Enabling Smart Factory with Edge Computing

Smart Factory is the future in manufacturing.  Integration of smart technologies such as IoT, AI and machine learning in factory equipment allows for real time analysis and immediate reactive actions that can maximize output and minimize downtime.  In this article, we will explore what is Edge Computing and its importance to a smooth operation in a Smart Factory.


What is Bluetooth Mesh Networking?

With many sensors and IoT devices installed in the many factory equipment, these devices communicate with each other and the main server through a Bluetooth network.  Such a network is also referred to as a Bluetooth Mesh.  The size of this mesh can be tens, hundreds or even thousands of nodes.  More details of how such Bluetooth Mesh can be found here:



What is Edge Computing?

On the wiki, Edge Computing is defined as, “In one vision of this architecture, specifically for IOT devices, data comes in from the physical world via various sensors, and actions are taken to change physical state via various forms of output and actuators; by performing analytics and knowledge generation at the edge, communications bandwidth between systems under control and the central data center is reduced.”, so what does this mean?


It is best to illustrate edge computing via an example.  There are five temperature sensors setup around the house to monitor temperature change and each of these sensors sends its current temperature to the main server directly.  Without edge computing, if there are no changes in the temperature, all five sensors send the same temperature reading continuously to the main server.  As you can see, this is a waste of energy and causes collision (we will discuss this in subsequent sections).  With edge computing, an algorithm can be developed on each of the sensors.


It is designed so that each sensor won’t sent temperature information to the main server unless it’s has a change of at least 0.5 degrees.  This way, the sensors will only send “useful” data and won’t update main server data unless there is a big enough change.  This is the main philosophy of edge computing.


MtM+ Technology’s Edge Computing Research and Solution

MtM+ Technology has implemented Smart Factory solutions through its Data Collection Stations (DCS) (Details and specification outlined here: https://www.mtmtech.com.tw/DCSV1.html).  DCS collects environmental data and communicates with other DCS units and the main server through a Bluetooth Mesh network.


MtM+ Technology partnered with a company in the Semiconductor Industry to implement a smart cabinet for silicon wafer storage.  Silicon wafers can oxide (rust) quickly with oxygen therefore must be stored in nitrogen cabinets.  Maintaining a proper level of nitrogen in these cabinets is proven a challenge and traditionally, nitrogen is just refilled regularly over time.  It is found in our research that the level of nitrogen in these cabinets has a strong relationship with temperature and humidity.  DCS machines are equipped with temperature and humidity sensors and will replenish nitrogen accordingly to changes in these two matrices.  Many units of these DCS machines were implemented in the many nitrogen cabinets and worked really well.  As more units were installed, problems arose.   Please see the following chart.

It is found that as the number of DCS machines increased, the percentage of data loss to the main server increased exponentially.  For example, if there are only five people talking to you at once, you are very likely to hear fully what they are talking to about.  Suddenly if there are twenty people talking to you at once, some of the voices will surely collide and you will miss some of the information.  This data collision  is precisely the problem and is the main cause of this data loss.  This data loss resulted in inaccurate temperature and humidity readings which interfered with the smart control of nitrogen replenishment to the wafer cabinets.


Looking at this issue, edge computing can fully resolve this.  An algorithm has been developed and implemented on each of the DCS units so that only “useful” information will be sent to the main server.  Referring to the chart below.

The black line represents humidity data measured in one DCS unit.  The red line shows humidity data received at the main server without edge computing while the blue line shows the same data with edge computing in place.   As can be seen in this plot, with edge computing, a very accurate representation of the real data is achieved (Pearson’s R of 0.9906).


Why Bluetooth and Edge Computing over Wifi?

Smart factory sensors and the communication interface with the main server can also be implemented using a Wifi network.  Unlike Bluetooth, a Wifi network is not limited by bandwidth so edge computing is not needed.  MtM+ Technology has done extensive research comparing the two technologies and found that Bluetooth implementation result in an average cost savings of 76-83% over Wifi.   Here are the findings.


Interface with the main server

Each Wifi sensor node connects to the main server through a wireless router (called Access Point or AP).  Each of these routers can connect up to 30 sensor nodes.   If there are 250 sensors, 9 routers will be needed.  For Bluetooth mesh connection to the main server, only one dongle is needed for up to 500 sensor nodes.  As the number of sensors in the network increase beyond 30, the need for addition routers for Wifi implementation increases cost dramatically.  This is before comparing the cost of a Wifi router over a Bluetooth dongle which is easily 8 times.


Network Infrastructure and physical size of sensor network

Each Wifi router in a Wifi network requires a physical wire (or fiber optics) to be connected to the main server and/or the internet.  A wireless repeater can be used but with each repeater, the network bandwidth will be cut in half.  As bandwidth decreases, it faces with issues like data loss.  A Bluetooth mesh sensor network, however, do not require a physical wire network.  As long as each Bluetooth sensor node is within 30 meters of each other, data can be transmitted to the main server through each of the nodes and the dongle.  For older factories, the cost to implement large scale physical wired networks will be extremely expensive.  Without many obstructions, each Wifi router has network coverage up to 300 squared meters.  For large factories, a large number of Wifi routers will be needed which makes it less feasible when compared to a Bluetooth mesh network.



Smart factory is the inevitable future of manufacturing.  Illustrated in our research, as the number of DCS units increased, the data loss to the main servers can be as much as 80%.  With the implementation of edge computing algorithms, very accurate representation of real data can be achieved.  The advantages of a Bluetooth mesh network over Wifi were also discussed and it was found that a cost savings of 76-83% was achieved.  It is concluded that a Bluetooth mesh network along with edge computing is proven to be a vital and cost effective part of any large scale smart factory solution.

Getting started with MtConnect05 on Nordic SDK

MtConnect05 packs with Nordic nRF52832 micro-controller. Nordic provides a rich and well-tested software development environment, nRF5 SDK, for the nRF52 Series modules. The nRF5 SDK is intended to be used as a foundation for any generic Bluetooth® low energy, ANT or proprietary 2.4Ghz product development.

To introduce you to development using the MtConnect5 applications with Nordic nRF5 SDK, we will provide you with step-by-step instructions.

Continue reading

Getting started with ARM mbed and MtConnect04S

ARM® mbed™ is an IoT development platform for ARM-based microcontrollers. The best thing about them is they build the IDE in the browser. You won’t need to install SDKs on your PC or MAC. There are already a lot of open source examples in the community. You can visit ARM mbed to learn more about it.

To introduce you to development using the MtConnect04S applications with ARM® mbed™, we will provide you with step-by-step instructions.

Continue reading