Four points of Multi-Sigma

Simple execution of prediction, factor analysis, and multi-objective optimization with AI

Multi-Sigma, which combines multiple AI technologies, enables prediction, factor analysis, and multi-objective optimization on a single platform. It is a scalable and innovative AI analysis application with no limit to the number of variables to be predicted, meeting the needs of various fields.

Dramatic increase in R&D productivity

Automatically tunes neural networks to make highly accurate predictions from as little as 30 data points. No need for big data. It makes predictions based on the minimum amount of data required, and searches for the best solution from among a huge number of combinations of conditions.

Bringing the world's most advanced AI skills to people everywhere

Empower any person with AI skills without the cost of learning. No programming skills or data science expertise required. Anyone can perform the world's most advanced analysis with the same quality of analysis as experts.

Safe and no cost for installation and maintenance

Since it is built on Google Cloud Platform, you can be assured of its security. With a reasonable usage fee, you can perform AI analysis from the browser of your PC or tablet device.

Streamlining R&D with Innovative Design of Experiments

Many of the challenges faced in the field of research and development are probably “multi-input, multi-purpose systems” with multiple inputs (material input, temperature, time, etc.) and multiple outputs (quality, cost, environmental impact, etc.).
Multi-Sigma was born from the strong desire of AI researchers to bring breakthroughs in research and development with the power of AI, which has been done by trial and error and relying on experience.
The original “Innovative Design of Experiments” method, which incorporates AI techniques into the conventional PDCA cycle, efficiently leads to optimal solutions for multi-input, multi-purpose systems.

Translated with (free version)

Optimizing design to both improve functionality and reduce side effects

Design Optimization of an Artificial Heart (Hydrodynamic Levitation Centrifugal Blood Pump) Using Innovative Design of Experiments
(Modified from Figure 2 on page 98 and Figure 3 on page 99 of the Abstracts of the 122nd Annual Meeting of the Japanese Society for Quality Control)

Multi-Sigma has been successfully optimized from approximately 60 simulation data for an artificial heart design with 7200 possible experimental conditions.
The AIST research not only succeeded in searching for the optimal solution with 1/120th of the effort, but also contributed to the discovery of completely new knowledge that was previously unknown and unique to AI.

We developed an innovative design of experiments method that combines neural networks and multi-objective genetic algorithms in the framework of design of experiments.
There were 7,200 combinations of the four input conditions of the artificial heart (number of grooves: 3-18, 16 ways; groove angle: 10-180 degrees, 18 ways; groove inlet depth: 0.05-0.25 mm, 5 ways; groove outlet depth: 0.05-0.25 mm, 5 ways). It is very difficult to conduct experiments and numerical simulations for all of these combinations in terms of time and money.

Using the design of experiments method with Multi-Sigma, we were able to optimize the generating force of the hydrodynamic bearing and the red blood cell damage coefficient of an extracorporeal hydrodynamic levitation centrifugal blood pump based on the results of about 60 analyses.

The details of this method were presented at the 122nd Research Conference of the Japanese Society for Quality Control on May 23, 2020.

Translated with (free version)