Application of artificial intelligence in manufacturing

With the advent of the smart manufacturing boom, artificial intelligence applications have penetrated into all aspects of the manufacturing industry such as design, production, management and services.

The concept of artificial intelligence was first proposed in the 1950s, more than sixty years ago. However, it is not until recent years that artificial intelligence has experienced explosive growth. The reason is mainly due to the increasingly mature Internet of Things, big data, cloud computing and other technologies.

The Internet of Things enables a large amount of data to be obtained in real time. Big data provides data resources and algorithm support for deep learning, and cloud computing provides flexible computing resources for artificial intelligence. The organic combination of these technologies drives the continuous development of artificial intelligence technology and has made substantial progress. The human-machine battle between AlphaGo and Lee Sedol pushed artificial intelligence to the forefront and set off a new round of artificial intelligence craze.

In recent years, research and applications on artificial intelligence have begun to bloom everywhere. With the advent of the smart manufacturing boom, artificial intelligence applications have penetrated all aspects of the manufacturing industry such as design, production, management and services.

PART ONE

##Product defect detection

Due to the application of deep learning, the defect detection process of manufacturing production lines is becoming increasingly intelligent. Deep neural network integration allows computer systems to identify surface defects such as scratches, cracks, leaks, and more.

This process is accomplished by data scientists training the visual inspection system on a given defect detection task by applying image classification, object detection and instance segmentation algorithms. Deep learning-driven detection systems, combined with high optical resolution cameras and GPUs, form perception capabilities beyond traditional machine vision.

For example, the AI-based visual inspection program built by Coca-Cola can already diagnose facility systems and detect production line problems, and promptly feed back detected problems to technical experts for resolution. Based on this, quality inspection personnel were listed by Kai-fu Lee as a job type that will be replaced by artificial intelligence in the future.

New detection technologies include synthetic data, transfer learning and self-supervised learning. In the synthetic data, Generative Adversarial Networks (GAN) data generation tools examine images that quality inspectors consider “normal” and synthesize defective images for training artificial intelligence models. At the same time, transfer learning and self-supervised learning are used to solve specific problems. As data accumulates, defect detection algorithms become more precise.

PART TWO

##Intelligent sorting

There are many operations that require sorting in the manufacturing industry. Manual operations are slow and costly, and a suitable working temperature environment needs to be provided. If industrial robots are used for intelligent sorting, costs can be significantly reduced and speed increased.

Take sorting parts as an example. The parts that need to be sorted are often not neatly arranged. Although the robot has a camera to see the parts, it does not know how to successfully pick them up. In this case, using machine learning technology, first let the robot perform a random sorting action, and then tell it whether the action successfully picked up the parts or caught them empty. After many times of training, the robot will know how to sort the parts. Sorting in the order will have a higher success rate; which position to pick up when sorting will have a higher success rate; knowing in what order to sort will have a higher success rate. After a few hours of learning, the robot's sorting success rate can reach 90%, which is equivalent to that of skilled workers.

PART THREE

Warehouse Management and Logistics

For example, JD.com A certain logistics warehouse needs to sort finished products according to orders and shipping locations, while recycling empty boxes and throwing some waste materials and waste products into the waste dump. This work is completed by two workers in each shift. There is dust and noise in the warehouse, and the sorting operations are repeated 2,000-3,000 times every day. Although the heavy objects are handled by robots, it is still a high-intensity, poor environment and technically demanding task. Low-content repetitive work.

The company uses a robot to replace two workstations that work in three shifts a day. The robot is equipped with a machine vision system. It can scan RFID codes when sorting orders and shipping locations. The judgment of finished products, empty boxes, and waste materials is learned by AI. The algorithm gradually improved the recognition rate. The initial recognition rate was only about 62%, and each shift required a worker to fill in the gaps. As data accumulated, the AI ​​recognition model continued to improve. After 9 months, the comprehensive recognition rate increased to 96%. The identification of finished products and the sorting of delivery places are completely accurate. There is no need to keep people in the warehouse to fill the vacancies. Only a very small number of empty boxes can be picked out during waste recycling.

PART FOUR

##Manufacturing

Ford once boasted : No matter what car you want, I will only produce it in black. This is a typical portrayal of assembly line mass production, but if Ford continues to think this way in the current situation, Ford cars will only die. Because there is more and more personalization now, but the cost of personalized production is very huge, then the only way is mass customization, which uses personal consumption data to analyze and form comprehensive orders, and then the platform distributes it for mass production to reduce costs. The unit price of finished products is the path Rhino Manufacturing is currently taking. However, although e-commerce has a large amount of consumer behavior data, the data always lags behind actual demand. This application scenario requires the analysis platform to maximize the accuracy in order to increase the accuracy.

PART FIVE

##Remote Operation and Maintenance Service

Remote Operation The dimensional platform uses technologies such as the Internet of Things, big data, and artificial intelligence algorithms to monitor key parameters of the production process and production equipment in real time, and provide timely alarms for faults. Functions such as predictive maintenance and auxiliary decision-making supported by industrial big data analysis and artificial intelligence algorithms can further reduce personnel travel and shutdown delays caused by unplanned downtime, making the operation and maintenance of industrial enterprises less manned, unmanned, and more efficient. Remote model change.

Looking around the world, companies involved in the field of industrial artificial intelligence have already proven the unique value of this technology. Artificial intelligence technology has great potential in improving the productivity, efficiency, quality and cost of enterprises, and will undoubtedly become a new engine empowering the future manufacturing industry. However, enterprises’ AI transformation journey has a long way to go. Companies that are the first to awaken must strengthen their beliefs, practice their internal skills diligently, and set out immediately to expand their territory in the field of industrial artificial intelligence, striving to turn themselves into a beacon shining the light of future intelligent manufacturing.

The above is the detailed content of Application of artificial intelligence in manufacturing. For more information, please follow other related articles on the PHP Chinese website!