Currently, KENKI DRYER of a high moisture contents sludge dryer is operated by PLC which is a device controlling machine movements with programs though, it is not completely automatic and rotation speed of the dryer’s main shaft need to be adjusted by manual according moisture contents after the drying.
To keep the moisture contents of the dried material constant as much as possible after the drying, rotation speed of main shaft in dryer’s main body need to be finely tuned by skilled workers according to moisture contents of the materials and change of temperature inside the dryer.
The skilled worker’s experience and intuition are introduced into the AI program which is based on changes of temperature at 4 measuring point in the dryer and moisture contents before and after the drying, and this makes the rotation speed of the main shaft to be adjustable and changeable automatically and accurately without any manual operation.
The AI program builds models based on estimated data to run its automatic control, and accuracy of the control is improved by the estimated data which keep changing by accumulating and utilizing more data.
Therefore, alignment with IoT remote access is very important in the AI automatic control and more accurate control becomes possible after the system start running by using data accumulated by IoT.
There will be no more labor cost for hiring skilled workers because of the automated operations and also COVID-19 can be prevented by avoiding direct contact with other workers at the local site.
AI model is built with 4 steps.
1.Development of estimated data infrastructure
2.Proof and development of AI model
3.Implementation of AI into PLC
4.Integration verification with IoT system
Development of data infrastructure is the first step in AI model development. The most important thing in this process is estimation of specific data.
Based on past accumulated experimental data, factors which have large effect is inferred and infrastructure of collecting log data about the factors is developed.
Since accuracy cannot be obtained with laboratory data alone, data from linkage with IoT is used to build data infrastructure which will make progress reliably.
Several patterns of AI model are built and proofed. SARIMA which is a multivariate time series model and Transformer which is a time series model are built.
Log data of built models are collected and verified, then various realization methods are explored and developed.
This process of AI model building is the same as ones that on-site skilled workers learn from their experience and intuition, and dryers can automatically run without the skilled workers.
The developed AI model is implemented into PLC.
There are 2 ways in the self-driving with AI, which are controlling on-site by setting a device on control board, or control via internet. In both ways, the developed AI model has flexibility based on its load calculation.
After AI model is implemented into PLC, the AI model is checked and verified in stand-alone.
After the stand-alone test, system integration with IoT is proofed and verified. This system integration is very important process which affects to reliability of the AI model in future.
|AI based on data accumulated every seconds of temperature at 4 measuring points.|
|AI automatically controls rotation speed of main shafts per unit time.|
AI (Artificial Intelligence) a term for software or system which is simulated human knowledge with computers.
IoT stand for Internet of Things and is a system to getting data from things and control the things remotely by incorporating information into things and connecting them via internet.
PLC is a device which automatically control a machine.
You must be logged in to post a comment.