280
D. Mishra et al.
8.3.5.4 Big Data Analytics
This tool refers to the large volume of dataset comprising of a combination of structured and unstructured data. Manufacturing industry generates huge amount of data
each day which could be explored by implementation of big data. The handling
and storage of this huge dataset would require cloud-based system. The constituents
of big data are: volume, velocity, variety and veracity. The “volume” refers to the
amount of data that could be gathered from a welding process. For instance, in case of
the GMAW technique, the dataset may contain: (a) “basic data”, which will include
the physical data of the workpieces, environmental conditions, (b) “machine data”,
which will include the identification tag of the welding power source utilized, torch
and consumables and (c) “sensors’ data”, which will include arc voltage, arc current,
wire feed speed, flow rate of the shielding gas, pressure of the shielding gas, arc
acoustics, acoustic emission, temperature and images of the welded sample. “Velocity” refers to the speed of data collection, and this will usually vary from one sensor
to the other. Of course, this dataset will be heterogeneous, as it consists of data from
different sources, which is referred as “variety". Finally, the term “veracity” aims at
deriving the useful information from the dataset, and discarding the irrelevant ones.
A recent article mentions the importance of big data in optimization of welding
process, documentation, production monitoring and quality management [105]. This
could be realized for an industry like automobile manufacturing unit, where welding
is one of the crucial operations in the production process, and it is highly imperative
to identify the rate of rejection, possible reasons for the rejection, building predictive
models using variety of data, and increasing the accuracy by using volume and
velocity of the data. For similar welding machines performing similar jobs in a shop
floor, this dataset can help in detecting, whether or not, a machine is degrading
by considering into the quality of the welds. Similarly, the warehouse for welding
consumables can be optimized by tracking the rate of consumption of the shielding
gas and filler rods. A limit value can be set for the tank containing the shielding gas,
which can alarm in real time, as the limit value approaches. This can help in ensuring
the quality of the weld.
8.3.5.5 Sensor Fusion, Signal Processing and Machine Learning
Sensor fusion is a process of combining signals from various sources. Here, the
information is extracted from multiple sources in order to integrate them into a
single indicator. The information extracted is analysed by the help of sensor fusion,
also known as data fusion algorithms, and have been found to be beneficial for
manufacturing process [5]. The automation of welding cannot be fulfilled by using a
single sensor, instead multiple sensors would be required. This is because a number
of characteristics of a welding process are required to be monitored, which may not
be fulfilled by using a single sensor. Further, benefits in terms of higher accuracy have
been reported with the use of sensor fusion. In order to fuse the data, one of the widely
utilized techniques is the “intelligent fusion” which uses neural networks. Earlier
D. Mishra et al.
8.3.5.4 Big Data Analytics
This tool refers to the large volume of dataset comprising of a combination of structured and unstructured data. Manufacturing industry generates huge amount of data
each day which could be explored by implementation of big data. The handling
and storage of this huge dataset would require cloud-based system. The constituents
of big data are: volume, velocity, variety and veracity. The “volume” refers to the
amount of data that could be gathered from a welding process. For instance, in case of
the GMAW technique, the dataset may contain: (a) “basic data”, which will include
the physical data of the workpieces, environmental conditions, (b) “machine data”,
which will include the identification tag of the welding power source utilized, torch
and consumables and (c) “sensors’ data”, which will include arc voltage, arc current,
wire feed speed, flow rate of the shielding gas, pressure of the shielding gas, arc
acoustics, acoustic emission, temperature and images of the welded sample. “Velocity” refers to the speed of data collection, and this will usually vary from one sensor
to the other. Of course, this dataset will be heterogeneous, as it consists of data from
different sources, which is referred as “variety". Finally, the term “veracity” aims at
deriving the useful information from the dataset, and discarding the irrelevant ones.
A recent article mentions the importance of big data in optimization of welding
process, documentation, production monitoring and quality management [105]. This
could be realized for an industry like automobile manufacturing unit, where welding
is one of the crucial operations in the production process, and it is highly imperative
to identify the rate of rejection, possible reasons for the rejection, building predictive
models using variety of data, and increasing the accuracy by using volume and
velocity of the data. For similar welding machines performing similar jobs in a shop
floor, this dataset can help in detecting, whether or not, a machine is degrading
by considering into the quality of the welds. Similarly, the warehouse for welding
consumables can be optimized by tracking the rate of consumption of the shielding
gas and filler rods. A limit value can be set for the tank containing the shielding gas,
which can alarm in real time, as the limit value approaches. This can help in ensuring
the quality of the weld.
8.3.5.5 Sensor Fusion, Signal Processing and Machine Learning
Sensor fusion is a process of combining signals from various sources. Here, the
information is extracted from multiple sources in order to integrate them into a
single indicator. The information extracted is analysed by the help of sensor fusion,
also known as data fusion algorithms, and have been found to be beneficial for
manufacturing process [5]. The automation of welding cannot be fulfilled by using a
single sensor, instead multiple sensors would be required. This is because a number
of characteristics of a welding process are required to be monitored, which may not
be fulfilled by using a single sensor. Further, benefits in terms of higher accuracy have
been reported with the use of sensor fusion. In order to fuse the data, one of the widely
utilized techniques is the “intelligent fusion” which uses neural networks. Earlier