Page 89 - Proceedings of the 2017 ITU Kaleidoscope

ADVANCED DATA ENRICHMENT AND DATA ANALYSIS IN MANUFACTURING
INDUSTRY BY AN EXAMPLE OF LASER DRILLING PROCESS
b
b
b
a,c
b
a
You Wang , Hasan Tercan , Thomas Thiele , Tobias Meisen , Sabina Jeschke , Wolfgang Schulz
a. Nonlinear Dynamics of Laser Manufacturing Processes Instruction and Research Department
(NLD) of RWTH Aachen University, Steinbachstraße 15, 52074, Aachen, Germany
b. Institute of Information Management in Mechanical Engineering (IMA) of RWTH Aachen
University, Dennwartstraße 27, 52068, Aachen, Germany
c. Fraunhofer Institute for Laser Technology, Steinbachstraße, 52074, Aachen, Germany

ABSTRACT etc. in manufacturing industry, the developed data-driven
methods and methodology can be applied in manufacturing
Nowadays, the internet of things and industry 4.0 from decision making processes [1]. However, the data density is
Germany are all focused on the application of data analytics not enough for extracting knowledge because the evident
and Artificial Intelligence to build the succeeding generation parameter dimensionality is not enough and “the number of
of manufacturing industry. In manufacturing planning and experiments” is limited by time or costs.
iterative designing process, the data-driven issues exist in In this work, a methodology to enrich sparse data by fast and
the context of the purpose for approaching the optimal frugal reduced models is introduced. Several typical model
design and generating an explicit knowledge. The multi- reduction methods such as mathematical methods, numerical
physical phenomena, the time consuming comprehensive methods and data-driven methods for generating reduced
numerical simulation, and a limited number of experiments models are reviewed. After obtaining sufficient and dense
lead to the so-called sparse data problems or “curse of data, machine learning methods such as clustering and
dimensionality”. In this work, an advanced technique using classification are applied to conduct the data analysis and
reduced models to enrich sparse data is proposed and knowledge extraction. The example case is a laser drilling
discussed. The validated reduced models, which are created process. The detailed enrichment of the data and the data-
by several model reduction techniques, are able to generate driven decision making process are demonstrated in this
dense data within an acceptable time. Afterwards, machine example.
learning and data analytics techniques are applied to extract
unknown but useful knowledge from the dense data in the 2. SPARSE DATA PROBLEMS IN
Virtual Production Intelligence (VPI) platform. The MANUFACTURING INDUSTRY
demonstrated example is a typical case from laser drilling
process. The processes in manufacturing industry are characterized
by multi-parameter models with high resolution. Moreover,
Keywords— sparse data problems, data analytics, the solvability of the process is considerably to be restricted
machine learning, virtual production intelligence, model by the complexity of physics. From a data analytics
reduction perspective, two main barriers are addressed in this paper,
which slow down the process to extract knowledge from
1. INTRODUCTION manufacturing processes.
The first reason is that the required number of sampling
Presently, production companies in the high-wage countries points is enormous. When the dimensionality of parameters
have to overwhelm the challenges of rapid responding to increases, the volume of sampling space increases
variant market demands, individual customer necessities and exponentially. That means a large amount of parameters
increasing labor costs. The rising complexity in the combination is needed and the existing available data
production processes motivates the generation of more becomes sparse. Especially, in manufacturing industry, the
reliable, efficient and flexible production planning and data from different process domains becomes heterogeneous
scheduling steps. The process parameter identification, and sparse within the high dimensional parameter space.
knowledge extraction, iterative and communicative process
design and multidisciplinary optimization are vital fields for The second reason is that the sampling process can be time
consuming. The sampling process is a process of selecting or
production planning and decision making. Since a large
amount of data are generated from machines, sensors, orders generating the dataset which can be used in knowledge
extraction. The sampling points can be collected from the
The investigation are partly supported by the German Research real experiments or the computer aided calculations. The
Association (DFG) within the Cluster of Excellence “Integrative number of real experiments is not only limited by the time
Production Technology for High-Wage Countries” at RWTH Aachen
University. restriction but also limited by the boundary of the

978-92-61-24291-6/CFP1768P-ART © 2017 ITU – 73 – Kaleidoscope