Xinsheng IntelligentIndustry NewsThe role of artificial intelligence in manufacturing transformation

Today, we have finally ushered in the fourth industrial revolution. The exponential growth of sensors, computing power, and network transmissions creates enormous power. From Germany's "Industry 4.0" to "Made in China 2025", the manufacturing industry has ushered in a new round of innovation.

The transformation of the manufacturing industry is a process from digitalization to automation and finally to intelligence.

  Digitizing

From the current point of view, the domestic industrial Internet of Things is in the early stage, both the network and the hardware equipment are not yet mature, and the infrastructure construction and data collection have not yet been completed. The factory first needs to install sensors and control devices for the production equipment, open up the production equipment, production management, manufacturing execution and planning systems, and control the production progress in a more real-time and transparent manner.

automation

Except for the automobile manufacturing industry, a large number of factories in China are still under-automated. In 2015, according to statistics released by The Economist, although China purchases the most robots in the world every year, the entire country is only equipped with 50 robots per 10,000 workers, while in Germany and Japan, where automation is relatively developed, the number is about 300. , and even as many as 500 in Korea. If production efficiency is to continue to be improved, automated production systems will surely become widespread in factories.

Intelligent

Historically, "automation" has meant that a machine can perform a specific, discrete task, such as turning a pump on and off according to defined rules. Automation is to replace people to do repetitive work, and intelligence is something that people can't do. A smart factory is defined as a flexible, collaborative system that runs the entire production process autonomously, optimizes itself on a global scale, and adapts to new environments in real time. It represents an ongoing adaptive process, rather than the "set and forget" upgrades of the past.

The main application of artificial intelligence in manufacturing

1. Big data analysis - equipment predictive maintenance

In traditional factories, production equipment still cannot be connected to the Internet, and only after the equipment fails to be repaired, or regular maintenance is adopted without considering the actual operation of the equipment. In the event of an unplanned downtime, it is necessary to temporarily purchase parts and spend a high cost for emergency repairs in order to resume normal production as soon as possible. Even if there is no downtime, when a human discovers that the machine is malfunctioning, it may have made substandard products, causing financial losses to the factory.

Uptake, an AI industrial forecasting platform in the United States, can collect various operational data of front-end equipment by placing sensors in the equipment of the factory, and combine big data analysis and machine learning technology to provide industrial customers with predictive diagnosis of equipment and management of energy efficiency optimization. Suggest. The factory can monitor the operating status in real time, compare historical data, predict potential equipment failures, and effectively avoid interruption of normal production.

If the data of equipment predictive maintenance is integrated into the ERP system in the future, the enterprise can optimize the production process and minimize the economic loss caused by equipment failure by dynamically adjusting the production plan. The integration and analysis of different data sources, production equipment and management systems will become a standard configuration for decision-making in future manufacturing companies.

2. Automatic Quality Control - Machine Vision Inspection

Before the development of deep neural networks, machine vision was already used in industrial automation systems such as pick-and-place, object tracking, metrology, defect detection, etc. Among them, nearly 80% of industrial vision systems focus on defect detection.

The human eye can also detect anomalies in products, even if such anomalies have never been seen before. However, because the eyes are easily fatigued, and human judgment is also very subjective, this will result in inconsistent product inspections or even missed inspections. It is also difficult for the human eye to adapt to the needs of high-speed production. For example, for printed circuit boards with complex graphics, manual inspection takes a long time. Usually, it can only be based on sampling inspections, and it is impossible to conduct real-time comprehensive inspections like automated systems. Currently on PCB and IC production lines, about 60% of inspection tasks are performed by machine vision.

Machine vision is more and more widely used in modern industry by virtue of its advantages of speed, precision and objectivity. On a production line, for example, automated inspection systems can inspect hundreds or thousands of components per minute. Equipped with cameras and optics of the right resolution, machines can also detect details that are invisible to the human eye. In addition, machine vision reduces the cost of component wear by eliminating direct contact between humans and the component being inspected, and also protects workers from hazardous environments.

But machine vision still faces the challenge of adapting to different industrial production environments, because few companies deploy automated inspection systems specifically for a certain type of product. In different environments, the direction of the camera's lens, the relative position to the component, and the strong reflected light on the component surface will affect the detection accuracy. Therefore, the vision algorithm itself must have strong adaptability.

3. Intelligent collaborative robots

Because the motion path of traditional robots is fixed, each action requires engineers to program, debug and manually configure to adapt to the specific production environment. Manual adjustments are useless when the robot has to deal with changing scenarios. Deep learning has brought about a revolution, giving robots "flexible" learning capabilities. Over time, robots can learn from data, switch between tasks autonomously, and import new tasks in minutes. Eventually these robots will not only be able to communicate with each other, but also work safely alongside humans, and even watch workers demonstrate production processes and automatically learn new skills.

At present, high-end industrial robots are mainly dominated by foreign companies. Switzerland's ABB, Germany's KUKA, Japan's Fanuc and Yaskawa Electric are collectively called the four major families, and they occupy more than 50% of the domestic robot market. The Big Four are also actively promoting intelligent collaboration. For example, KUKA robots have the ability to communicate with each other, and can cooperate and assign their respective production tasks according to the processes on the production line. Similarly, ABB has also launched the dual-arm robot YuMi (abbreviated from You and Me) for welding and assembly scenarios on 3C product lines. It can safely work side-by-side with the worker and automatically slows down or stops movement as soon as the arm in motion may touch the worker.

In addition to selling cobots directly, many companies are experimenting with new leases that allow the use of robots to be billed on the hour like a hired worker. Traditional robots are not safe and need to be isolated from workers, which not only cannot meet the plug-and-play scenario but also causes additional deployment costs. Robot as a Service (RaaS, Robot as a Service) is an emerging business model that lowers the initial payment threshold and places more emphasis on software and services other than hardware products. It can be repeatedly reprogrammed to complete new tasks, helping companies meet the production challenges of small batches and multiple orders.