Applications of AI and IoT in IR4 digital transformation

AIoT live sharing #8 Aiotmission
smart manufacturing with AI and industrial IoT

 AI and IoT serve as critical cornerstones of smart manufacturing, bringing about digital transformation in the industrial sector.

Welcome to the first episode of “Sembang AIoT,” our live channel dedicated to discussing Industrial IoT and AI for smart manufacturing.

Before we delve deeper into our subject, let’s remind ourselves of the four foundational design principles of Industry 4.0, which guide the evolution and implementation of the Fourth Industrial Revolution:

  1. Interoperability: This principle emphasizes the need for machines, devices, sensors, and people to connect and communicate with each other. Interoperability allows for the seamless sharing of information across different systems and stakeholders. This can involve the Internet of Things (IoT), Internet of People (IoP), and other connected technologies.
  2. Information Transparency: With Industry 4.0, the collection and communication of information become much more seamless and transparent, providing operators with vast amounts of useful data. By utilizing digital context to physical processes, a virtual copy of the physical world can be created. This is often referred to as a “Digital Twin”. These digital models can then be used to analyze data and predict trends, helping decision-makers understand and anticipate issues before they arise.
  3. Technical Assistance: This principle concerns the ability of assistance systems to support humans by aggregating and visualizing information comprehensibly for making informed decisions and solving urgent problems on short notice. Moreover, it includes the capability of cyber physical systems to physically support humans by conducting a range of tasks that are unpleasant, too exhausting, or unsafe for humans.
  4. Decentralized Decisions: Industry 4.0 systems are capable of making decisions on their own and performing their tasks as autonomously as possible. This principle refers to the ability of cyber-physical systems to take decisions on their own and to perform their tasks as autonomously as possible. If a conflict arises, a decentralized decision is required to be made in real-time.

The combination of Industrial Internet of Things (IIoT) and Artificial Intelligence (AI) forms the backbone of the fourth industrial revolution, as these technologies inherently support the processes integral to Industry 4.0.

IIoT enhances connectivity from the most fundamental sensors and input-output systems to controllers, equipment, and SCADA systems, reaching all the way to cloud platforms. This interconnectivity ensures decentralization and transparency, enabling swifter, informed decision-making. Additionally, AI and machine learning, driven by the rich data environment of IIoT, allow for precision in decision-making.



In today’s interactive session, we are excited to introduce the application of an industrial computer system and a vision camera to perform deep learning and inferencing on the sample production of a 3-pin AC plug. The computer system employed is the Axiomtek ebox 640-521, an 8th/9th gen Intel-based fanless system with built-in IO, which can seamlessly interface with existing controllers for automation tasks.

The proof-of-concept for Vision AI is composed of four steps:
 
  1. Gathering a dataset with the cameras
  2. Labelling the collected datasets
  3. Utilizing deep learning AI processes to create a model
  4. Running inference to yield desired results
 
This approach leads to two substantial benefits:
 
  • Minimization of human intervention, thereby conserving manpower
  • Boosting efficiency and accuracy
 
However, our exploration doesn’t stop there. We can collect essential data such as the number and types of defects, which can be uploaded to the cloud or big data for further analysis. Such analytics, powered by AI, can enhance forecasting and optimize the overall process.

Watch the streaming from youtube with the link below :-

https://youtube.com/live/1eVTd6PnEn0

please help to subscribe if you like this sharing. your like and subscribe is much appreciated!

 

Vision AI-IOT SCADA integration

AIoT Live session 7 on AIOT SCADA Integration

In the 7th ” Sembang”  AIoT session, CC Lee shared the topic about the SCADA Vs IIoT where there are many areas of overlaping between the SCADA and IIoT ( industrial IoT) at the same time they have their differences. 

Both SCADA (Supervisory Control and Data Acquisition) and IIoT (Industrial Internet of Things) are technologies that allow for data acquisition, monitoring, and control, but they are different in terms of architecture, scalability, and their typical use-cases.
 
1. Use Cases: SCADA systems are traditionally used in industrial processes for controlling large-scale processes where data from multiple sensors is combined and analyzed in real time. Common applications include Kl no water treatment plants, power generation, and oil and gas refineries. On the other hand, IoT is a more general concept that can be applied in numerous scenarios, not limited to industrial processes. It’s widely used in smart homes, healthcare, agriculture, and urban infrastructure, among others.
2. Architecture: SCADA systems typically have a centralized architecture with all the data flowing to a single central location for processing and control. IoT, however, often operates on a decentralized architecture where data can be processed and actions taken at the edge, closer to the source of the data.
3. Scalability: SCADA systems are usually designed for specific purposes and their scalability can be limited. IoT, due to its flexible and modular nature, can be easily scaled up or down depending on the application.
4. Connectivity: SCADA systems traditionally use a closed, private network for security and reliability reasons, whereas IoT devices are often connected to the internet, making them potentially more vulnerable to cyber-attacks but also more accessible and capable of integration with other systems and technologies.
5. Data Handling: SCADA systems focus more on real-time control and less on data analysis and prediction. IoT, on the other hand, leverages cloud computing and big data analytics, enabling not only real-time control but also predictive maintenance and other advanced features.
 
In the modern industrial landscape, these two technologies often work together. Many companies are integrating IoT devices into their SCADA systems to enhance data collection and analysis capabilities.

Demonstration on Vision AI Object detection

Live demonstration on object detection on the type of Sushi on a conveyor belt moving.  the AI model was built using the Mobilenet V2. 

AI data was captured and pushed to the SCADA system within the system to indicate the counting of the type of sushi passing through the conveyor belt. 

The Vision AI operation in this case was powered by Axiomtek industrial PC with a very affortable web camera system. 

To watch the live session, please visit the link below:-

https://www.youtube.com/watch?v=6wmwtmc28Bk




Jom! lets Sembang AIoT!

We are delighted to announce the commencement of our live sharing session, titled “Jom! Let’s Sembang AIoT,” which began on May 24th, 2023. The primary objective of this initiative is to provide a comprehensive overview of our collective experiences and activities in the fields of IoT and AI. Given the rapid pace of technological advancements, we have taken this opportunity to engage the audience and solicit public comments to better understand the current trends and requirements in the realm of AI and IoT.

Today, on July 2nd, we are pleased to share that we have successfully completed our fourth live session. These sessions occur every Thursday or Friday, depending on our scheduling constraints, as we are committed to fulfilling other responsibilities during weekdays. Nonetheless, we assure our audience that we dedicate ourselves to delivering the highest quality content. Our discussions cover a wide range of topics, including real-life experiences in working on AIoT projects and development, as well as fundamental concepts of Industrial IoT and AI. We highly recommend following our AIoTmission’s YouTube channel or Facebook page to stay updated with our latest content. The most recent live session is included at the end of this post, and we encourage you to subscribe to our channel if you find the information pertinent and valuable. Hope to see you in the upcoming live channel!

 

 

Live session ” JOM! lets Sembang AIoT”  a Malaysian’s style of openly talk about technology in AI and IoT. 

OEE Tracking for Smart manufacturing with AIoT

OEE Tracking Training with AI and IoT

Overall Equipment Effectiveness tracking is one of the most classical and proven methodologies of tracking production equipment and processes. With the help of Industrial IoT and AI technology, it can be digitally transformed to meet the requirement of Industry 4.0 where connectivity and data transparency is of main concern.

RiSE4WRD for Industry4WRD 

RiSE4WRD for Industry4WRD is HRD Corp’s initiative to support the national agenda of embracing the Fourth Industrial Revolution (IR 4.0). This programme is designed to assist SMEs in the manufacturing and related sectors that have participated in the Readiness Assessment (RA) under the Ministry of International Trade & Industry (MITI), to start or accelerate their digital transformation journeys. If you have enrolled for this program, you may look out for our training program on AIoTmission IOT OEE and AI OEE HRDF claimable training programmed.
 

There are 3 main source of data that dictate the OEE index, Availability, Performance and Quality. By using the Iot devices and tool, data can be tapped from the machines and processes. Those data can then be calculated based on the formula in order to produce the online data of OEE index. 

By looking at the OEE index, you will know how effective or how efficient your process is running. IoT devices and tools allowed you to make this available and connect to the wanted servers or platform. The visualization is make possible at the local site as well as Cloud.

AIoTmission AIoT OEE Training provides you a very practical training outcome where after the training, you should be able to know the overall concept of IoT and AI in the application of making up the OEE Tracking system and you should be able to put that into the implementation of the system. 

There are 4 main key skill sets in the training:-

  1. IoT connectivity from sensors to the system
  2. AI application on Quality factors
  3. Data integration to databases and related production system.
  4. Data manipulation and visualization at local SCADA and  Cloud.

Book your interactive chatting session (15 mins) with our Trainer about your needs with no obligation.

Exclusive AIoT Mission Training

Exclusive AIoT Mission Training

AI and IoT are key technologies in the Era of digitalization. In the context of AIoT, AIoTmission focus on proving innovative tools in data collection for IoT and optimized AI inferencing emphasized by Intel on PC based platform to achieve the AI inferencing result with Intel core CPU. 

In this training, practical experience in shop floor data acquisition is made into practise with Axiomtek IoT edge gateway and Stack light IO module that acted as a simulation of the production process and machine’s status.

 
Tesla IoT edge acted as a  tool to construct and build an intuitive data collection method at the simulated shop floor.
The industrial communication protocol was introduced and IoT protocol set was taught in the training.
Cloud computing and experience was brought in as a platform for participant to realize the how the IoT cloud connectivity and as well as the dashboard’s  presented data is realised.
 
On the AI portion, the concept and theory behind the AI engine was introduced and Hands on in building the AI model was put into
the training for participants to get to know how the AI model is built as that is the key about the whole AI process in machine Learning and also the AI Deep learning.
 
We hope that this training will spark up the idea of more applications of AI and IoT in their working environment as we know the technology will be only become powerful when it is applied.
 
We innovate . We Apply and we sure train !
#AIoTmissionforAIOT
 
 
 

SCADA Training OEE series

On site SCADA Training on OEE integration to our client.

It was a great 5 days of training just before the announcement of endemic. Thank you to all the team members for making this happen. As OEE still stay relevant in all the manufacturing process. Indusoft Web studio was brought to the front to gather, automate and present the shop floor data. it is a successful transfer of skill to the client and they are able to expand on their own from there on!
A quick introduction, AIoTmission is a training and consultancy arm of Axiomtek Malaysia to research, to share and transfer knowledge and skill that has been accumulated in the past 25 years in the industries. One of the key focuses will be the Digital transformation in the industries or IR4 activity that require the engagement of technology such as IoT, AI and any of technology development.

Non intrusive analogue meter reading using Axiomtek AI suite

In the Era of digitalization, almost all the old apparatus in measurement is now being relooked as data analytics has become vital in many important decision-making processes.

The smart meter enables remote and automated meter reading, however, in many industries, there are still a huge number of analog or gauges in operation. Human operators need to read the meter reading and log it down to enable the data to be used elsewhere. Someone might suggest replacing those analog meters with smart meters, but most of the time those gauges are not easily touched or tempered in some critical process as it might cause some unforeseen circumstances.

As far as digital transformation is concerned, Axiomtek Malaysia provides both Industrial AI and IoT solutions based on the Intel platform in most of the digital transformation processes specifically in the area of the manufacturing industry and industrial applications. One of the challenges is being able to tap onto some analogs meters or gauges and convert that to a digital format for data analytic purposes.

In achieving the above objective, we went into using Axiomtek AIS to make a non-intrusive analog to digital meter reading based on vision AI. A camera will be located in front of the meter/gauge to generate live video stream to Axiomtek AIS. Axiomtek AIS uses deep learning approach to locate the position of meter pointer and convert it to digital value. The digital value will then send to SCADA engine for display and logging purposes.

you may look at the short video below, it shows the successful first experiment by using the Axiomtek AIS and thanks for AIoTmission team ( One of the axiomtek AIoT traning and consultancy company) in making this happen.

We look forward to any of your feedback and if you found any needs like the above situation like what we have mentioned. do contact with us at +603-77731203 or Whatsapp us at 017-9698026.  Get to know more Axiomtek’s AioT solution. visit us at :

https://axiomtek.com.my/Default.aspx?MenuId=Products&FunctionId=ProductMain&Cat=276&C=AIoT%20Solutions

How to choose the right camera for your AI application in smart manufacturing?

Axiomtek AI starter kit supports USB webcam, IPCAM, and GENICAM compliance vision camera. But how do we know which camera is suitable for our AI application?  To answer this question, we need to understand the characteristic of these cameras.

A USB webcam is a camera that connects to a computer through a USB cable. USB webcams are compatible with a variety of operating systems including Windows, Mac, Linux, and even some gaming systems like the Sony PlayStation but please double-check the camera datasheet if you are developing your AI application in Linux. USB webcams are generally used for multi-person meetings, video chats, online teaching, etc. Therefore, examples of AI applications that use USB webcam are human/object detection and face recognition. Most of the USB webcam support autofocus, which means the USB webcam will automatically adjust the camera focus when the object is out of focus. But sometimes we might experience the USB webcam cannot focus on the object very well. The three most important factors influencing autofocus are the light level, subject contrast, and camera or subject motion. For human detection or face recognition, slightly out of focus does not affect the deep learning model performance. However, for smart manufacturing, we do not recommend using an autofocus enabled USB webcam to perform defect detection on the product on a moving conveyor belt especially when the defect area is small. This is because the USB webcam might keep adjusting the camera focus when the object is on the moving conveyor belt and resulting in all captured images being blurred.

An Internet Protocol camera or IPCAM is a type of camera that sends images or video streams to the computer through a wireless (WiFi) or wired network (Ethernet cable). Unlike a USB webcam that needs to gain power from the computer USB port, IPCAM can be powered up through a power adapter or PoE (Power over Ethernet) port. RTSP (Real-Time Streaming Protocol) is the most common communication protocol used by IPCAM to send video streams to other devices. There is an RTSP server running inside the IPCAM and all the devices that receive video stream is known as RTSP client. IPCAM usually can support 5 to 20 RTSP clients simultaneously depending on camera brand and model. The first thing that comes to our mind when talking about IPCAM is it is a camera that is used for surveillance & security purpose. Therefore, examples of AI applications that use IPCAM are human/object detection, person search, vehicle detection, license plate recognition, etc. For smart manufacturing, if the camera needs to operate in an extreme temperature environment, IPCAM will be a better choice compared to a USB webcam.

Two types of vision cameras are commonly used in machine-vision applications: area-scan and line-scan. Area scan vision camera is like a USB webcam and IPCAM which use an imaging sensor with width and height for example 1920 × 1080 pixels. A line-scan camera uses a sensor that is long and very narrow for example 8000 × 1 pixels. Axiomtek AI starter kit currently only support area-scan vision camera. Vision cameras send images or video streams through USB 3.0 or GigE (Gigabit Ethernet). USB 3.0 provides 3 times higher throughput (camera frame per second) compared to GigE but the USB 3.0 cable length is limited to 5 meters while GigE cable can go up to 100 meters. Axiomtek AI starter kit supports both USB 3.0 and GigE interface as long as the vision camera is GenICAM (Generic Interface for Cameras) compliant.

Examples of AI applications by using Axiomtek AI starter kit in smart manufacturing are defect detection (detect scratch, dings, dents, missing parts), product counting, product classification, product orientation (detect front view or back view of the product is facing to the camera), etc. Some simple tasks for example product counting can be handled by using a USB webcam or IPCAM if the object size is large. A vision camera is more suitable for handling tiny objects.

How to choose the right camera for your AI application in smart manufacturing?

Axiomtek AI starter kit supports USB webcam, IPCAM, and GENICAM compliance vision camera. But how do we know which camera is suitable for our AI application?  To answer this question, we need to understand the characteristic of these cameras.

A USB webcam is a camera that connects to a computer through a USB cable. USB webcams are compatible with a variety of operating systems including Windows, Mac, Linux, and even some gaming systems like the Sony PlayStation but please double-check the camera datasheet if you are developing your AI application in Linux. USB webcams are generally used for multi-person meetings, video chats, online teaching, etc. Therefore, examples of AI applications that use USB webcam are human/object detection and face recognition. Most of the USB webcam support autofocus, which means the USB webcam will automatically adjust the camera focus when the object is out of focus. But sometimes we might experience the USB webcam cannot focus on the object very well. The three most important factors influencing autofocus are the light level, subject contrast, and camera or subject motion. For human detection or face recognition, slightly out of focus does not affect the deep learning model performance. However, for smart manufacturing, we do not recommend using an autofocus enabled USB webcam to perform defect detection on the product on a moving conveyor belt especially when the defect area is small. This is because the USB webcam might keep adjusting the camera focus when the object is on the moving conveyor belt and resulting in all captured images being blurred.

 

An Internet Protocol camera or IPCAM is a type of camera that sends images or video streams to the computer through a wireless (WiFi) or wired network (Ethernet cable). Unlike a USB webcam that needs to gain power from the computer USB port, IPCAM can be powered up through a power adapter or PoE (Power over Ethernet) port. RTSP (Real-Time Streaming Protocol) is the most common communication protocol used by IPCAM to send video streams to other devices. There is an RTSP server running inside the IPCAM and all the devices that receive video stream is known as RTSP client. IPCAM usually can support 5 to 20 RTSP clients simultaneously depending on camera brand and model. The first thing that comes to our mind when talking about IPCAM is it is a camera that is used for surveillance & security purpose. Therefore, examples of AI applications that use IPCAM are human/object detection, person search, vehicle detection, license plate recognition, etc. For smart manufacturing, if the camera needs to operate in an extreme temperature environment, IPCAM will be a better choice compared to a USB webcam.

Two types of vision cameras are commonly used in machine-vision applications: area-scan and line-scan. Area scan vision camera is like a USB webcam and IPCAM which use an imaging sensor with width and height for example 1920 × 1080 pixels. A line-scan camera uses a sensor that is long and very narrow for example 8000 × 1 pixels. Axiomtek AI starter kit currently only support area-scan vision camera. Vision cameras send images or video streams through USB 3.0 or GigE (Gigabit Ethernet). USB 3.0 provides 3 times higher throughput (camera frame per second) compared to GigE but the USB 3.0 cable length is limited to 5 meters while GigE cable can go up to 100 meters. Axiomtek AI starter kit supports both USB 3.0 and GigE interface as long as the vision camera is GenICAM (Generic Interface for Cameras) compliant.

 

Examples of AI applications by using Axiomtek AI starter kit in smart manufacturing are defect detection (detect scratch, dings, dents, missing parts), product counting, product classification, product orientation (detect front view or back view of the product is facing to the camera), etc. Some simple tasks for example product counting can be handled by using a USB webcam or IPCAM if the object size is large. A vision camera is more suitable for handling tiny objects.

Configure the DIO of Axiomtek EBOX640-521-FL with Python

Configure the DIO of Axiomtek EBOX640-521-FL with Python

DIO stands for Digital Input/Output. It is a hardware device that receives or sends digital signals. Examples of applications that use DIO are alarms, control relays, fans, lights, horns, valves, motor starters, solenoids, etc. Recently I wrote a Python binding for the Axiomtek BSP to configure the DIO of EBOX, so that my AI program that runs on Python is able to control PLC through DIO.

Some industrial PC required users to enable and configure the DIO in BIOS first, but EBOX640-521-FL does not require users to enter BIOS to enable and configure DIO. Figure 1 shows the DIO pin number of EBOX based on Axiomtek BSP.

dio-pin-768x432
Figure 1: The DIO pin number of EBOX based on Axiomtek BSP

Step 0: Install the Axiomtek BSP (with Python bindings)

Please contact us to get a copy of this Python version Axiomtek BSP. The Axiomtek BSP that you found online only supports C code. Then follow the instructions in the README file to install.

 

Step 1: Load Axiomtek BSP

c_lib = ctypes.CDLL(“/usr/lib/libaxio.so”, mode = ctypes.RTLD_GLOBAL)
libname = pathlib.Path().absolute()/”libdio.so”
c_lib = ctypes.CDLL(libname, mode = ctypes.RTLD_GLOBAL)

 

Step 2: Configure the DIO mode.

All the pins can be configured as input or output, but we need to decide which pin becomes input and which pin become output based on the device that will connect to the EBOX. Figure 2 shows a DIO relay module that will be connected to the EBOX. Pin 0 to pin 3 of the DIO relay module are output pins, while pin 4 to pin 7 are input pins. Therefore, we need to set pin 0 to pin 3 of EBOX as output pins and pin 4 to pin 7 as input pins.

dio_mode=0xf0 # pin 0~3 is output, pin 4~7 is input
res = c_lib.set_dio_mode(ctypes.c_uint32(dio_mode))

dio-module-768x432
Figure 2: DIO relay module.

Step 3: Configure output pin to high or low

res = c_lib.set_low(1)
res = c_lib.set_high(1)

Step 4: Read signal from the input pin

The input pin of EBOX always gives 5V when there is no input signal from DIO relay module. When the DIO relay module output signal to EBOX, the input pin of EBOX will become 0V.

value = c_lib.read_input(7)

That’s all. Only 4 steps to configure the DIO of EBOX.

OpenVPN Tunneling in Axiomtek AI Starter Kit

Let say we want to send the AI data from the factory back to a PC in headquarter, there are few methods to do this. The simplest method is using some remote login software like Anydesk or Teamviewer to transfer the data but users need to do this manually and the format of data must be a file. The second method is users can subscribe to a cloud service like Google Cloud Platform (GCP) or Amazon Web Service (AWS), then use reverse SSH tunneling to transfer the data. But this method required the user to possess computer networking knowledge because the user need to set up the cloud service and configure the firewall of the cloud service to enable certain SSH ports. The third method is to use Virtual Private Network (VPN) to link these private subnets together. There is a lot of VPN service available e.g., OpenVPN, Hotspot Shield, NordVPN, etc. In this example, I will show you how to use the OpenVPN cloud that comes with an intuitive web interface to set up a VPN network for AI starter kit.

The free version of OpenVPN Cloud provides 3 connections which means you can link 3 different private subnets together or in other words 254×3=762 devices can communicate with each other. After you create an account on OpenVPN Cloud, click on the Networks tab, then click Create Network.

Select Remote Access since we want to connect the AI starter kit to OpenVPN Cloud.

Create a network name. The term “Connector” is referring to AI starter kit in this demo. Always choose the closest region to reduce latency.

Enter your private subnet. OpenVPN Cloud need this info to forward incoming packet to the correct destination. Private domain is optional.

Since the AI starter kit is Linux based with Ubuntu 18.04, we select Linux and Ubuntu 18.04.

Open a new terminal and then copy-paste the command and run. The first command is to download a shell script named ubuntu_18.04.sh. The second command is to change the file permission of the downloaded shell script so that we are able to execute it. The third command is to execute the shell script. The shell script will perform some port forwarding and configure firewall rules in the AI starter kit.

Copy the token to terminal.

Ignore the offline warning, later reboot AI starter kit will solve this issue.

Click Next.

Click Finish and reboot your AI starter kit. The AI starter kit will automatic join the OpenVPN network during startup.

To verify the OpenVPN connection, open a new terminal and type ifconfig. You will see the tun0 interface if the AI starter kit successfully connects to the OpenVPN cloud.