.