DSEY : Data Science & Engineering for Yield

DSEY provides the Industry 4.0 smart factory solution, so customers can focus on doing what they do best! Innovation of manufacturing to meet ever changing business environment or improving equipment efficiency, process quality, and productivity.

Based on the experiences of optimization from the early days of semiconductor/display production automation to the present day, the company provides Industry 4.0 smart factory total solutions (Signe-S Series) of the fourth industrial revolution.



DSEY aims to maximise business competitiveness of the customer around the world.

- Next-generation MES & ECM for bigdata processing
- Machine learning framework for accelerating intelligent factory
- Intelligent VM, APC, FDC, and SPC for process quality innovation
- Big messaging infra for high performance transporting equipment data
- Mesh network for stable wireless communication of all IoTs

Signe-S Philosophy


For intelligent factories, manufacturing execution system must ensure high performance, high reliability and equipment data must be 100 percent secured in order to build AI model. DSEY generates 100% measurement data through artificial intelligence and provides high-precision process quality control.

02. Machine Learning

Due to lack of data science capability at manufacturing sites, it is difficult to select statistical techniques for data analysis, difficulty in using algorithms for feature selection, difficulty in using algorithms for machine learning, and time spent due to trial and error. That's why we need the one-touch machine learning product with built-in data science capabilities specializing in manufacturing. This will increase the accuracy and efficiency of data science at the manufacturing site.

03.Intelligent Factory

Metrology data is a very important factor for improving the quality of products. However, in order to maximize productivity, some measurements may not be made or some may be made. There is a solution that enables complete quality control even without metrology data

Our Products

Signe-S MES & Big Messaging

To build an intelligent factory, MES must be an architecture that can collect, deliver, store, and process high-volume equipment data with high performance and reliability. DSEY revolutionized big data processing in production control systems.

MES's big messaging foundation based on Solace is the only advanced event broker technology that supports pub/sub, queuing, request/reply, streaming and replay, and that’s available as run-anywhere software, purpose-built hardware and a managed service.

You can control it all through a single panel of glass and deploy it anywhere—public cloud, private cloud, no cloud. 65 billion messages a day. That’s 750,000 messages a second, 24 hours a day. That’s huge volume, and we do all that completely seamlessly without any data loss.”
DSEY is also a professional partner of Solace.

Signe-S ECM

DSEY revolutionized big data processing in equipment control systems.

As long as the facility is operating normally, all data must be collected from the equipment without loss. You must be able to collect data for less than 1 second in real time. SECS-II messages must also be able to be delivered to other applications as a hub than ECM.

DSEY also provides SECS-II driver, HSMS driver, Logger and APIs.
- SML/MDF Parsing
- SML Factory
- No SML/MDF Definition Required
- Binary format SECSII <->Object based SECSII <-> Formatted SML, XML
- Easy Application Programming Interfaces,
- SECSII Encode/Decode by Item Path
- Open, Close, Send, Stop & Events.

Signe-S ML Framework

ML Framework is DSEY's proprietary machine learning framework for implementing intelligent factories, enabling AI models to be created more accurately and fast. MLF includes manufacturing domain consulting and so It provides fast and accurate manufacturing-specific data science methods that do not require complex statistics, algorithm selection, and coding.

Signe-S VM & FDC

Signe-S VM monitors process, predicts the results, and identifies key variables which contribute mostly to the results.

Traditionally FDC is implemented in the unsupervised way which is using only input data without the process results. The specifications of input data can not be specified because input data is related interactively. Because the unsupervised FDC tells us that it is not actually a failure, the engineers who have to check it out have a lot of work, and eventually the equipment operation rate is reduced.

User can implement own supervised based FDC using Signes-S MLF. Signes-S VM generates faults alarm using Hotelling’s T2 and SPE against inputs if the predicted output is abnormal (out of specifications or out of control limits). Signes-S VM also generates contributed input variables to the faults. Signes-S SPC generates actual alarms for the faults based on user SPC rules.

Signe-S APC

As the manufacturing process becomes more complex and the acceptable range of critical dimension becomes narrower, a small change in a recipe setting during the process may cause a significant loss to the final yield. Traditional APC does adaptive control rather than re-estimating control parameters. This make big fluctuations in changing setpoints.

APC provides information about a current run (a wafer, a glass, a lot, or a batch) to the next process step or the previous process step based on the metrology result of the next step for adjusting recipe settings. A proper APC leads to an increase the final yield of a wafer as well as a high quality of product. Signe-S APC is a new way of APC which minimizes input changes and also re-estimates control parameters using RL (Reinforcement Learning).

✓ Model (SISO, MISO, and MIMO)
✓ Dynamic system identification
✓ Dynamic parameter estimation
✓ Only model structure is required
✓ Self trainning through RL
✓ Users can apply their own RLs
✓ Offline model optimization
✓ Analyzing random/systematic errors
✓ No code and No maintenance

Signe-S SPC

- Detects process changes using Hotelling’s T squared and SPE

- Detects how close virtually measured data and actual measured data using Chi-square distribution and normal distribution.

- Monitors product quality using the desired Cpk and the target
. 𝜎 = (𝑈𝑆𝐿 − 𝐿𝑆𝐿)/6𝐶𝑝𝑘

- Monitors process capability using the sigma and the mean.
. 𝐶𝑝𝑘 = (𝑈𝑆𝐿 − 𝐿𝑆𝐿)/6𝑠

- Monitors western electric rule violations using SPC charts:
. X bar control chart
. Range “R” control chart
. Standard Deviation “S” control chart
. “u” and “c” control charts
. “p” and “np” control charts
. Pre-control Charts

- Monitors trend of time series data using EWMA chart
. 𝑋𝑛 = 1 − 𝜆 𝑋𝑛−1 + 𝜆𝑋𝑛

- Traces random and systematic errors(𝜀) in measured data
If 𝜀 > k 𝜎, adjust the error and weighting factor




B-dong, 309-ho, Hill state Eco songpa, Jeongui-ro 7-gil 13, Songpa-gu, Seoul, Korea (Zip05835)



Email Address