What we provide you about the MT System
Features of MT System
Normal data and correlation networks
There are many issues to be solved such as "normal/abnormal judgment from waveform data" and "prediction of when a failure will occur based on multiple measurement data".
The MT system learns only normal data and calculates the distance to the target from there. The set of normal data is called the “Unit Space" and the Mahalanobis Distance(MD) is calculated using the correlation between items.
If the MD to the target data is large, then it is out of the normal balance. Also, by knowing the speed of the MD, we can estimate when a failure will occur.
The figure shows the "correlation network" created by the normal data. The circles represent items and the lines represent correlations. Correlation is an indicator of mutual relationship. In the case of the relationship between faucet opening and water volume, the correlation is a large value (close to 1).
Since only normal data is needed for learning, the amount of data required is small.
MT systems, especially MT methods, are often compared to deep learning, a representative of artificial intelligence (AI). For more information on the differences between them, see the followingThe MT method is used in many situations, especially in the field of manufacturing.
Similarities and differences are described.
Excellent features of MT System
The advantages of the MT system are its clarity, lightness, and accuracy of results. The output and cause diagnosis results obtained by the MT system can be easily understood by the engineers, making it easier to respond to abnormalities.
Cost for implementation
Angletry provides a wide variety of software lineup from entry to production use, as well as expert consulting services.
Our software solves the problem of multicollinearity, which has been an issue with the Mahalanobis distance, in a unique way, and performs highly accurate calculations for all kinds of data. Our introductory software MTRT-AddIns runs on Excel, and provides the five most frequently used calculation methods. First, you can check which calculation method is appropriate, etc. at the beginning stage.
Furthermore, you can efficiently verify with ultra-fast computing software such as ATMTS.
We also provide feature extraction software from waveforms(WaveTool).
You can verify the best method in the shortest period of time and at the lowest cost, and proceed step-by-step to on-site implementation. It has been used for many applications, including vibration problems and rocket autonomous diagnosis.
In the MT system, feature extraction from images and waveforms was also proposed as a set. It is a feature called differential property (amount of change) and integral property (amount of existence). It is highly versatile and effectively extracts features from any time series or image data.
Dr. Taguchi applied for a patent in the U.S. for his idea on image features, which was granted (U.S. Patent No. 5,684,892: see below). However, Dr. Taguchi filed the application out of curiosity to see how long it would take for the U.S. patent to be granted. Feature extraction from waveforms can be implemented with WaveTool.
MT System and Quality Engineering
The MT system is a recognition and prediction technology proposed by Dr. Genichi Taguchi, the founder of Quality Engineering.
He hoped to apply the theories of Dr. Mahalanobis, an Indian statistician with whom Taguchi had a close relationship for 20 years, to Quality Engineering.
Inspired by the words "Happy families are uniformly happy, but unhappy families are unhappy in their own way" in the novel Anna Karenina by the Russian writer Tolstoy, he developed the MT system (MTS) applying Mahalanobis distance. A happy family corresponds to a healthy and normal state, while an unhappy family to a defective and unsuccessful state. As the phrase "in their own way" suggests, the causes of defects in manufacturing are varied, and unknown defects may eventually occur. If we focus only on the normal state and use it as a reference point, we can define it as abnormal if the Mahalanobis distance is far.
Quality engineering is the field of study that seeks robustness, stability, and minimizing variability. In other words, it is a means to achieve happy and healthy manufacturing.