MathWorks offers extensive support for deploying machine learning models
Published on : Wednesday 06-03-2024
Prashant Rao, Head of Application Engineering, MathWorks India and Philipp Wallner, Industry Manager – Industrial Automation, MathWorks.
How does MathWorks envision the integration of artificial intelligence (AI) in industrial automation and machinery, and what specific tools or platforms does MATLAB offer to facilitate this integration?
MathWorks foresees the integration of AI in industrial automation to enhance efficiency, reliability, and innovation. The company envisions AI streamlining operations, enabling predictive maintenance, and providing deep insights into process optimisation. To facilitate this integration, MATLAB® offers a comprehensive suite of tools and platforms, such as the Machine Learning Toolbox, Neural Network Toolbox, and Predictive Maintenance Toolbox. These tools provide an accessible environment for designing, training, and deploying AI models. Simulink® is the platform for Model-Based Design, which is crucial for developing complex dynamic systems that integrate AI algorithms with control systems and other machinery components.
Using MATLAB and Simulink, engineers can embed AI and data science algorithms in industrial automation applications without expertise in data science or machine learning. MATLAB and Simulink enable engineers to address the growing complexity of modern production equipment and the demand for higher flexibility.
Industrial Automation and Machinery engineers use Model-Based Design in MATLAB and Simulink to:
Design and test machine controls and supervisory logic
Run automatic tests on equipment functions.
Develop and train AI algorithms for predictive maintenance and operations optimisation.
Generate real-time code (C/C++, IEC 61131-3) for industrial controllers and PLCs.
Automated inspection and defect detection systems use AI to inspect manufacturing parts for failures and defects. This approach enables industries to automatically detect flaws on manufactured surfaces such as metallic rails, semiconductor wafers, and contact lenses.
Can you provide examples of successful applications where MATLAB or Simulink has been utilised to implement AI algorithms for optimising industrial
processes or improving machinery performance?
One notable example is the use of MATLAB in the energy sector for optimising power grid operations. AI algorithms developed with MATLAB have helped forecast energy demand and manage supply more efficiently. In manufacturing, Simulink has been instrumental in designing control systems that integrate AI to improve the performance of robotic arms, resulting in more precise and efficient assembly lines. Another application is in the automotive industry, where engineers use MATLAB and Simulink for developing AI-driven systems that enhance vehicle dynamics and fuel efficiency.
As the industry adds more AI for defect detection, vision-based AI systems become an important component of manufacturing machinery. In the manufacturing industry, automated visual inspection systems with high-resolution cameras efficiently detect microscale or even nanoscale defects that are difficult for human eyes to pick up. However, false detection sometimes happens when various defects are present, which is a major challenge. Software with deep learning technology such as MATLAB and the Computer Vision Toolbox™ Automated Visual Inspection Library is playing a more important role.
For example, Musashi Seimitsu Industry, an automotive parts manufacturer, was inspecting approximately 1.3 million parts per month with manual visual inspection. Using MATLAB® to develop deep learning-based approaches to detect and localise different types of anomalies, they built an automated visual inspection system for inspecting bevel gears. The updated approach is expected to considerably reduce the company’s workload and costs.
Another example is Hyundai Steel. Hyundai harnessed MATLAB® to enhance their steel classification process with AI. They crafted a labelling algorithm using unsupervised machine learning for efficient data annotation, significantly reducing the need for manual corrections. Using the labelled data, they trained a neural network for pixel-wise image classification to distinguish various steel types. The AI model achieved an impressive 85% prediction accuracy, comparable to human experts, but with the added benefits of consistency and repeatability. This innovation exemplifies the integration of AI in industrial settings, where MATLAB® enabled the automation of complex tasks, leading to increased efficiency and consistent production quality.
What support and resources does MathWorks provide for developing and deploying machine learning models on embedded systems commonly used in industrial automation, such as PLCs or edge devices?
MathWorks offers extensive support for deploying machine learning models on embedded systems. This includes MATLAB® Coder™and Simulink® Coder™, which generate C and C++ code from MATLAB® and Simulink® models, respectively. GPU Coder™ can be used to implement and deploy deep learning models on NVIDIA® CUDA® GPUs. Or MATLAB® Coder™ and Embedded Coder® can be used to generate C code for deployment on Intel® and Arm® boards. Vendor-optimised libraries create deployable models with high-performance inference speed.
The generated code is designed to be production-ready for 24/7 operation and hardware-independent, allowing for integration with PLCs or edge devices from all major vendors across various platforms. For specialised hardware, HDL Coder and GPU Coder allow for the generation of code that can run on FPGAs or GPUs. MathWorks also provides Embedded Coder for optimising code efficiency, which is critical for resource-constrained environments. Additionally, MathWorks offers comprehensive documentation, webinars, user forums like MATLAB® Community, which offer knowledge sharing and libraries that accelerate AI system development, and training courses to assist developers in deploying their machine learning models on embedded systems.
In the context of predictive maintenance, how can MATLAB® assist in creating models that leverage AI techniques to predict equipment failures and optimise maintenance schedules for industrial machinery?
MATLAB assists in predictive maintenance by providing a rich set of tools for data preprocessing, feature extraction, and machine learning. Engineers can use Predictive Maintenance ToolboxTM to develop models that analyse historical sensor data, detect anomalies, and predict when equipment might fail. These models can be trained using machine learning algorithms available in MATLAB to recognise patterns that precede failures. By deploying these predictive models, companies can schedule maintenance more effectively, reducing downtime and extending the life of their machinery. For example, Hyundai Steel first created a labelling algorithm to label data for neural network training. They also used unsupervised machine learning to help them cluster certain objects in the images and to semi-automate this labelling process. So labelling was very efficient. They only had to manually correct small parts of the image. And then they trained, in this case, another neural network, which pixel-labelled this image and classified different types of steel. After implementing this model, the prediction accuracy was about 85%, which is comparable to human precision.
Are there any specific features or updates in recent MATLAB® releases that cater to the evolving needs and challenges of AI implementation in the industrial automation sector?
Recent MATLAB releases have increasingly focused on lowering the bar and enabling engineers and scientists to leverage technologies such as AI and deep learning easily. Apps enable a low-code/no-code approach to following established workflows and leveraging the technologies. Additionally, the new releases focus on enhancing deep learning capabilities, including the introduction of new neural network architectures, training options, and performance improvements for both training and inference. Updates have also improved the integration with deep learning frameworks like TensorFlow and PyTorch, allowing for easier model exchange. Additionally, there have been advancements in automated machine learning (AutoML), enabling engineers to find optimal models with less manual effort. These features and updates address the industrial sector's need for more robust, efficient, and scalable AI solutions.
How does MathWorks address the challenges of real-time processing and control in AI applications for industrial automation, particularly in scenarios where low-latency decision-making is critical?
MathWorks addresses real-time processing and control challenges by providing tools that enable the design and implementation of real-time systems. Simulink® Real-Time™ allows for the creation of real-time prototypes and hardware-in-the-loop (HIL) testing, which is essential for developing low-latency control systems. Additionally, MATLAB's support for code generation and integration with embedded systems ensures that AI algorithms can be deployed on hardware capable of meeting the stringent timing requirements of industrial automation applications.
Can you share insights into the role of MATLAB® in facilitating the development and validation of AI algorithms for robotics used in industrial automation, such as autonomous vehicles in manufacturing plants?
MATLAB plays a crucial role in the robotics field, particularly for developing and validating AI algorithms. Robotics System Toolbox™ provides tools for designing and testing robotics algorithms, including motion planning and control, which are essential for autonomous vehicles in manufacturing plants. Simulink, with its ability to simulate complex systems and integrate with physical hardware, is invaluable for validating the performance and safety of these AI-driven systems before deployment. This combination of tools ensures that industrial robots can operate efficiently, safely, and autonomously within their environments.
Prashant Rao is a regular contributor at industry forums, sharing his views around megatrends in technology and how Artificial Intelligence (AI) is getting adopted across industries and other technologies. He works closely with the academic community to help develop analytical and AI-related skills that make them industry-ready. He can be reached at [email protected]
As Industry Managers for Medical Devices, Industrial Automation & Machinery, and Utilities & Energy at MathWorks, Philipp and his team work closely with their worldwide offices, with innovation leaders among our key customers and with our development teams to sharpen MathWorks' strategy and offering in these industries. Prior to joining MathWorks, he worked in the machine builder industry where he held different engineering and management positions. Philipp has a M.Sc. in Electrical Engineering from Graz University of Technology and an Executive MBA in Project and Process Management from Salzburg Management & Business School.
_______________________________________________________________________________________________
For a deeper dive into the dynamic world of Industrial Automation and Robotic Process Automation (RPA), explore our comprehensive collection of articles and news covering cutting-edge technologies, robotics, PLC programming, SCADA systems, and the latest advancements in the Industrial Automation realm. Uncover valuable insights and stay abreast of industry trends by delving into the rest of our articles on Industrial Automation and RPA at www.industrialautomationindia.in
