INTELLIGENT SOLUTIONS FOR INDUSTRY 4.0
Research, Development and Training in Automation, Deep Learning Machine Vision, Manufacturing Modeling and Simulation, and Robotics (002769122-V)
HRDF MyCOID: 20173481604
DZUKI SuperAI Camera
The camera system is powered by the world's smallest AI supercomputer, the Jetson Xavier NX. It comes installed with Tensorflow and Keras, The operating system is Ubuntu 18.4. An inbuilt object detection module and a classification module allow users to easily deploy deep learning image processing for defect detection and parts classification such as OK/NG. The modules are also designed to communicate with a PLC thus allowing the camera system to be part of an automation system for example rejecting or binning parts. The modules are also IOT ready which means data on detection and classification can be sent to cloud.
DZUKI MiniAI Camera
The AI Camera Nano is powered by the Jetson Nano. This system uses a built in camera with fixed focal length. It has the same deep learning modules as installed in the DZUKI SuperAI camera but runs at a lower speed. Therefore it is suitable where speed is not critical. It is also designed to communicate with a PLC and is IOT ready.
AI Deep Learning Vision Development/Educational System
A general purpose Automated Optical Inspection (AOI). It can be installed with a conveyor or rotary table system. A self contained unit with monitor, cpu/gpu, PLC for external communication, LED lighting, camera and lens, wifi/ethernet and IOT enabled and mini keyboard. Options for application-specific camera / optics / illumination module. NO programming knowledge needed. For classifying OK or NG assembly, just take pictures of OK and NG assembly. Put them into two separate folders, train and then deploy. For detecting different types of defects, just take a large number of pictures of each type of defect, label them, train the system and then deploy. Automatically generates reject and accept signals to a PLC that can be used for communication with robots or other automation equipment.
An AOI system for inspecting metal strips. A self contained unit with monitor, cpu/gpu, PLC for external communication, LED lighting, two cameras, wifi/ethernet and IOT enabled and mini keyboard. Automatically generates reject and accept signals to a PLC that can be used for communication with robots or other automation equipment. Materials are fed through by an inbulit conveyor belt.
An AOI system with an open conveyor system. The DZUKI SuperAI or MiniAI camera system can be used depending on the speed of the detection or classification required. Suitable for research and development on deep learning AI or university laboratory classes on deep learning
DZUKI Deep Learning Image Classification Software
DZUKI Deep Learning Defect Detection Software
Defect detection involves drawing a bounding box around each defect of interest in the image and assigns them a class label. Our Deep Learning defect detection software was developed to make it easy to set up for deployment in a step by step approach. It's menu driven and runs on Windows 10 OS. For defect detection the steps involve capturing or importing image samples, labelling the defects of interest on the sample images, training the samples, evaluating and selecting the models for real time deployment. The models generated can be imported into the defect detection module in our AI Deep Learning cameras. Application is when the user needs to identify various defects for analysis in improving processes.
DEEP LEARNING DETECTION OF PCB DEFECTS
Six types of defect was trained using our DZUKI Deep Learning Detection Software namely missing hole, mouse bite, open circuit, short, spur, and spurious copper. Figure below shows examples of the detection of mouse bite and missing holes on different PCB design
DEEP LEARNING CORROSION CLASSIFICATION
According to the NACE International — The Worldwide Corrosion Authority, the global cost of corrosion is estimated to be US$2.5 trillion. Implementing corrosion management practices could save between 15-35% – around US$875 billion annually. The traditional maintenance process involves an engineer spending a month studying around 6,000 images to manually mark up the corrosion. Each oil rig has an average of 20,000 images, meaning that would take over four months of human analysis per rig. Worn equipment, rusty tanks and other evidence of wear and tear can cause unplanned shutdowns. This can lead to costly repairs or, worse still, environmentally devastating oil spills, resulting in significant economic loss. Manual processing and reviewing of photographic data is inefficient causing a backlog of maintenance. Also existing manual process cannot track and predict corrosion development over the years.
In this research the DZUKI Deep Learning Classification software was trained with hundreds of corrosion images categorized into 3 stages. Both pictures or video can be processed to produce annotated video or images of the probabilities of the different stages of corrosion.
DEEP LEARNING SYSTEM FOR DETECTING DIABETIC RETINOPATHY
WHO estimates that in year 2030, Malaysia would have 2.48 million people with DM (NHMS 2006). Within 20 years of diagnosis of DM, nearly ALL people with Type 1 DM and almost 2/3 of people with Type 2DM will have some degree of retinopathy. About 15,000 to 39,000 people lose their sight because of diabetes (NHMS 2006). It is the commonest cause of visual loss among working adults in Malaysia. In the early stages of DR, patients are usually asymptomatic and are unaware of their retinopathy changes. Screening is necessary to identify the group at risk of visual. Retinal screening contributes to early detection of diabetic retinopathy and timely treatment.
Our DZUKI Deep Learning Classification software was trained with a dataset of 4 levels of DR fundus images. A smartphone with an optical lens mounted on a 3D printed part (oDocs Fundus) can be used to stream video of the retina to the software which will predict the level of DR.
COVID-19 DEEP LEARNING DETECTION FROM X-RAY AND CT SCAN IMAGES
The Covid-19 pandemic has pushed the frontier of medicine to its limits. In particular the detection of Covid-19 and its variants is very crucial in its early stages. CT Scan and X ray images still remains the definitive diagnostic tool to detect its onset. Our DZUKI Deep Learning Classification software was trained with a dataset of X-Ray and CT scan images of patients with Covid and Non-Covid. Our system is able to achieve an accuracy of more than 95% accuracy of detection for both type of images. This demonstrates the versatility of our software with various types of images.
DEEP LEARNING FRUIT INSPECTION
Manual inspection of fruits is laborious. Automated Optical Inspection of fruits will reduce the time taken to grade fruits. Our DZUKI Deep Learning Classification software was trained with 100 image samples of rotten and fresh bananas respectively. The prediction model was tested on 2000 images of rotten and fresh bananas with an accuracy of 98%.
Dr Zahari Taha is a Chartered Engineer (UK), Member Institution of Engineering Designers (UK), Fellow Academy Sciences Malaysia and Fellow and Board Member Asia Pacific Industrial Engineering and Management Society. He was formerly a Professor at the Faculty of Engineering, Universiti Malaya and Faculty of Manufacturing Engineering, Universiti Malaysia Pahang. He has more than 30 years of experience in education, research and development in machine learning, automation and robotics, and simulation and modelling. Dr Zahari has published over 300 papers and graduated over 70 PhD and Masters students. He has developed numerous software and hardware solutions for industry and universities. Dr Zahari is a certified HRDF trainer.
Machine Design and Prototyping
Design and prototype development of new machines
Machine Vision Applications
Development of Machine Vision system for various applications such as defect detection, parts differentiation etc
Machine Learning Software Development
Software development for the application of machine learning algorithms
Industrial Automation and Robotics Application
Consultation and studies on industrial automation and robotics solutions
Simulation and Modeling
Consultation and studies on simulation and modeling of manufacturing systems
Development of IOT applications
Introduction to Machine Learning (One day Course)
Machine Learning involves the use of algorithms to build a good and useful approximation to data and then make a determination or prediction. However machine learning can be very complex and most people are not trained to program or do not have the time to do so. Orange is an open source machine learning and data visualization toolkit that does not require programming skills. It is suitable for both novice and expert. It is packed with features for data analytics and has a large toolbox for machine learning. In this course participants will have hands on training exploring various types of data with Orange. Participants are also encouraged to bring their own data. The course is opened to participants from any background.
Introduction To Industry 4.0 (One Day Course)
Industrial Revolution 4.0 is going to have a major impact on the future of Malaysia. Knowledge and skills of the core technologies of Industrial Revolution 4.0 are essential in order fulfil the requirements of the workforce of the future. This is particularly so when the economy is badly affected by COVID-19. Therefore, it is imperative to understand basic principles of Industry 4.0 to meet the challenges facing the manufacturing industry in post COVID-19. The concerted objective of the course is the basic knowledge for Industry 4.0 in manufacturing. The course will consist of four parts
● Design and Planning of Automated Production Facilities via Simulation and Modeling
● Automated Quality Control with Deep Learning and Machine Vision
● AGVs and Industrial Robots
● Industrial Internet of Things (IIOT)
An Introduction to Industrial Internet of Things (IIoT) (One Day Course)
The Internet of things (IoT) will have a significant impact on manufacturing. It helps to create new technologies to solve problems, enhance operations, and increase productivity. IoT can be explained as a network of physical objects, systems, platforms and applications that contain embedded technology to communicate and share intelligence with each other, the external environment and with people. (IoT) impact on Industrial Automation is very high and it makes us to use tablet computers, smart phones, virtualized systems, and cloud storage of data and so on. IoT generates a large amount of data for in manufacturing. This gives rise to Big Data. Big Data are large sets of data that may be analysed computationally to reveal patterns, trends and association, especially relating to human behaviour and interactions. The combination of Big Data and IoT is called Industrial IoT (IIoT) . The objectives of this course are to give an understanding of Industrial Internet of Things IIoT and to be able to develop basic applications of IIOT using Thingspeak and Blynk. Participants will also be introduce to Node Red and PLC IOT.
Introduction to Manufacturing Modeling and Simulation (One Day Course)
Industrial Revolution 4.0 is going to have a major impact on the future of Malaysia. Knowledge and skills of the core technologies of Industrial Revolution 4.0 are essential in order fulfil the requirements of the workforce of industrial revolution 4.0. One of the core technologies of Industrial Revolution 4.0 is Simulation.
Simulation involves experimenting with a model of an operations system on a computer as it progress through time for the purpose of better understanding and/or improving the system
Simulation allows a company to analyse and experiment with their processes in a virtual environment. The simulation model can be used to evaluate inventory, assembly, transportation and production to explore solutions for improvement at minimum cost.
The objectives of this course are
To give an understanding of how simulation software works and the process involved in a simulation project
To be able to collect and analyse the data required for a simulation model
To be able to perform simulation experiments, analyse the results and draw conclusions
IF YOU ARE INTERESTED IN THE ABOVE COURSES , PLEASE EMAIL US AT email@example.com
4-3A, Jalan Pandan 3/7, 55100 Kuala Lumpur, Malaysia
DR67, Kampung Kuala Pahang, 26660 Pekan, Pahang, Malaysia