Author: Peng Zhao (Founder of Zhicifang, Co-Founding Partner of Cloud and Capital) Internet of Things Think Tank Original This is my 311th column.
In 2024, we are entering AIoT 2In the new stage of 0, a large number of mainstream devices will be intelligent, and the application of generative AI in the industry is an indispensable piece of the puzzle.
Recently, the application of generative artificial intelligence (GENAI) in the manufacturing industry is progressing imperceptibly.
Taking Siemens as an example, following last year's joint development of the "AI Industrial Co-Pilot" with Microsoft, this month Siemens and AWS joined forces to promote the popularization of generative AI in the field of industrial software.
AI industrial co-pilotThe goal is to enable workers to operate machines more efficiently, and tasks that used to take weeks to complete can be achieved in minutes by the industrial co-driver, which can significantly reduce the time and increase productivity.
Genai applications for industrial softwareIt involves revolutionizing the way enterprises approach generative AI programs, by integrating Amazon Bedrock, an AI-based model service, with Siemens' low** platform, Mendix, to accelerate the industrial software development process with just a few clicks, using a simple graphical interface and drag-and-drop instructions.
With the rapid development of generative artificial intelligence (GENAI) technology, its application prospects have attracted much attention. Whether GenAI will become a "sharp weapon" in the field of industrial manufacturing, promote the intelligent upgrading of traditional manufacturing and improve the industrial ecology, the industry is currently receiving mixed reviews.
Behind these discussions, some tech companies are embracing GenAI with active actions, pushing all sorts of explorations forward.
Well-known research institutes** also support the assertion that genai is about to take root in multiple industries.
A representative study, such as Goldman Sachs, believes that GenAI's breakthrough will bring unprecedented changes to the world. With the introduction of new tools such as NLP, GENAI and NLP could drive global GDP growth by 7% over a decade, equivalent to adding $7 trillion to the global economy.
A recent research report by Boston Consulting Group BCG analyzes the application of generative AI in the factory of the future, which is more informative.
Key takeaways include:
Rather than replacing traditional artificial intelligence or existing industrial control systems, genai plays a complementary role to pave the way for the factory of the future.
With the development of genai solutions, the autonomy of machines is constantly improving, allowing devices to self-regulate and adapt to unfamiliar environments.
In today's article, we will focus on this research report by BCG, and analyze the application potential, implementation path and precautions of GENAI in industrial manufacturing through specific cases, in order to provide reference for the transformation and upgrading of industrial manufacturing.
BCG recently surveyed manufacturers to gain insight into their views on emerging technologies.
The survey found that regardless of their enthusiasm for digitalization, manufacturing executives, including GenAI, see artificial intelligence, including GenAI, as the technology most likely to revolutionize operations.
BCG analysis shows that AI can increase shop floor productivity by more than 20%, with a return on investment of just 1 to 3 years.
For example, an automotive company has seen an AI application increase in productivity by 21%. Among them, the AI-driven defective product consultant optimizes the parameters, so that the scrap rate is reduced by 25%; The pump valve health monitor almost eliminated the failure of key production pumps, and the equipment efficiency was increased by 7 percentage points; The quality inspection system reduces the quality inspection manpower by 65% and improves the detection accuracy.
Artificial intelligence technology has many sources and is widely used. Machine learning and deep learning are mainly used for data analysis, classification, clustering, etc.; GenAI, such as ChatGPT, can create new content based on prompts.
Since generative AI offers new opportunities for innovation in manufacturing, pilot projects are an ideal starting point for companies to practice generative AI. Industry experts from companies such as NVIDIA, Siemens, and Invisible AI shared three typical cases of industrial GenAI empowering smart factories.
Case 1: "Synthetic data" allows robots to pick up and place different objects
With the help of artificial intelligence training, the robot has acquired the ability to process a variety of objects, even chicken wings.
NVIDIA and Soft Robotics have partnered with food manufacturers to enable robots to accurately identify piles of chicken wings and pick up individual slippery chicken wings through generative AI solutions.
In the past. This is an extremely challenging task as the shape and posture of the wings are difficult to judge in advance, and there are many combinations. Artificial intelligence is unique in building realistic 3D digital twins and simulation environments. Compared with shooting a large amount of real life**, using the "synthetic data" generated by the algorithm to train the model can greatly save time and cost.
Pictured: Soft Robotics' robot is able to identify and pick up a single slippery chicken wing from a pile of chicken wings.
Case 2: Using outlier detection, the throughput of the production line is doubled
Although the factory director can't be everywhere, smart devices can. Invisible AI helps manufacturers optimize assembly lines with GenAI smart devices.
Once an anomaly is found in the execution cycle of a part of the work site, the AI becomes a "clairvoyant" that provides insight into the production landscape, identifies the anomaly, and directs engineers to key issues.
Figure: Using artificial intelligence tools, car manufacturers find abnormal time hours at workstations.
With the help of Invisible AI, an automotive manufacturer doubled the production capacity of the production line. In another case, an automotive OEM partnered with Invisible AI to identify underutilized sites, and the OEM used this insight to consolidate workstations and increase throughput by 5% per shift while redividing 20% of its workforce.
Case 3: Agile simulation of new production lines and processes
Digital twin technology reduces the risk of new plant design and process changes. It establishes a 3D simulation environment of the virtual factory, which is connected to the existing system and looks and operates like a physical factory.
Taking it a step further, the industrial metaverse makes this possible, building virtual spaces specifically for manufacturers. NVIDIA and Siemens are bringing virtual technology to a wide range of industrial users through digital twins.
Figure: The entire planning phase of manufacturing production can take place in the industrial metaverse.
Digital twins cover a wide range of technologies, including the use of genai. The use case is very fresh in this area, and Freyr Batteries has built a complete virtual model of the battery factory, covering details such as infrastructure, equipment, ergonomics, safety, etc., to achieve a realistic simulation of product production, and greatly reduce the risk of actual factory planning.
Genai introduces a range of innovative features, but it is not well suited for tasks such as fault detection, production analysis, or fixed-point optimization. For these tasks, traditional AI has a good solution.
Still, GenAI has an important supporting role to play in helping manufacturers realize the smart factory of the future. Its unique capabilities enable manufacturers to automate and enhance factory processes and assist employees in new ways.
According to BCG, GenAI can play a role at all levels, transforming factories from reactive to proactive, ultimately to intelligent and autonomous operations. It is an important enabler for the realization of the smart factory of the future.
Summarized as GenAI capabilities, there are three typical types of manufacturing use cases: assistance systems, recommendation systems, and autonomous systems
The first category is assistance systems.
This type of genai application increases the efficiency of practical work such as programming, equipment maintenance, etc. For example, engineers have traditionally had to manually program machines and logic controllers. The genai tool can automatically generate**, reducing the amount of engineering and time cost, and the engineer only needs to review and adjust**.
Similarly, genai can aggregate the operator's extensive experience and knowledge and translate it into data-driven recommendations.
It can build models that validate operator recommendations for parameter adjustments or handling exceptions for optimized equipment through data analysis. By automating coding and translating employee experience and knowledge, genai can play an important role in improving work efficiency.
The second category is recommender systems.
Genai can provide advice and guide staff in choosing the best option.
The application value of Genai can be seen in the ** sexual maintenance. In the past, manufacturers used fixed-cycle maintenance to prevent failures. With the application of machine learning, it is possible to identify patterns and faults by analyzing different sensor data.
This state-of-the-art maintenance process is further enhanced by the automatic generation of text or image instructions for maintenance steps, including spare parts lists. This frees up maintenance personnel to focus more time on execution, which increases efficiency and reduces costs. Even inexperienced technicians can repair equipment efficiently with the help of genai tools.
The third category is autonomous systems.
The developers are exploring the use of Genai to achieve machine autonomy. For example, many handling operations are still manual and difficult to automate, so Genai can translate the engineer's voice prompts, such as "Give me spare parts 47-11", into a series of actions that the robot performs automatically. This reduces training for specific environments and tasks, reduces engineering costs, and increases productivity.
Another example is the use of genai to synthesize training data for quality control of machine vision, allowing for a quick start-up of the system without the need to collect large amounts of real-world data in production.
Through simulation learning and content generation, GENAI can autonomously adapt to new environments and greatly advance the level of automation in the manufacturing industry.
To successfully roll out AI in manufacturing, it's not enough to identify application areas, but also to build a solid foundation in both people and technology. The human capabilities required for the development and operation of genai applications are similar to those of traditional artificial intelligence. However, the GenAI technology architecture is more complex, including: models**, platforms and infrastructure, as well as application operations.
The combination of options for these technology architectures has resulted in a variety of operating models for GENAI in the manufacturing industry, which can be divided into the four types in the diagram above, each with its own advantages, and manufacturers can choose the best solution according to the actual situation.
In general, there are a variety of solutions to choose from for the implementation of the GenAI technology architecture, each with its own advantages and disadvantages. Manufacturers should make decisions based on their own realities and needs.
All things considered, here are five steps manufacturers can follow to integrate GenAI into their operations:
The first step is to diagnose the current situation and identify opportunities and value enhancement opportunities for GNAI applications.
The second step is to design a target vision, strategy, and roadmap. Evaluate the benefits of various genAI applications and identify personnel and technical measures. Choose the right genai model to take into account the effect, cost, and response speed.
The third step is to develop the genai solution and supporting measures.
The fourth step is to pilot the GenAI solution and supporting initiatives to spark momentum for widespread adoption within the organization.
The fifth step is to roll out the proven genai application portfolio in production and launch more pilot projects to continuously expand the application scenarios.
Generative AI is quietly changing our world, and its application in manufacturing has become a hot topic.
Through case analysis, this paper sorts out the value and role of GENAI in intelligent manufacturing. GenAI can realize intelligent map recognition, voice interaction, intelligent decision-making, etc., greatly improving the level of automation and autonomy in factories.
At the same time, we should also be aware of the limitations of GenAI, which still has advantages in detection and analysis, and the two should be applied in complementary applications. To truly realize industrial intelligence, enterprises also need to pay attention to the selection of technical frameworks, talent training, and the gradual application of GenAI.
While genai injects new momentum into the industry, it also brings new challenges to enterprise management. We need to respond prudently to promote the steady and intelligent upgrading of the manufacturing industry.
References: 1Generative Ai's Role in the Factory of the Future by Daniel Küpper, Kristian Kuhlmann, Monika Saunders, John Knapp, Kai-Frederic Seitz, Julian Englberger, Tilman Buchner, Martin kleinhans，**boston consulting group
2.Turning Genai Magic into Business Impact by Nicolas de Bellefonds, Sylvain Duranton, Vladimir Lukic, Jessica Apotheker, Rich Lesser, Theo Breward, **Boston Consulting Group
3.Four AI Case Study Successes in Industrial Manufacturing, by Carrine Greason, **Control Engineering
4.AI Agents Help Explain Other AI Systems by Rachel Gordon, Control Engineering
Selected articles
1.2.With the support of open source, how does the first power IoT operating system release the new kinetic energy of AIoT + smart power? 3.