For industries around the world that have adopted manufacturing quality management, finding steel surface flaws is one of the toughest difficulties. The conventional method of inspection of the steel's surface detection takes a long time. Steel faults detection is possible via image analysis through machine learning techniques, which could further helps to save the time of fault detection and enhances the steel fault detection process. This further helps to adopt the machine learning techniques in steel industries. The objective of this endeavour is to create models that can detect steel defects. Using the SEVERSTAL dataset, which includes 12,568 photographs of the steel surface, the models were created for binary categorization (defect and no defect). These models were then evaluated for multiclass classification. In this research, image pre-processing is used to enhance the detection of steel faults.