NewsHow to Achieve Deep Learning Automation in Smart Factories?

How to Achieve Deep Learning Automation in Smart Factories?

Category articles

It is a principle that lies at the heart of manufacturing in industrial settings. Since the Second Industrial Revolution to the rise of Lean Manufacturing, industry giants have improved production processes. They have realized that eliminating inefficiency and waste from the process allows them to manufacture more components with little change else. This is the basis for the Toyota Production System, which seeks to deliver “Just-in-Time” production while minimizing stock and batching. The concept of process is based on the notion that even though the Ideal State of 100 % efficiency is not possible in the real world but the effort to achieve and progress toward an Ideal State is worthwhile. Making steps to minimize mistakes, eliminate redundant and unnecessary activities, and boost production can help meet customer demands while reducing the production cost.

What is Deep Learning

Robotics, a fairly modern technology that has its roots going back to the 1920s industrial age and was further developed over the course of time. For instance, Westinghouse Electric Corporation had a robot known as Televox that could recognize human voices and complete basic tasks. Nowadays, far more complex and dangerous tasks rely on robots.

The manufacturing industry is accelerating through Industry 4.0 by introducing new techniques and methods that are as transformative as those of the past. With connected devices growing their IoT along with 5G rolling to a broader range of coverage, deep learning makes use of these technologies in conjunction using robotics and artificial Intelligence (AI) for bringing these to manufacturing.

A paper from The Journal of Manufacturing Systems defines deep learning as the process by which engineers make use of comprehensive analytical tools to process and analyze large amounts of data. That is, the machines learn what users want to do and when they can alter. This process results in “smart” manufacturing by incorporating advanced analytics to improve decisions and opportunities for system performance. These enhancements significantly expand the manufacturing process beyond the previous models. They assist in moving further towards an Ideal State of just-in-time manufacturing that is defined by Toyota.

Deep learning can be applied to specific applications such as the predictive maintenance of equipment, and analytics that provide predictive steps for improvements in process and product development (to forecast the manufacturing effects of design choices) and quality assurance and logistical supply chain management. The benefits of this approach are significantly more than its components; they open up innovative areas for efficiency, while providing more flexibility and adaptability to the products and processes. Technology also assists with disruptions in supply chains that are discovered and further exacerbated due to the pandemic. These include the utilization of capacity and manufacturing shortages.

How Manufacturing Sites Incorporate Deep Learning

With the concept of deep learning being established the main problem is how to incorporate this transformational approach to manufacturing plants to boost performance. Since there are significantly lower manual steps, incorporating deep learning could require a change in the concept and process of the layout of the plant through several layers of integration

  • Horizontal integration using operational systems
  • Connected manufacturing and vertical integration
  • Holistic integration throughout the entire value chain

The resultant style of work is often called “smart factories.” With these pieces of equipment installed, the machine will automatically transmit the output of one process to be the source of the subsequent. The reduction in processing change time and transition time is the key distinction.

Leading companies around the world are already using this revolutionary technology to improve their strategic position like Whirlpool, Siemens, Hirotec and Hewlett-Packard. It’s not an easy transition and will require companies to invest the training, capital, and assistance to make the massive change. It’s a good one, but McKinsey recommends that applications of deep learning for supply chains and manufacturing could result in an increase in annual revenue of between $1 and 2 trillion.

While the financial and psychological investment are crucial however, the technology utilizes and improves processes already in place and optimizes them after the deployment. Engineers can gather data by incorporating sensors that collect current-state information. They establish the standard for determining the areas to enhance the process. The more precise the grid of information collected and the more accurate the predictive models will be able to analyze the performance and suggest improvements.

A case study of the Westinghouse Televox robot. The robot today can perform more complicated tasks thanks to technology for predictive analysis. In addition, with the use to deep learning the robot can also learn to perform new tasks, taking the process of optimization to a new level. Additionally, the enhancements in the robot’s process can eliminate a manual task leaving the operator to focus on a specific task. This is just one example of how deep learning can improve production processes, enhancing accuracy and decreasing errors while enhancing the use of human capital.


Industry 4.0 is underway and technologies like IoT 5G, IoT, as well as AI are accelerating the pace of technological advancements in the field of manufacturing. With deep learning as a key component manufacturing innovations, manufacturers will join technological advancement to propel the industry like never before.

Michal Pukala
Electronics and Telecommunications engineer with Electro-energetics Master degree graduation. Lightning designer experienced engineer. Currently working in IT industry.