The plus of artificial intelligence in the semiconductor industry

Companies are increasingly turning to artificial intelligence to find useful solutions for their business. Cutting-edge technology quickly became a real opportunity in many production sectors. The adoption trend is consolidated, and the AI will be a key driver for development in the coming years. According to data from the dedicated observatory of the “Politecnico di Milano”, in 2020 the sector grew by 15%. The focus is mainly on AI to extract information from data. The Intelligent Data Processing covers 33% of the investments, while automating processes is 11% of the expenditure.

In the semiconductor industry, companies can greatly benefit from artificial intelligence and machine learning at every stage – from chip research and design to manufacturing and sales. AI will be increasingly important throughout the value chain. You can use artificial intelligence to develop demand and inventory optimization forecasts, gaining insights relevant to production, procurement, and operations and sales planning.

In semiconductor manufacturing, companies can use artificial intelligence to better adjust instrument parameters. By combining real-time data from the tool sensor with data from previous process steps and metrology readings, machine learning models can be developed to capture non-linear relationships between process time and results, such as engraving depth. Collected data may include values of electric currents in the etching process, light intensity in lithography and temperatures in firing. With these models, you can implement optimal process times on a wafer or batch basis to shorten processing times and improve yield, thereby reducing the cost of goods sold and increasing productivity.

Visual inspection of wafers, to check quality and detect defects, is often conducted during production, using cameras or scanning electron microscopes. The images are then manually evaluated by the operators. They are therefore subject to errors and arrears which increase costs. Modern wafer inspection systems, made possible by advances in deep learning for machine vision, can be trained to automatically detect and classify defects, with equal or better accuracy than human inspectors. By eliminating defects and out-of-tolerance process steps, companies can avoid time-consuming iterations, increase yield and reduce costs.

Finally, semiconductor companies can implement machine learning algorithms to identify patterns in component failures, predict likely malfunctions and failures in new designs, and propose optimal layouts to improve yield.