Logiciel Edge Computing

Application Edge

MELSOFT MaiLab

30-day free trial

30-day trial of MELSOFT Mailab

MELSOFT MaiLab

MaiLab is a software offering a variety of machine learning and statistical analysis methods, including AI features like deep learning and multiple regression analysis. It enables data analysis for various purposes without requiring programming, making data analysis solutions easy to implement.

One software for all types of data analysis

This software alone covers the phase of analysis of production data in the office and the phase of diagnosis in production based on these analysis results, so that it is possible to apply the learning models obtained from data analysis directly online in the production.

AI data scientist - an AI-based analysis supporting system for everyone

Customer benefits:

Very short training phase for the software as no specialized knowledge is required, anyone can do data analytics

MaiLab is supporting the customer in all phases of the data analysis project

Customers benefit from MaiLab as many companies see a lack of manpower for data analyzing

Customers are empowered to improve their production efficiency quickly and efficiently

An UI bringing you data analytics with a great experience

Customer benefits:

Quick Return-on-invest as MaiLab Software is a single tool for both Offline Analysis and Real-time diagnostic including direct feedback to the production site. Rich possibilities for data visualization.

Longevity and future-proof design and operation of MaiLab through integrated open concepts like Python programming language or web-based environment.

Flexibility through different licensing schemes available (Yearly subscription for OPEX, perpetual model for CAPEX) and many different application scenarios.

Tech Talks - MaiLab

Analysis Process

MELSOFT MaiLab is a tool that enables easy data analysis in 4 basic steps.

Offline analysis

Step 1: Data set creation

First, read the data to be analyzed into MELSOFT MaiLab and register them. A group of registered data is called a “data set”.The data set can be shown in various kinds of graphs, so that it can also be easily checked by human eyes before performing diagnosis using AI.

Step 2: AI creation

Learning from the data set is performed.A model that enables diagnosis of unknown data is called “AI”.When “What you want to do (objective)” is selected, the regularity and rules of the data are automatically derived, and MELSOFT MaiLab creates the “AI”.

Real-time diagnosis

Step 3: Task creation

Settings for performing diagnosis of unknown data are called a “task”.MELSOFT MaiLab will define the data input/output methods and threshold values for whether diagnostic results are good or bad. The accuracy is displayed as a score, which serves as a guideline for judgment.

Step 4: Task execution and monitoring

You can execute tasks and monitor the diagnosis status of unknown data.Deployment to equipment can be easily performed with just a click. Data flow and good or bad judgment status can be confirmed on a graphical display via the learning server.

Offline Analysis

Prepare the data. (Data set creation)

In order to analyze the data and create the diagnosis model, it is necessary to register the data subject to analysis in MELSOFT MaiLab. A group of registered data is called a “data set”. By registering the data set, the data can be visualized in tables or graphs, and diagnosis models can be created.

Data registration can be performed by simple mouse operations

The original file of data to be registered as a data set is called the “data source”.Data sources which can be registered are CSV-format and TSV-format text files.

Create diagnosis rules. (AI creation)

Perform pre-processing of the data set and create AI by performing learning according to analysis methods.

Interactive and easy. Automatic AI creation saves time and effort

Automatic

MELSOFT MaiLab selects the optimum pre-processing and analysis methods based on the objectives and data set contents, and automatically creates the AI. Select this when you don’t know what analysis method to use for what you want to do (objectives).

Manual

In this method, you select the analysis methods yourself and create the AI. Select this when the appropriate method for what you want to do (objectives) is clear.

Complete the AI

Proceed to creating the task to perform real-time diagnosis while referring to the displayed scores and comments, When learning has been completed, AI creation will be completed. You can manually change the completed AI to customize it to increase reliability.

You can customize the AI to increase its accuracy

In MELSOFT MaiLab, each process of AI is performed in a block, and the AI processing flow is created by connecting the blocks. You can edit the AI flow prepared by the AutoML function to freely customize it or create an original AI from scratch.

Original processing can be performed with Python blocks

MELSOFT MaiLab is also equipped with function expansion blocks that are useful for customizing learning models. You can also perform coding in Python, which is often used in data analysis. By performing customization, you can create learning models with higher accuracy.

Real-time Diagnosis

Implementing in the device (task creation)

A group of processes (process flow) using the created AI to perform diagnosis on unknown input data and output the diagnosis results is called a “task” in MELSOFT MaiLab.* A simple task can be automatically created by setting the necessary parameters for the operation of each process.

*There are 2 types of tasks: simple and advanced. For details, please refer to the manuals.

When using Mitsubishi Electric FA equipment, devices can be specified directly

MELSOFT MaiLab has high affinity with Mitsubishi Electric FA equipment.Since direct specification of compatible devices can be performed, device deployment (arrangement) can also be performed easily.

The status is shown in real time during task execution

Diagnosis results are shown in line graphs and pie charts.Diagnosis results and data input to AI are shown in table format.

System Configuration/License/Operating Environment

System Configuration

Data collection and diagnosis can be started in MELSOFT MaiLab with just a basic license. In addition, systems can be freely configured according to the scale of facilities, increases in the number of analysis users, etc.

License

Operating Environment

Learning operating environment
In the minimum operating environment, it is possible to execute methods such as multiple regression analysis, etc. with relatively low calculation processing when no other tools are running. To execute methods such as deep learning, etc. that require lots of calculation processing, the recommended operating environment is necessary.

Item	Minimum	Recommended
Computer	PC, industrial PC, server	PC, industrial PC, server
CPU	Intel® Core™-i3 equivalent or better	Intel® Core™-i7 equivalent or better*1
Memory	4 GB or more	16 GB or more*1
OS	Windows® 10 (Pro, Enterprise, IoT Enterprise)Windows Server 2019 (Datacenter, Standard, Essentials)Windows Server 2016 (Datacenter, Standard, Essentials)	Windows® 10 (Pro, Enterprise, IoT Enterprise)Windows Server 2019 (Datacenter, Standard, Essentials)Windows Server 2016 (Datacenter, Standard, Essentials)
	64-bit	64-bit
Available storage space	16 GB or more	64 GB or more

Collection/diagnosis operating environment

Item	Minimum	Recommended
Computer	PC, industrial PC, server	PC, industrial PC, server
CPU	Intel® Core™-i3 equivalent or better	Intel® Core™-i7 equivalent or better*1
Memory	4 GB or more	16 GB or more*1
OS	Windows® 10 (Pro, Enterprise, IoT Enterprise)Windows Server 2019 (Datacenter, Standard, Essentials)Windows Server 2016 (Datacenter, Standard, Essentials)	Windows® 10 (Pro, Enterprise, IoT Enterprise)Windows Server 2019 (Datacenter, Standard, Essentials)Windows Server 2016 (Datacenter, Standard, Essentials)
	64-bit	64-bit
Available storage space	16 GB or more	32 GB or more

*1 Required when executing not just methods such as multiple regression analysis, etc. with relatively low calculation processing, but methods such as deep learning, etc. that require lots of calculation processing.