RDIS: Web Application to Detect Rubber Manufacturing Process Defects

Abstract

The rubber manufacturing industry plays an important role in the Sri Lankan economy. And it is one of Sri Lanka's five main subsectors by GVA (Gross Value Added). It is contributing in various ways to the growth of the economy of Sri Lanka. In Sri Lanka as a country, they are manufacturing top-quality ribbed smoked sheets(rss)within the minimum cost range. But, the rubber product manufacturing process, commonly faced various challenges. Rss related problems are one of the major challenges that occur in that manufacturing process. Due to technical limitations rss sheets can be overheated, also rss, are faced with coagulation problems. Usage of chemicals is another reason for ribbed smoked sheet-related problems. Those problems are mainly occurring reduce of grade quality. Presently, rubber sheets manufacturers are identifying rss related defects by using a version-based approach. Hence, RSS makers are facing different types of problems such as rss buyers are not giving proper prices, buyers nominating invalid grades, etc. For such a problems minimization, a Proper rss defect identification system is essential for the industry. As minimizing those problems, a web application called RDIS has been designed. The rubber sector workforce (processors, graders, buyers, other rubber product manufacturers) can easily use this web application and they can get their sheets defects by wasting minimum time. The system (RDIS) is mainly developed by using image processing techniques. During the implementation in the model development phase, have various steps. The steps are to collect data from the rubber research institute located in Ratmalana, Crop images by using python, create a dataset, Color images convert into wavelet format, classify images by using their color frequencies, and model training by using a neural network. The model will train only to identify two defects namely reeper marks and mould growth. Displaying results for defects, created the Html based web interface. As the result, image processing by using frequency based classification and neural network related model training can be the most suitable way to detect defects