IMAGINE, Franco-Taiwanese (ANR-MoST) Project

Integrated MAnufacturinG decIsions for NExt generation factories

x

News

11/2017: ANR Revues Mi-Parcours is held on Nov. 20-21 in Paris. IMAGINE team will present the progress updates.

10/2017: The work "Automatic equipment fault fingerprint extraction for the fault diagnostic on the batch process data" supported by ANR IMAGINE has been accepted for publication in Applied Soft Computing Journal.

08/2017: A new post-doctoral fellow, Dr. Margaux NATTAF, is hired and will start working on the smart sampling problem in Task 3 of IMAGINE on Sept. 1, 2017.

08/2017: The ANR-supported research, Equipment health diagnosis and prognosis using a wavelet-based windowing approach in semiconductor manufacturing, is accepted for APC Conference, which will be held on Oct. 9-12, 2017 @ Austin (Texas), United States.

07/2017: IMAGINE mid-term report (D+18) has been submitted to ANR for review. A review meeting is foreseen on Nov. 20/21, 2017.

04-05/2017: The ANR-supported Ph.D. student, Hamideh Rostami, visited the project partner, NTU-IIE, for four weeks. She has made decent discussions with the team members there and brought back fruitful results.

05/2017: The ANR-supported research, Equipment health modeling for deterioration prognosis and fault signatures diagnosis, is accepted and presented by Hamideh ROSTAMI in ASMC, which is held on May 15-18, 2017 @ Saratoga Springs (New York), United States.

04/2017: The ANR-supported research, Equipment deterioration prognosis and fault diagnosis, is presented by Jakey BLUE in APC|M, which is held on Apr. 10-12, 2017 @ Dublin, Ireland.

12/2016: The ANR-supported research, Equipment condition diagnosis and fault fingerprint extraction in semiconductor manufacturing, is accepted and presented by Jakey BLUE in ICMLA, which is held on Dec. 18-20, 2016 @ Anaheim (California), United States.

08/2016: The ANR-supported research, Equipment Anomaly Detection and Automatic Fault Fingerprint Extraction in Semiconductor Manufacturing, is accepted and presented by Hamideh ROSTAMI in ISMI, which is held on Aug. 8-10, 2016 @ HsinChu, Taiwan. This work has also won the "Best Paper Award" in the conference.

05/2016: Invited to participate the Journées STP du GDR MACS @ Grenoble, France. Project contents and latest progress are shared by Jakey BLUE and Hamideh ROSTAMI, in the workshop. Several brainstorming questions are received.

01/2016: ANR kick-off meeting was held on Jan. 15, 2016 @ Paris, France.

11/2015: Project kick-off via conference call with Taiwan partner, NTU-IIE. Monthly technical meetings are settled to update the mutual progress.

x

Background & Motivations

In terms of the significant advancements of information, manufacturing and robotics technologies, it is foreseeable that the epoch of a fully automated factory is within a stone’s throw. In the past, factory automation is invested in various aspects in order to manage fundamental production issues such as utilization, quality, cycle time, cost and due date. On the software side, information systems, such as the Manufacturing Executive System (MES) and advanced planning system for production scheduling and dispatching, are upgraded and installed. From the hardware perspective, equipment automation and automatic material handling system are implemented as well. Through devoting massive labor efforts in incorporating these systems, we are able to visualize and solve arduous problems in production environment, but still with less efficiency as these systems are usually monitored separately and the interfaces are not well integrated. If the productivity problem cannot be solved through these manners, more employments, more machines and loosen process constraints are usually adopted to treat the superficial symptoms but not the root cause.

Nowadays the speed of data generation/collection is faster than that of data processing. What has been integrated and analyzed in the factory automation in order to link up the interfaces remains a tip of the iceberg. With the enormous data collected continuously, the novelty, efficiency and effectiveness in data analytics must be elevated. For example, equipment prognostics shall provide precise input for scheduling and dispatching in order to avoid unexpected process interruption. Meanwhile the product quality can be better controlled by simply dispatching the product to an “analytically-suitable” machine to compensate the inter-processes effects. Moreover, unstructured data like hand-written logs in maintenance book, troubleshooting memos, on-site images/videos or even dialogues shall be taken into account in the analysis such that new norms will be formalized by learning rules from human decisions.

By abstracting a fabrication system, four fundamental elements can be identified: equipment, process, product and manpower (Figure 1). As mentioned above, automations/systems have been deployed within each element to facilitate the decision making process which is still led by human. What is lacked now and anticipated in the “future factory” would be the “automations” of the links between equipment, process and product to enable flexible and agile manufacturing environment. The competence of human resource, at the same time, will need to be enhanced so that smart decisions will be made according to the synthetic and condensed information. Furthermore, the decision cycle shall be analyzed and documented to become an empirical rule in the expert system.

Into the details of the three elements around manpower, individual advancements are still demanded such as diagnostic and prognostic maintenance policy for equipment control; optimized dispatching, scheduling and sampling algorithm for process management; and predictive product quality for product monitoring. Most importantly, information sharing and integration among the elements will be the key to the future factory. From the process perspective, predictive equipment condition and product quality would facilitate efficient and effective automation in production scheduling and dispatching. For the equipment side, knowing the process settings and product assignments will help identify the maintenance schedule. For the product side, the predictive quality model definitely needs the inputs from equipment behavior and process parameters at the same time. Therefore, manpower shall be elaborated to incorporate expert decision systems that consolidate all the information around.

Fig. 1. Fundamental fabrication elements and their relationships.

The proposed intelligent fabrication scheme gives general but critical concepts on how a production facility can be significantly improved. By consequence, it can be directly applied to production systems in pharmaceutical, food processing, health care systems, and e-commerce, and supply chain systems. By taking supply chain systems as an example, demand and inventory data are readily collected at all sort of facilities, including production sites, warehouse, and retail locations. Meanwhile, massive amount of real time information are also generated from production, transportation and logistics systems. Although supply chain management has been widely discussed, it remain unclear how a large network for participants and collaborate and coordinate with each other. By developing an intelligent scheme for large decentralized systems, research results from this project can also enhance supply chain system performance.

x

IMAGINE Structure

The structure of IMAGINE project is shown in Figure 3. It is divided into four scientific tasks. Each task is surely collaborated between the partners and is managed by one partner for report consolidation. As can be seen in Figure 3, the task managers are highlighted in red. In Task 1, the state of the art of the academic and industrial knowledge will be studied. The survey will be shared and updated constantly. Task 2, managed by EMSE, deals with equipment behavior modeling in terms of fault diagnosis and failure mode prognosis. Task 3 will handle smart sampling for proactive product monitoring which is managed by EMSE as well. As can be imagined there exists a strong link between Tasks 2 and 3 in terms of the interactions between the equipment condition and sampling strategies. NTU team will manage Task 4 and 5 which integrate the equipment behavior model and smart sampling strategies for proactive production dispatching/scheduling and equipment maintenance planning, respectively. It’s evident that equipment behavior modeling will be a critical input for product sampling, production scheduling and, of course, maintenance scheduling. The last one, Task 6, indicates that all the developed methodologies will be tested and validated through the practical case studies with the local partners (STMicroelectronics in France and TSMC in Taiwan). The two local industrial companies have been constantly (but separately) collaborated with the two academic partners in many projects. IMAGINE project will be in a good position to tackle and investigate the key issues that exist in two different manufacturing industries.

Fig. 2. Task breakdown and relationship of project IMAGINE.

While the project is carried out in close collaboration between EMSE and NTU, Claude Yugma from EMSE and Cheng-Hung Wu from NTU will serve as project coordinators. Close coordination will be ensured through bi-monthly teleconferences to review progress and, more importantly, through some face-to-face meetings at both institutions at critical junctions of the project where intensive joint effort would be needed to move forward. Short and long stays in France and Taiwan are planned every year. Dissemination activities in the tasks will be coordinated.

x

For more detailed information, please contact

Claude YUGMA, Professor & Project Coordinator, yugma@emse.fr

Jakey BLUE, Associate Professor, lan@emse.fr

You may also visit:

Manufacturing Sciences and Logistics Department (SFL)