【2021 Solutions】 You can import AI without writing programs. Yonghong Intelligence shortened the AI project from 180 days to 3 days.
Can AI be mass-produced? How to mass-produce AI? Chimes AI, a company established less than a year ago, has built enterprise-level No-Code AI to help Taiwan's heavyweight petrochemical plant monitor 400 machines of different brands and models in just three months. Through the Tukey platform, the modeling speed is accelerated and customer-built algorithms are supported, reducing the execution time of the AI project from 180 days to 3 days.
As the name suggests, No-Code means developing application tools without writing any line of code. The method of use is very simple: through a graphical user interface. , you can easily complete the abnormal monitoring task with simple operations such as click, drag, and menu.
Raise a virtual data scientist to introduce AI in the factory to quickly achieve the goal
Xie Zongzhen, CEO of Yonghong Intelligence, serves as a lecturer at the Taiwan Artificial Intelligence School, through the Tukey platform He has led corporate teams to build AI projects step by step. The projects he has worked on cover petrochemical, textile, energy, telecommunications, finance, medical and other industries, and he has accumulated considerable experience.
Everyone who works on AI projects knows that when executing an AI project, the most difficult thing is "communication requirements". Engineers need to confirm the company's needs, communicate with the masters, define the problem, and then collect data , cleaning data, annotating, building AI models, etc. It takes a lot of time and communication costs to go back and forth. If you can use No-Code AI tools, it will be like raising a virtual data scientist in your own factory. Data cleaning, modeling and deployment are all completed by domain experts who are familiar with factory-side operations, achieving rapid deployment of AI technology. Successfully complete the transformation goal.
▲ Use the Tukey tool to lead the corporate team step by step in building AI projects.
Dr. Xie Zongzhen said that the specific method of Tukey tool is that after field experts collect the on-site data into a machine-readable format, they feed it to the No-Code AI software. Data cleaning and data modeling are performed in the software. After confirming that the modeling effect is satisfactory, it is encapsulated into an API of the AI model and connected to the BI (Business Intelligence) dashboard.
To put it simply, the available data is entered into No-Code AI for analysis, and the analyzed modules are automatically packaged and then thrown to the war situation board, whether it is equipment performance abnormality warning, process Energy-saving suggestions or raw material product supply and demand assessments can all be used as references for decision-making.
The petrochemical plant completed abnormal detection of 400 machines in 3 months and increased the accuracy by 5%
For example, the global petrochemical giant Formosa Plastics Corporation There are dozens of factories in Taiwan. If the AI projects had to be completed one by one, one equipment and one model would take about 180 days. Not only is it time-consuming and labor-intensive, but it is also slower and slower in terms of efficiency. The Tukey tool was used to build an AI model, and 15 senior equipment maintenance engineers on site conducted data modeling. In just three months, abnormal monitoring of 400 machines, including machines of different brands and models, was completed. Compressors, rotating fans, condensers...etc. The most important thing is that after the introduction, the availability rate of the equipment is increased by nearly 5%, which is not only better than entrusting professional AI engineers, but also greatly improves the construction speed.
▲ Friendly interface design, you can easily complete abnormal monitoring tasks through simple operations such as click, drag and drop, and menu.
Fan Wenxuan, associate director of Yonghong Intelligent Marketing, pointed out that Turkey’s greatest value to enterprises has two points:
First, it is a simple and fast development tool that allows enterprises to Internal concepts can be quickly verified. Generally, AI projects take 6-12 months, but using Tukey can be completed in just 3 days;
Second, it is a flexible management tool to complete the AI model. It allows enterprises to simply use it in the production (production), sales (sales), human resources (human resources), development (research and development), and finance (finance) of the company management, and is not limited to the manufacturing production line. And can be integrated with existing systems such as ERP.
Xie Zongzhen also emphasized that there are many AI solutions. For enterprises, it is a chore to introduce so many systems at once. If they can choose a stable working platform, they can be centralized. Therefore, Tukey's positioning is the concept of "microservices". The biggest difference from traditional AI projects is that the price of the project is high, which means that it can only solve problems with high commercial value. However, there are many complex problems within the enterprise, which business and administrative units have to face every day, and cannot be solved by a project approach. In this case, Tukey is a good method of "agile development".
No-Code AI significantly reduces the threshold for non-IT personnel to enter AI applications
Tukey has the function of model retraining and adjustment. When the enterprise's business strategy changes , new data must be accumulated. When the new data is accumulated, it can be fed to the same Tukey project to perform the data analysis process. "Cut the workflow into segments, similar to the concept of card swapping." Xie Zongzhen emphasized that as long as Change the original data of the first workflow and replace it with the new one. The original data cleaning, modeling, parameter setting, model export and other processes will be completed in sequence.
The model is used to monitor whether on-site equipment is abnormal. When the monitoring model is set, it is the peak state of the equipment. After monitoring for a long time, the equipment will age, the equipment will age, and the original old model will also crack and decline. Therefore, the model needs Retrain every six months. Xie Zongzhen believes that No-Code AI is a solution that significantly lowers the threshold for non-IT people to enter AI applications. How to make people on site willing to use it must create incentives to improve their work efficiency, prompting more cases to appear, such as customer service centers It can be used to develop potential sales lists. Tukey can be used to generate 120 accurate e-marketing lists from 2,000 lists, reducing the erroneous marketing strategy of shooting at random.
At this stage, companies still view AI solutions in terms of one-off projects. In fact, the performance of AI models is often related to data quality and data distribution trends. When the operating model changes (for example, process upgrading and downsizing, equipment updates), data drift occurs, and the performance of the original AI model declines. This A rolling correction is required. Therefore, the greatest benefits can be achieved by using No-Code AI tools and allowing the professional team most familiar with the business situation to monitor and update the AI model on their own.
In the future, Yonghong Intelligence will continue to accumulate application cases and cooperate with SI Company, and is expected to promote Tukey tools to overseas markets.
▲ Xie Zongzhen (left), chief executive of Yonghui Intelligent and sales associate Fan Wenxuan.
「Translated content is generated by ChatGPT and is for reference only. Translation date:2024-05-19」