Image credit: Lutsina Tatiana/Shutterstock.com
In the past, when people imagined 2020, they always included a lot of sci-fi smart devices, such as robots providing services at home, self-driving cars and planes on the road and even in the sky, and virtual reality for audio-visual entertainment that people can immerse themselves in, etc. Although such a life will not come in 2020 due to the impact of the COVID-19 epidemic, Industry 4.0, which has this vision, has not stopped.
As early as 2013, with the maturity of Internet and computer technology and the gradual improvement of related infrastructure, Germany took the lead in proposing the concept of "Industry 4.0", which is a new generation of revolution that uses cyber-physical systems to improve people's lives in all aspects. This concept was then written into the development plans of many countries, aiming to create new growth points based on traditional industrial technology and service industries by combining industrialization with informatization.
The development and popularization of technologies related to Industry 4.0 are also in full swing. In terms of software, augmented reality technology can bring a new audio-visual experience and has been applied to the training of special professions (such as police and doctors); the Internet of Things technology uses sensor clusters to achieve all-round monitoring of electrical appliances; the progress of industrial network security technology can monitor in time and avoid hacker attacks on corporate networks. In terms of hardware, 3D printing technology allows ordinary people to quickly manufacture any design; the popularization of industrial robots will make product manufacturing more standardized and efficient.
Typical scenarios of Industry 4.0
Data and machine learning technology closely integrated with data are the core of Industry 4.0.
Data is obtained from sensors, transmitted to cloud computing servers via the Internet, and then analyzed using machine learning and artificial intelligence algorithms. The results are returned to service terminals or industrial robots to complete the entire workflow.
Typical scenarios of Industry 4.0 include understanding of users, product manufacturing, product quality monitoring, product distribution logistics, and user feedback. Data and machine learning are widely involved in each part.
1
User portrait
Many mobile phone and computer software are already storing and analyzing user data. Some physical stores also use radio frequency identification chips (RFID) to record user preferences and use recommendation algorithms and other methods to analyze user data to recommend and update products and content.
In Industry 4.0, user data will be recorded in all aspects, such as user usage frequency, preferences, methods, and time periods. The recording media ranges from mobile phone apps to household appliances, from office supplies to medical devices.
After the data is analyzed by machine learning algorithms, multi-dimensional classification labels can be predicted. Each user will be described by multiple labels, thereby achieving increasingly accurate user portraits.
2
Manufacturing Process
User portraits will bring very direct benefits, namely, the improvement of production personalization.
Similar to the personalization of browsing content on the Internet today, in the era of Industry 4.0, these detailed user portraits will be directly applied to the product manufacturing process, and merchants will also be able to more easily produce personalized products that meet user needs.
Some personalized products can be predicted based on user data, which invisibly provides users with more possibilities.
Not only will it affect production decisions, but the control of different steps in the manufacturing process will also be fully automated with the help of the Internet of Everything and industrial robots in Industry 4.0. Each step in the production process will comprehensively analyze the status of the previous steps and product requirements, and make fine adjustments in time. In such a smart factory, the controllability and robustness of the production line are improved, and the participation of workers is changed from repetitive labor to supervision of robots. Tesla, an electric car company, has been committed to building smart factories since its inception. Not only is the assembly work on the production line completed by industrial robots, but the warehousing, material management, order and sales links are also highly intelligent, which makes this car company stand out in terms of technology content and sales in the entire industry.
3.
Quality Control
In addition to the analysis and control of process data, the combination of machine learning and machine vision technology can automatically complete large-scale, high-precision product inspections, which is especially effective for complex defects that are difficult for the human eye to distinguish. Landing.AI, an artificial intelligence algorithm company led by Professor Andrew Ng, a famous artificial intelligence scientist, recently launched a bubble detection device based on artificial intelligence and machine vision to detect gas leaks in equipment. Through the machine vision system, the computer can capture tiny bubbles very accurately and then determine the location of the gas leak. Its recognition error rate is much lower than the 30% average error rate of workers' naked eye recognition. Combined with the data of the entire production process, it can not only quickly locate the location and production line of the problem, but also greatly reduce labor costs and recognition error rates.
4Fast
logistics
生产过程的最后，还会为物流做出准备。
工业机器人可以对产品进行自动打包，并在包装上打印包括产品信息、邮寄地址等特异性的二维码标识，为产品分发做准备。
分发过程中，自动驾驶系统会发挥很大作用。
预计在未来十至十五年内，以计算机视觉、机器学习、控制技术等为基础的自动驾驶技术可以实现全面商用，这样会使物流的传递更加简单高效，而且可以显著降低人力成本。
电
商巨头阿里巴巴的首批智慧机器人仓库已于2017年投入使用，其旗下的菜鸟物流已逐步实现刷脸取件、无人机派件等技术。
2019年末，菜鸟物流估值已达2000亿人民币，在未来物联网和自动驾驶技术的加持下，未来的“包裹找人”替代“人找包裹”指日可待。
5Service
and Feedback
On the user side, the data uploaded by the product's sensor system can be analyzed by the machine learning algorithm in the cloud, which can determine whether there are any anomalies in the usage data, thereby achieving real-time supervision of the product's performance.
When users encounter usage problems, the trained artificial intelligence system can efficiently handle tasks such as text chats, answering calls, and video calls, so that problems can be quickly fed back and solved in a timely manner.
The BERT model published in 2018 has surpassed humans in the field of chatbots, and related applications have already occupied a place in products of Internet giants (such as Microsoft XiaoIce, Ali Xiaomi, IBM Watson, etc.) and emerging AI companies (such as 4Paradigm, iFlytek, etc.).
Characteristics of Industry 4.0
From the above applications, the characteristics of Industry 4.0 are demonstrated, which can be summarized as follows:
1Integration
and Interconnection
In Industry 3.0, people around the world can quickly connect with each other through the medium of the Internet.
In Industry 4.0, sensors are integrated into each hardware to enable communication between machines.
For example, in the printing and dyeing industry, the management system serves as the center of the production system, coordinating the mother liquor configuration, dye positioning, automatic dripping, automatic water supply, proofing system, etc. on the entire assembly line to achieve intelligent dyeing, greatly improving production efficiency and product stability.
Coupled with the machine learning engine provided by the cyber-physical system and cloud computing, the Internet of Everything can be truly realized, that is, seamless connection between people, people and machines, machines and machines, and services and services.
When "interconnection" becomes the norm, all steps from production to service, that is, equipment, production lines, factories, services, etc. can be closely linked.
2.
Data and Digitalization
In Industry 4.0, the introduction of information technology has made data the blood of industrial production.
These data include all aspects of production and services, including product data, equipment data, R&D data, supply chain data, operational data, user data, etc.
On the one hand, data is of decisive significance for training and optimizing machine learning algorithms. On the other hand, after the machine learning algorithm is deployed, the algorithm also needs to control the production process by processing the newly generated data.
This means that all aspects of life and production processes should be digitized as much as possible, that is, everything can be quantified with reasonable indicators, otherwise it cannot be embedded in the automation system.
This requires data scientists to
fully
consciously guide the system to collect appropriate data, and design reasonable indicators.
3.
Refinement and personalization
In Industry 4.0, since the requirements for data flow will be relatively detailed, the various modules in production will become more and more refined accordingly.
The increasing modularization and detail of each part of the production line will make personalized production possible, better reflect and predict user needs, and enable the product's "production-sales-feedback" cycle to enter a virtuous development.
Opportunities and Challenges of Industry 4.0
Industry 4.0 brings many new opportunities. Although the entire production process can be integrated into one, the workload in the data flow may be distributed to multiple departments or even multiple companies. Therefore, a small company's single-point breakthrough in the entire process will become more and more valuable. The same smart device can be split into multiple models such as classification, segmentation, trend prediction, as well as multiple modules such as data transmission system, data acquisition equipment, data feedback system, etc. Each module can be embedded in other production processes. For example, data acquisition equipment can share production lines with other precision instrument production processes. Therefore, different small companies can rely on Industry 4.0 and use their own strengths to embed themselves in different dimensions of the market.
According to current forecasts, infrastructure construction will be a hot industry in the next few years. Whether it is the friction between China and the United States over 5G technology from last year to this year, or the cloud computing platforms that various Internet companies have established in recent years, they all show the importance of infrastructure in Industry 4.0 to ensure corporate profits and national security. In addition, data is another form of infrastructure. Internet giants with a large amount of data will occupy more opportunities, but small companies can also devote themselves to finding parts of production lines and life that have not yet been digitized. Such opportunities are particularly prominent in industries that are not digitized (such as traditional heavy industry) and industries where data has not been well used (such as medical care).
Industry 4.0's demand for data and machine learning technology is also a challenge that traditional large companies are facing. At this stage, large companies mainly rely on large-scale industrial production, so adding sensor systems and Internet of Things systems to production lines requires relatively high costs. Large companies also need to introduce machine learning technology into production lines and product design, which requires the input of talents and innovation in management methods. The popularity of machine learning technology in recent years has caused companies to fall into an obsession with "intelligence" when making decisions, which has also posed new challenges to decision makers' ability to discern.
As Industry 4.0 gradually enters people's lives, many new applications that are still unknown will gradually grow. When the Internet of Things can really enter thousands of households, and when autonomous driving is deployed on a large scale, humans will be freed from the current large amount of repetitive labor. So under the influence of Industry 4.0, will future careers be concentrated in the computer industry or the data analysis industry? Will people have more free time waiting to be filled? How will the relationship between people and people, and between people and machines develop? We are about to enter the third decade of the 21st century. Although these questions are still difficult for humans to answer, it is certain that Industry 4.0 will be a major theme of future development.
Author: Wang Dongang
Wang Dongang is a PhD candidate at the University of Sydney. His research covers medical imaging, artificial intelligence, neuroscience, video analysis and other fields, and he is committed to applying artificial intelligence technology in practical systems. He has published papers in international conferences such as CVPR and ECCV, and has been invited to review papers for academic journals such as IEEE Transactions on Circuits and Systems for Video Technology, IEEE Transactions on Multimedia, and international conferences such as ICML and AAAI. He has more than 5 years of development experience in machine learning and computer vision, and has cooperated with many companies and institutions in China, the United States, and Australia to develop projects, including behavior recognition in multi-angle videos, road condition prediction based on road monitoring, and automated brain CT screening systems.
京公网安备 11010802033920号