Huawei Technologies, China Academy of Information and Communications Technology (CAICT), and China Telecom
Smart Factory: Manufacturing quality management in home appliance industry to renovate and modernize the existing manufacturing facilities to meet the challenges of high quality standards in the future. The methodology and practice can be extended to other manufacturing-oriented sectors.
China has been the “World’s Factory” especially for manufacturing labor-intensive products, for more than two decades. In recent years, as the world is moving towards Smart Manufacturing, the Chinese industry has been feeling the pinch from facing the enormous challenges of the number and diversity of factories needing to be upgraded.
Although superseding the old ones with modern factories through the “green-field” approach seems to be straightforward, the excessive capital investment for demolishing the old factories and replacing them with contemporary ones has become formidable. Furthermore, workers need to be temporarily furloughed during the period of the new factory construction, which can cause severe social challenges as well.
As a result, in addition to simply building new modern factories, it will be crucial to effectively retrofit the existing factories with new sensing, control and intelligent platform to help improve decision-making technologies developed through supporting the Industrial Internet of Thing (IIoT).
Furthermore, it is necessary to create a repeatable process to renovate and retrofit the legacy manufacturing facilities.
By managing the quality measures throughout the production line and using them as the key factor to demonstrate the effectiveness of the improvement while modernizing the existing factory, the testbed team adopted the IIC’s Industrial Internet Reference Architecture (IIRA) and constructed the testbed with advanced data collection, analysis and process management capability. The Manufacturing Quality Management (MQM) Testbed is implemented through retrofitting an existing factory for modern production by adding elements, such as sensory networks as well as an analytical engine. Many unexpected challenges, ranging from the stability of the power source to the accuracy of the deep learning analytic engine, will impact the Testbed. These experiences, although painful, are good lessons learned and will be fully documented to benefit parties interested in engaging in similar activities.
- MQM is a process to retrofit the existing manufacturing facilities with modern technologies
- MQM leverages the IoT and sensory network technologies for data collection, transmission and intelligent storage to support an effective data processing scheme
- MQM employs cognitive data analysis schemes to adaptively fine-tune the manufacturing process
- MQM will include energy efficiency and environment control into manufacturing process.
how it works:
The physical platform of the MQM will be used to renovate an air conditioner production line in a factory with focus on improving the yield rate of the welding process for the condenser tubes. The quality check to determine a pass or fail of the product is based on pumping air into the tube. Experienced examiners listen to the noise and make an assessment manually. This setting is far from satisfaction and the high false passing rate is costing the company dearly in RMA.
Following the IIRA architecture, an upgrade was performed on the Physical Platform of the MQM Testbed by adding the noise detection analytic engine. In order to maintain the data integrity, a housing unit was used to lower the ambient noise around the testing area.
The software platform to support the MQM Testbed consists mainly of four components: (1) Cognitive Analytics, (2) Data Processing, (3) Intelligent Storage and (4) Applications. Only a portion of these software elements are deployed in the existing system so a major overhaul in the Software Platform is expected.
1. Cognitive Analytics
The main purpose of this part of the platform is to process data collection from multiple devices and tools. Data-driven analytic is applied for teaching the machines and engines to learn from experiences. Distinctive features of the platform include cognitive analytics and soft computing, which will bring statistical modeling and business intelligence into reality. Novel learning algorithms and control protocols will be developed for making recommendations. This platform also provides APIs to users to develop customized solutions for specific use cases.
Inspired by the physical structure and functionalities of the human brain, the cognitive computing models are developed. Predictions and inferences for quality management are generated by a variety of learning algorithms by providing data to the cognitive model. In addition, heterogeneous computing and dynamic resource allocation are applied for further improvement in the performance of the proposed platform.
2. Data Processing
The Data Processing Platform provides a computing framework for the Cognitive Analytics including computing task scheduling and resource management. The platform also provides a high throughput messaging path between the brain-like Cognitive Analytics and the Intelligent Storage.
3. Intelligent Storage
Intelligent Storage is the basis of the whole testbed. It supports file content analytics, a data-aware-based data layout and a unified operation interface to upper layer users through three key components: Data Aware Engine, Content Analytics and Unified Storage Engine, thus making it distinctive from existing storage systems in complex factory environments. It will be designed and implemented on integrated hardware architecture to offer plentiful storage resources.
The MQM Testbed will focus on fault prediction and quality management. In the future, the customer's personalized requirements will be taken into account. The customer's personalized requirements are virtual simulation by PLM (Product Lifecycle Management). When the user's personalized requirements are verified by the PLM, the order will be issued to the factory. The iMES (Manufacturing Execution System) will execute the order and the SCADA (Supervisory Control and Data Acquisition) will monitor the production process. The production process of the material will be pulled by the iWMS (Warehouse Management System).
To reduce the dependency on the physical plants and to avoid unrecoverable failures, a simulation platform is deemed necessary and rewarding. The simulation platform exhibits the main features of the manufacturing processes and it can be used for solution validation. The essence of the simulation environment lies in a novel mechanism that enables it to interact with the physical plant. The simulation platform, which is also intended as a “remote mirrored factory” of the Testbed, may not be fully developed until 2018.
The key process of the Testbed is the detection of the deviation in product quality and MQM is the process to meet the challenges. The initial target is to increase the productivity by 15% while lowering the CAPEX and OPEX by 27% compared with the status quo.
A remote mirrored factory enables the Data Center to monitor the physical factory while experimenting with improvements to the production processes without actually interrupting the existing practice.
The objective of the testbed is to develop and validate the MQM, which is a repeatable “Brown Field” quality management process, to help renovate and modernize the existing (legacy) manufacturing facilities to meet the challenges of higher quality manufacturing standards in the future.
In China, there are dire needs, from both the government and the industry, to modernize the existing manufacturing facilities to meet the future higher production quality standards; while providing the least impact to the personnel during the renovation. Most of the existing manufacturing facilities were designed for massive production, which may not be fit for producing high quality products with high efficiency. Although building new facilities may also be a viable solution to meet this purpose, the investment and impact to the existing employees may cause severe financial and social impacts during the transition. Therefore, there is a need for a systematic way to examine and evaluate the existing manufacturing facilities and decide the best ways to retrofit the existing facilities with adequate technologies. More specifically, the Testbed will start with an existing automatic welding facility provided by a collaborating partner to establish the following:
- Stable data collection and transmission mechanism in a noisy and lossy environment, (such as welding stations in home appliance manufacturing) to maintain the integrity of a large amount of the time-series data;
- Effective storage and data processing schemes with a cognitive analytical process to provide real-time feedback to correct the production process on a timely basis; and,
- Adequate energy efficiency and environmental control to meet the future government regulations.
The trigger of the process is the detection of the deviation in product quality and MQM is the process to meet the challenges. The initial target is to increase the productivity by 15% while lowering the CAPEX and OPEX by 27% compared with the status quo.
Furthermore, MQM is an essential building block for “Intelligent Manufacturing”, which is the core of “Made in China 2025.” The proposed MQM testbed will be incorporated into part of the 94 State approved projects.
Interested in learning more about the MQM Testbed? Email us!