IEEE Systems, Man, and Cybernetics Magazine - January 2018 - 10

Standards are another crucial element to enable intercompany networking and integration through value networks in future IoT-based manufacturing. First, a global
standard for the unique identity of each manufacturing
object is needed, e.g., IPv6, an IoT enabler with an almost
unlimited number of globally reachable addresses. Then, data
exchange standards should be designed to facilitate seamless data exchanges between objects, systems, and organizations (e.g., MTconnect [29]) and across the stages of the
manufacturing and business planning processes.

This needs to be explored through modeling and analysis,
e.g., using game-theory-based methods to model the investment, rules, and revenues. Proper pricing mechanisms should
be built to accommodate different use cases and maximize
mutual benefits. Duan et al. [32] analyzed and compared different incentive mechanisms for a client to motivate the collaboration of smartphone users on both data acquisition and
distributed computing applications. Similarly, incentive
mechanisms should be designed for IoT operators and service consumers, based on business models.

Deployment and Business Models
Undoubtedly, the IoT is becoming an attractive paradigm
that can bring great benefits to the manufacturing industry.
However, there are still many challenges involved with its
practical implementation. The deployment and business
models of IoT devices/systems are always a central issue. To
enable pervasive sensing and actuation in real time, large
amounts of sensors/actuators need to be deployed. Then,
questions such as what kind, how many, and where are likely
to arise. While we believe IoT devices will become increasingly cheap, large-scale deployments will still cause a huge
expense. A cost-benefit analysis is necessary to determine
whether the application warrants the high cost or a reasonable investment plan makes more sense.
There is some literature on the return-on-investment
analysis of RFID in supply-chain management [13], [30].
However, much effort is needed to build precise models to
predict costs and benefits for various application scenarios, as the deployment and operation of WSNs, RFID, and
cloud/big data applications are complex. For example, the
applications can be deployed in private clouds, community
clouds, public clouds, or hybrid clouds. Small and medium
enterprises can choose public clouds to better serve their business targets without huge up-front investments, while big
corporations can afford to build private clouds under their
absolute control. Also, varying pricing strategies can bring
different costs. Moreover, technical plans, costs, and benefits intertwine with each other. To tackle this challenge, at
a minimum, the following five questions should be answered,
taking WSNs applications as an example:
1) What is the immediate problem without WSNs?
2) How can the costs and benefits of deploying WSNs be
balanced?
3) Where and how many sensors should be deployed?
4) What process should be used to deploy WSNs (a one- or
multistep process)?
5) What is the update and maintenance plan?
For strictly privately owned IoT facilities, enterprises need
to cover the whole expense. In other cases, IoT facilities can
be shared among companies to improve the utilization rate
and reduce the cost, e.g., the sharing of physical assets and
service in industrial parks [31]. Designing a feasible business
model so that multiple sides can obtain their benefits through
information and resource sharing plays an important role in
the successful implementation of the IoT infrastructure.

Manufacturing Big Data
The wide adoption of smart-manufacturing devices gives rise
to huge volumes of heterogeneous data that are generated
and collected. The storage and processing of those manufacturing big data are usually conducted in the cloud. Real-time
data from in-use products can elicit an unbounded development of novel online manufacturing applications, like intelligent prognostics. For RFID systems, readers can identify the
information contained in tags and store it directly to a (cloud)
database. However, data collection in WSNs is much more
complex and challenging.
First, proper strategies are needed to balance the ondevice/in-network data processing and the cloud-based
data processing. The former method can be energy efficient
for WSNs, but this may cause the discarding of some useful
raw data. Measuring the effectiveness of sensor data is difficult and probably varies on a case-by-case basis. To
decide whether the local data should be processed on the
base node or uploaded to the cloud is still a challenge.
Some applications require a very fast (even real-time)
response, e.g., the detection of errors in computer numerical control machines and production systems. In such
cases, local data processing is more suitable to enable fast
feedback control. The cloud is strong at scalable storage
and the powerful processing of big data, but some preprocessing is still required on the base node to prevent the network congestion caused by the transmission of large data
sets. An alternative method is to gradually transfer local
data sets to the cloud during idle time. A flexible method of
collaborative data processing between local nodes and the
cloud is needed.
Second, heterogeneous big data (e.g., structured data with
different schema and sampling frequency, unstructured or
semistructured data) are gathered from various devices.
How to correlate big data from different sources and organize those related big data that may be incomplete and/or
inconsistent should be explored to lay a solid foundation for
the upper-level applications. Machine-learning algorithms
expect data that are carefully structured, so adding structures to unstructured data before processing them on a massive
scale is the norm [26]. General approaches that provide flexible schema-based big data manufacturing are required to
handle multisource data after the preprocessing. The real
challenge also lies in how to responsively find enough useful
data in manufacturing big data generated from multiple

IEEE SYSTEMS, MAN, & CYBERNETICS MAGAZINE Janu ar y 20 18

Table of Contents for the Digital Edition of IEEE Systems, Man, and Cybernetics Magazine - January 2018