The global automotive industry is in the midst of major changes and challenges against the trend of electrification and intelligence. According to the recent statistics released by a local auto industry association, the sales of China’s fuel vehicle market have declined for three consecutive years. With the intensification of the price war, some car companies have even withdrawn from the market. The auto parts manufacturers caught in it are facing the problem of how to survive and grow against the increasingly fierce competition.
Yanfeng Auto International Automotive Technology Co., Ltd. (hereinafter referred to as “Yanfeng Auto”) is a leading Chinese automotive parts supplier specializing in automotive interior and exterior trim, car seats, cabin electronics and safety systems. Headquartered in Shanghai, Yanfeng Auto has nine R&D centers, more than 240 factories and technical centers in 20 countries around the world, with over 55,000 employees. Facing challenges, Yanfeng Auto’s approach is to work with companies like IBM with advanced technology, industry experience and technical expertise to accelerate its own data-driven digital transformation to reduce cost, improve efficiency and scale for company-wide innovation.
Yanfeng Auto receives a huge number of orders from automakers and downstream manufacturers every day, and it previously had to manually convert external general orders to internal orders based on experience. In each factory, it took an average of 150 minutes a day for two staff members to sort the orders manually, accompanied by 15% classification errors. That presents big challenges for the company in terms of labor cost and efficiency.
Leveraging the powerful natural language learning capability of IBM Watson Discovery and the hands-on support of IBM Customer Success Manager (CSM) team, Yanfeng Auto successfully built up an AI model that was trained with its mixed data of structured data and unstructured text, covering 180 million historical data, with over 200 permutations and combinations. The model has learned the rules behind the internal orders corresponding to general orders. The AI model helps the company realize a fully automatic execution process without manual operation, increasing the order classification accuracy rate from 85% to 97%.
The Intelligent Manufacturing Department of Yanfeng Auto hopes to work with IBM CSM team to explore the way of building up its intelligent inventory platform with predictive capabilities. To drive predictive decision making and automatic recognition, they need a large amount of data across the company for AI model training. However, the first roadblock is its outdated way of data transmission.
To maintain real-time situational awareness of the parts inventory in various manufacturing workshops in more than 240 factories around the world, Yanfeng Auto needs to quickly transmit back to headquarters the thousands of real-time photos taken at each plant. Previously, the Intelligent Manufacturing Department used the traditional copy and paste method to transfer the photo files by batches. Due to slow transmission speed, big network delays, and serious packet loss, they had to manually select and copy the photo files by batches and repeat the process multiple times. This was not only time-consuming but also made it easy to make mistakes. At the same time, if the transmission was interrupted, it could not be reconnected and resume transmission automatically, nor could they customize the transmission speed, or fully utilize the transmission bandwidth of the backbone network.
With support of IBM CSM team, Yanfeng Auto successfully deployed Aspera Module of IBM Cloud Pak for Integration within only one day to build up a lightweight enterprise-level file transfer solution for Yanfeng Auto, which increased its file transfer speed by 10 times, saved manual waiting time, avoided human mistakes, realized automatic transmission resumption and automatic network reconnection. With this solution, Yanfeng Auto now can dynamically configure transmission bandwidth and speed limit without affecting the performance of its ERP core system and maximize the transmission efficiency of its real-time monitoring files, laying the data foundation for realizing the department’s vision of building up an intelligent inventory platform with predictive capabilities.
Yanfeng Auto has previously deployed an open-source Kafka cluster in each branch factory to extract data from multiple real-time production data in its MES system and provide them to the MI Kanban (Dashboard) System of each factory for query and display. However, this open-source system poses multiple operational complexities.
For example, for each manual installation, deployment, configuration, upgrade, and maintenance, it would take days or weeks and incur a huge labor cost. Moreover, it was not able to ensure enterprise-level security and high availability, and it did not support natural integration with the core business systems and the common production systems. Finally, there was no Kafka technical support or after-sales guarantee, resulting in the need for ongoing investment in staff training and expert consulting services.
With the support of IBM CSM team, Yanfeng Auto has successfully adopted Event Streams Module of IBM Cloud Pak for Integration in one of its factories as a prototype for real-time data extraction. The data-generating application extracts data—such as parts production shifts, production quantities, demand quantities, rework quantities, sequencing and other relevant production data—from the MES system and sends them to the corresponding data topic channel. Applications that extract data can use the data directly by subscribing to the corresponding topic channel of Event Streams. The MI Skynet Kanban (Dashboard) system can select specified table fields for subsequent dashboard display and early warning analysis.
By deploying Event Streams, the enterprise-level data extraction solution, Yan Feng can achieve “one-click” deployment, out-of-the-box use, zero downtime rolling upgrades, and always have the latest stable version of Kafka. Event Streams comes with a graphical operation interface, which requires little additional skills training. It also takes advantage of high-security, geo-replication, and enterprise-grade disaster recovery capabilities of the product. Moreover, other functionalities like advanced schema registries and rich Kafka connectors and extensible REST APIs make it easy to scale. In addition, IBM provides enterprise-level after-sales service, expert consultation, and timely troubleshooting, helping the client to obtain the technical expertise they need.
In auto parts manufacturing, injection molding is one of the important processes. Yanfeng Auto provides various automakers with interior parts such as instrument panels, which require a core equipment of injection molding machine to produce. Due to the different specifications of the instrument panels of various vehicle models, the production process requires equipment changeovers. For example, when the material is switched from black to white, the equipment needs to be cleaned; when switching from white to black, it does not need to be cleaned. Switching from gold to red requires other additional actions.
Equipment switching will not only involve cost, but affect production scheduling and inventory management. For example, how do they determine the optimal economic batch size of different products while reducing inventory costs? How do they balance the capacity of multiple machines for an entire year while meeting customer needs? How can they estimate the capacity of the machine to better adjust the plan, maximize the efficiency of the machine, increase productivity, and reduce overtime? How will they ensure that the plan can be implemented in production and that changes can be responded to in a timely manner when they come?
In the context of the continuous expansion and change of demand for various auto parts, and the fact that productivity and production resources are very limited, the traditional production planning method is based on experience and manual calculation, which easily causes problems such as low production efficiency, high inventory cost, heavy labor burden and more. All of this could seriously affect the production efficiency of the company, so it is necessary to find new ways to develop reasonable production planning and scheduling schemes.
After rounds of discussions with experts from IBM CSM team, IBM Experts Lab and IBM China Development Lab, IBM experts developed a comprehensive and agile solution for Yanfeng Auto with Decision Optimization Module of IBM Cloud Pak for Data. The solution has two complementary parts—the overall multi-machine, multi-month planning scheme and the single-month fine scheduling scheme.
The solution supports nearly 100 staff in more than 20 factories to plan the capacity of hundreds of injection molding machines with highly detailed and specific plans. Each outline plan is accompanied by an accurate scheduling plan, which is highly practical for follow-up guidance for production. It is also an agile solution: planning a set of schemes only takes a few or dozens of minutes, greatly improving the responsiveness to future changes. If the customer needs or production resources change, Yanfeng Auto can adjust the plan at any time. Based on an agile and common platform, this business-friendly solution can be easily adapted and scaled to other production facilities and other similar areas.
The IBM CSM team has accompanied Yanfeng Auto on its journey of digital and intelligent transformation for two years: from the initial realization of automatic conversion of external orders to internal orders; from solving the problem of high-speed data transmission in branch factories around the world to the headquarters; from replacing its open-source tool with an IBM enterprise tool built on open source to simplify its IT operational complexity for high-availability data extraction, to the latest AI-powered solutions to realize manufacturing capacity estimation and planning for its core production equipment. From data integration and management to applying AI to its business process and planning, Yanfeng Auto has been actively co-creating with IBM technical and business experts to turn technologies to tangible business values.
Yanfeng Auto is one of the industry pioneers in China to address business challenges with a “data-first” strategy. Yanfeng Auto is also the pioneer client of IBM in China that has been working closely with IBM to co-create first-of-a-kind scenario-based solutions with IBM Cloud and AI technologies.
Yanfeng Auto International Automotive Technology Co., Ltd. (referred to as “Yanfeng Auto”) is a global automotive parts supplier, committed to providing carmakers and other users with interior and seating solutions that meet the needs of today’s and tomorrow’s driving, redefining the way you relax, work and play in the car. Headquartered in Shanghai, the company has 9 R&D bases, more than 4,200 R&D teams, more than 240 factories and technical centers in 20 countries around the world, and more than 55,000 employees worldwide, providing global vehicle manufacturers with the design, development and manufacture of auto parts products. With product innovation and forward-looking research, Yanfeng Auto will help automakers explore the future, bring better human-car interaction experience to global auto consumers, and actively promote the evolution of car driving experience.
The post Yanfeng Auto adopts IBM Data, AI and Intelligent Automation Software to accelerate digital and intelligent transformation appeared first on IBM Blog.
One of the significant challenges statisticians and data scientists face is multicollinearity, particularly its most…
Artificial intelligence is not just altering the way we interact with technology; it’s reshaping the…
This post was co-written with Mickey Alon from Vidmob. Generative artificial intelligence (AI) can be…
Prompt Poet allows you to ground LLM-generated responses to a real-world data context, opening up…
NASA has three more operational Starliner missions on the books. It hasn't decided whether it…
When training a machine learning model, you may sometimes work with datasets with a large…