MathWorks Japan List of Products and Services

With the realization of "smart factories" through the AI x IoT transformation in the manufacturing industry, interest in predictive maintenance is growing daily. Predictive maintenance allows for the monitoring of equipment conditions to prevent future equipment failures. By using data from equipment sensors, it is possible to identify the root causes of failures and predict the time until failure using classification, regression, and time series analysis. It also helps in identifying complex machine issues and determining which parts need repair or replacement. This minimizes downtime and maximizes the lifespan of the equipment. This ebook provides explanations of terms, examples, tutorials, and access to trial software to help you get started with developing predictive maintenance algorithms using MATLAB.

• Software (middle, driver, security, etc.)

Recent news has highlighted how reinforcement learning algorithms have defeated professionals in games like Go, Dota 2, and Starcraft 2. Reinforcement learning is a type of machine learning that enables the use of artificial intelligence in complex applications such as video games, robotics, and self-driving cars. If you are interested in leveraging reinforcement learning techniques in your projects but have never used them before, where should you start? This ebook will help you get started with reinforcement learning in MATLAB and Simulink by explaining terminology and providing examples, tutorials, and trial software.

• Software (middle, driver, security, etc.)

In recent years, AI and deep learning have been applied in various fields, and the number of companies successfully improving their operations has been increasing. However, there are also many cases where projects have stalled at the Proof of Concept (PoC) stage. What differences exist in the approaches of the companies that are succeeding? It is essential to understand the strengths and weaknesses of AI and deep learning, decide on a theme, and devise ways to incorporate it into existing workflows. In this ebook, we will introduce case studies from the following seven industries that have successfully implemented deep learning using MATLAB: - Manufacturing: Visual Inspection - Infrastructure Maintenance - Chemicals and Cosmetics - Automotive and Control - Healthcare - Aerospace - Education

• simulator

MATLAB can be used for modulation identification and target classification, as well as for communication applications. The white paper explains the following: 【Contents】 - Synthesizing radar and communication waveforms for labeling - Generating radar signals for moving objects - Training deep networks with synthetic data - Waveform modulation ID and target classification using deep learning and machine learning - Testing the system with data collected from software-defined radio and radar systems

• simulator

Based on extensive experience working collaboratively with engineers in the automotive industry, MathWorks consultants have created 11 best practices for ISO 26262 development using model-based design. This ISO 26262 guide provides proven best practices for model architecture that can be used to reduce verification and implementation workload when complying with the ISO 26262 functional safety standard. [Contents] - Management of model interface complexity and data exchange - Generation of file-split code with no interference (FFI) - Improvement of overall efficiency during the verification, validation, and documentation phases

• Other embedded systems (software and hardware)

MATLAB and Simulink support the development and verification of robotics and advanced, complex autonomous mobile robots. [Contents] - Design of hardware platforms - Acquisition of sensor data utilizing ROS and ROS 2 - Object recognition through image processing, deep learning, point cloud processing, and sensor fusion - Mapping and self-localization using SLAM and Visual SLAM - Path planning and path following - Simulation, verification, and hardware implementation By utilizing MATLAB and Simulink, algorithms in complex areas such as cognition, judgment, and control can be developed and verified in a single environment. Additionally, automatic code generation allows for seamless implementation to real-time hardware, GPUs, and embedded CPUs.

• Other embedded systems (software and hardware)

Do you want to develop AUTOSAR ECU software while minimizing costs, risks, and issues? Simulink provides APIs for workflow automation in addition to AUTOSAR functionalities. Based on extensive experience assisting automotive manufacturers and OEM suppliers, MathWorks consultants present ten best practices for applying AUTOSAR using Simulink and model-based design. Download the guide to see the following content: 【Contents】 - Strategies for migrating Simulink models to AUTOSAR: migration from scratch, full migration, model variation - Selection of data management strategies (Simulink, AUTOSAR authoring tools, or external tools) and establishment of modeling standards for AUTOSAR adoption - Simulation and automatic code generation for AUTOSAR and non-AUTOSAR targets from a single Simulink model - Future-oriented migration plans for evolving AUTOSAR and ISO 26262 standards

• Software (middle, driver, security, etc.)
• Other embedded systems (software and hardware)

In the development of autonomous driving and ADAS (Advanced Driver Assistance Systems), there are many technologies that need to be considered and developed across a wide range of areas related to "perception, decision-making, and operation." This white paper systematically and thoroughly explains the increasingly complex technological domains of autonomous driving and ADAS. Furthermore, it highlights the following seven important applications and functions in autonomous driving development, introducing features and solutions using MATLAB/Simulink for each: - Image processing and deep learning - LiDAR signal processing - Sensor fusion - Advanced driving control (model predictive control, reinforcement learning) - Generation of driving scenarios for various algorithm verification - External collaboration (Unreal integration, ROS integration) - Code generation

• Software (middle, driver, security, etc.)

In model-based development, the system model becomes the central focus of development, from recording requirements to design, implementation, and testing. At any stage, the model can be simulated to instantly verify system behavior without risks or delays, and without the use of expensive hardware, allowing for the testing of multiple what-if scenarios. But how does model-based development actually work in practice? And how should one get started? [Contents] - Basics of model-based development - Best practices for implementation - User case studies that achieved reduced development time, minimized issues related to component integration, and delivered higher quality products.

• Software (middle, driver, security, etc.)

The Battery Management System (BMS) includes a wide range of logic to ensure the safety, performance, and lifespan of the battery, such as protection logic for overcharging and over-discharging, estimation logic for State of Charge (SOC) and State of Health (SOH), and cell balance control logic. To verify the validity and reliability of the BMS logic, it is necessary to conduct system-level test verification under various conditions, including abnormal states. With simulation, testing and verification can be carried out safely and efficiently using a model that simulates the behavior of actual batteries. In the development of BMS using Simulink, the following can be achieved: - Monitoring cell voltage and temperature - Estimating State of Charge (SOC) and State of Health (SOH) - Limiting power input and output for thermal and overcharge protection - Controlling the charging profile - Balancing the charging state of the cells - Disconnecting the battery pack from the load as needed *Currently, we are offering a "white paper" that explains the benefits and procedures of BMS development, and a "seminar material" that introduces our solutions. You can view them immediately from the [PDF download].

• others
• simulator
• Development support tools (ICE, emulators, debuggers, etc.)

When designing digital control for power converters, there are many reasons to build models and simulate them. For example, the behavior of the converter during fluctuations in power supply and load, circuit topologies that combine passive components (such as resistors and capacitors) with active components (such as power transistors), and the design of feedback and monitoring control algorithms necessary to control voltage and meet stringent design requirements. This white paper introduces 10 methods for rapidly designing digital control power converters using "Simulink." [Contents] - Simultaneous simulation of analog and digital components - Automation of controller analysis and tuning in the frequency domain - Simulation of control algorithms to improve power quality - Verification of fault detection, mode logic, and monitoring control across all operating conditions - Validation of power converter operation within larger electrical systems *For more details, please refer to the PDF document or feel free to contact us.

• Software (middle, driver, security, etc.)

This ebook outlines the basics of machine learning and introduces supervised and unsupervised methods as a sequel to "Machine Learning with MATLAB." Using a heart sound classifier as an example, it explains the entire workflow for developing a real machine learning application, from data loading to the distribution of the trained model. It presents essential techniques for deriving accurate models at each learning stage and supports the acquisition of more challenging learning tasks, such as algorithm selection, model parameter optimization, and avoiding overfitting. [Contents (excerpt)] ■ Introduction ■ Development of Heart Sound Classification Application with MATLAB ■ Essential Tools for Machine Learning ■ Want to know more? *For more details, please refer to the PDF materials or feel free to contact us.

• Software (middle, driver, security, etc.)

The important procedure in the development of predictive maintenance algorithms is the identification of situational indicators. These are characteristics of systems that change behavior in a predictable manner as the system deteriorates. Situational indicators help distinguish between failure and normal operation. The materials include an overview of situational indicators, as well as the predictive maintenance workflow. [Contents] ■ What are situational indicators ■ Visual exercises ■ Feature extraction using signal-based methods ■ Predictive maintenance workflow ■ Overview of good features and why they are important *For more details, please refer to the PDF materials or feel free to contact us.

• Software (middle, driver, security, etc.)

This ebook is a sequel to "What is Deep Learning?" which provides an overview of deep learning. This time, we will explain practical methods for deep learning. "MATLAB" offers tools and functions for managing data, labeling, monitoring learning progress, and visualizing results, making it easy to perform time-consuming and complex tasks involved in deep learning. [Contents] ■ Introduction ■ Practical Example 1: Training a model from scratch ■ Practical Example 2: Transfer learning ■ Practical Example 3: Semantic segmentation ■ Data other than images *For more details, please refer to the PDF document or feel free to contact us.

• Software (middle, driver, security, etc.)

DC-DC converters have various topologies such as buck, boost, SEPIC, and Ćuk that convert voltage levels. With digital control, it is possible to manage various power sources and loads in a DC-DC converter to maintain the desired power quality within the operating range. This white paper explains how to develop digital control for DC-DC converters using system-level simulation in "Simulink." [Contents] - Modeling and simulation of SEPIC - Verification of converter power loss and thermal behavior - Design of digital controllers - Generation of code for implementation on microcontrollers - Summary *For more details, please refer to the PDF document or feel free to contact us.

• Software (middle, driver, security, etc.)

In various industries, the increasing reliance on energy storage through battery packs has heightened the importance of Battery Management Systems (BMS) that can ensure maximum performance, safe operation, and optimal lifespan under diverse charge and discharge conditions as well as environmental conditions. This white paper explains how to develop BMS algorithms and software by performing system-level simulations using "Simulink." Model-based design using "Simulink" allows for a better understanding of battery pack behavior, adjustment of software architecture, testing of various operational cases, and early execution of hardware tests, thereby reducing design errors. [Contents] ■ Desktop Simulation: Modeling of BMS Software ■ Modeling and Characterization of Battery Cells ■ Modeling of Power Electronics and Passive Components ■ Development of Monitoring Control Algorithms *For more details, please refer to the PDF document or feel free to contact us.

• Software (middle, driver, security, etc.)

Supervised learning algorithms use a known input dataset (training dataset) and the known responses (outputs) for that data to train a model, enabling the model to make reasonable predictions for new input data. When there is existing response (output) data for the event you are trying to predict, supervised learning is used. The materials include methods of supervised learning, as well as binary classification, multi-class classification, and general classification algorithms. [Contents] ■ When to consider supervised learning ■ Methods of supervised learning ■ Choosing the appropriate algorithm ■ Binary classification and multi-class classification ■ General classification algorithms *For more details, please refer to the PDF materials or feel free to contact us.

• Software (middle, driver, security, etc.)

Unsupervised learning is useful when you want to investigate data in detail but do not have a specific goal in mind regarding what you can learn from the results, or when the information contained in the data is unclear. This document provides a detailed explanation of the hard clustering and soft clustering algorithms used in unsupervised learning. Additionally, we will introduce tips for selecting the appropriate algorithm for your data and methods for reducing the number of features in your dataset to improve model performance. [Contents] ■ When to consider unsupervised learning ■ Unsupervised learning methods ■ Common hard clustering algorithms ■ Common soft clustering algorithms ■ Improving models through dimensionality reduction ■ Common dimensionality reduction techniques *For more details, please refer to the PDF document or feel free to contact us.

• Software (middle, driver, security, etc.)

In machine learning, it is rare to progress straight and unhesitatingly from start to finish. You will often find yourself repeatedly trying various ideas and methods. This document explains the systematic workflow of machine learning while focusing on several key decision points. [Contents] ■ It is rare to progress in a straight line ■ Challenges in machine learning ■ Points to consider before starting ■ Overview of the workflow ■ Training a model for classifying physical activities *For more details, please refer to the PDF document or feel free to contact us.

• Software (middle, driver, security, etc.)

Machine learning means teaching computers things that are taken for granted by humans and animals. In machine learning algorithms, information is learned directly from data using computational methods without relying on pre-defined equations called models. These algorithms are designed to adaptively improve their performance as the number of samples available for learning increases. The materials include the types of machine learning, how to determine the algorithms to use, and real-world examples. [Contents] ■ What is machine learning? ■ More data, more questions, and better answers ■ Types of machine learning ■ Supervised learning ■ Unsupervised learning *For more details, please refer to the PDF materials or feel free to contact us.

• Software (middle, driver, security, etc.)

Deep learning is one of the methods of machine learning, where the model learns classification methods directly from images, text, and audio. In this ebook, we provide a concise explanation of the fundamental techniques. Deep learning is not difficult at all, and you can start right away even if you are not an expert. [Contents (excerpt)] ■ What is deep learning? ■ Application areas of deep learning ■ How deep neural networks work ■ Learning methods for deep neural networks ■ About convolutional neural networks *For more details, please refer to the PDF materials or feel free to contact us.

• Other information systems

The background for the growing attention on anomaly detection and predictive maintenance includes factors such as the increasing complexity of failures and the rising costs of maintenance. This document introduces the construction of predictive maintenance systems. Additionally, it explains the image of predictive maintenance systems, case studies of their construction, and the construction workflow. 【Contents (Excerpt)】 ■ Image of Predictive Maintenance System ■ Predictive Maintenance by Three "Functional Levels" ■ Background for the Attention on Predictive Maintenance ■ Case Studies of Predictive Maintenance System Construction ■ Workflow for Constructing Predictive Maintenance Systems *For more details, please refer to the PDF document or feel free to contact us.

• Other information systems
• Software (middle, driver, security, etc.)

This document introduces predictive maintenance using MATLAB, including three methods for predicting Remaining Useful Life (RUL). "MATLAB" is software that combines a desktop environment suitable for iterative analysis and design processes with a programming language that directly represents matrix and array mathematics. Additionally, it includes explanations of Remaining Useful Life (RUL) and the differentiation of models used to predict RUL. 【Contents (excerpt)】 ■ What is Remaining Useful Life (RUL)? ■ Three methods for predicting RUL ■ Quick reference: Differentiating models for predicting RUL ■ Applying RUL predictions to the field ■ Related information *For more details, please refer to the PDF document or feel free to contact us.

• Other information systems
• Software (middle, driver, security, etc.)

The initiative of predictive maintenance, which involves predicting the timing of equipment failures and implementing maintenance, is gaining attention. From the perspective of complementing the intuition, skills, and experience of engineers with AI, it is expected that the demand for predictive maintenance will further increase in the future. This white paper introduces examples of effective features for detecting anomalies from time series data. Additionally, it includes discussions on the limitations of machine learning and features from the perspective of system integration. [Contents (excerpt)] ■ Limitations of machine learning ■ Examples of features that enable failure prediction ■ Features from the perspective of system integration ■ Summary *For more details, please refer to the PDF document or feel free to contact us.

• Software (middle, driver, security, etc.)
• Other information systems

In recent years, the utilization of data has been advancing in various fields such as manufacturing and infrastructure, and the efforts in "Predictive Maintenance" are gaining attention. This white paper introduces case studies of predictive maintenance system development and related technologies. [Contents (excerpt)] - Insights hidden in large amounts of sensor data - Case studies of predictive maintenance system development - Predictive maintenance workflow using MATLAB - Summary *For more details, please refer to the PDF document or feel free to contact us.

• Software (middle, driver, security, etc.)
• Other information systems

In this ebook, we explain the key differences between deep learning and machine learning approaches. When considering whether to start with deep learning or machine learning, we introduce three important factors (project, data, hardware). Additionally, assuming a basic understanding of AI technology, we address the often confusing question of "which algorithm should I use?" once you get started. [Contents (excerpt)] ■ Introduction ■ Terminology Explanation ■ Project ■ Data ■ Hardware ■ Conclusion *For more details, please refer to the PDF document or feel free to contact us.

• Other information systems

This white paper explains the basics of deep learning, as well as three examples of signal processing (voice command recognition, remaining useful life prediction, and signal noise reduction). Through these examples, it describes how deep learning using MATLAB can help perform signal processing tasks more quickly and achieve more accurate results. MATLAB is software that combines a desktop environment suitable for iterative analysis and design processes with a programming language that directly represents matrix and array mathematics. [Contents (excerpt)] ■ Introduction ■ Basics of Deep Learning ■ Deep Learning Networks ■ Choosing a Network ■ Considerations Regarding Signal Data *For more details, please refer to the PDF document or feel free to contact us.

• Software (middle, driver, security, etc.)

In this ebook, we will introduce essential trial-and-error methods for achieving a recognition accuracy of 99% in image classification through neural network training. Why not start with deep learning using a simple sample? 【Contents (excerpt)】 ■ Introduction ■ Practical Example 1: Training a model from scratch 　1. Accessing the data 　2. Creating and configuring network layers 　3. Training the network 　4. Verifying the network's accuracy, etc. *For more details, please refer to the PDF document or feel free to contact us.

• others

ThingSpeak(TM) is a cloud-based IoT platform for prototyping and small-scale operational applications. Data can be exchanged between various edge devices, including Raspberry Pi, and ThingSpeak using MQTT or REST API. Live data can be visualized instantly on an internet-connected web browser, enabling remote monitoring such as creating and sending alerts. 【Benefits of Use】 ■ Data can be checked and analyzed from anywhere on the internet ■ Build low-cost experimental systems using Arduino/Raspberry Pi ■ Check program completion/data anomalies on your smartphone *For more details, please download the PDF or feel free to contact us.

• Other embedded systems (software and hardware)