Blog | cnix

Anomaly Detection with Local LLM

In manufacturing, equipment anomalies or operational status faults are generally identified through signal alarms. Aging equipment, in particular, exhibits early warning signs before a breakdown occurs. However, while early warning signs that occur near the alarm threshold can be detected in advance, those that appear as irregular pattern disruptions within the normal range are difficult to spot. Consequently, these undetected issues can lead to equipment failure and ultimate

Changdeuk Kang

7 days ago1 min read

Use-Case for AI Agent of Data Service Platform…

The Agent for the Data Service Platform’s local LLM bridges the probabilistic outputs of artificial intelligence with the deterministic systems of the manufacturing floor. By delivering ultra-fast response times of under 1 second for users’ basic queries, it is well-suited for use in manufacturing environments. Cnix’s AI Agent is directly connected to the Data Service Platform, enabling it to monitor and control all data, retrieve and analyze stored data, and deliver the resu

Changdeuk Kang

Jul 251 min read

Agent for Data Service Plaform

A manufacturing site, composed of numerous processes and equipment, has just as many proprietary tools provided by equipment suppliers. To manage the shop floor, one must be proficient in all of these tools—which, in reality, is practically impossible. To solve this challenge, Cnix provides an integrated engineering environment as a Local LLM agent. Regardless of the equipment type, this agent enables users to perform supervising functions—such as reading data from devices co

Changdeuk Kang

Jul 71 min read

Local LLM

When a problem occurs with the production process or equipment on your shop floor, what is the very first thing you do to find the root cause? The fastest ways are to either ask a senior engineer or check the logic and status through the PLC program. Most manufacturing sites are automated using PLCs, and the know-how of senior engineers is embedded right into these PLC programs. In other words, it is no exaggeration to say that the manufacturing floor is driven by PLCs. Howev

Changdeuk Kang

Jun 301 min read

Pascal2 - Digital Twin

Pascal2 consists of Vision and Digital Twin modules. While Pascal2-Vision is an inference platform that enables CPU-based inference using vision AI models, the Digital Twin is a proprietary AI model. It is a system capable of simulating real-world environments with high fidelity by executing the entire process from training to inference. Furthermore, when the number of input features is optimal, it can perform high-speed inference efficient enough to apply the inference outpu

Changdeuk Kang

Jun 91 min read

Boosted Pascal 2

PASCAL2 is a container that enables trained AI models to run directly on CPUs without requiring any additional GPU cards. On modern CPUs equipped with NPU or integrated GPU cores, PASCAL2 can boost performance to a level that rivals dedicated GPU cards. In industrial settings, the vast majority of AI applications in manufacturing involve inference using pre-trained models. Through this inference, smart manufacturing environments are realized — including digital twins, image a

Changdeuk Kang

May 71 min read

Safety in PASCAL2

PASCAL2 now features a high-performance engine capable of running advanced object detection AI models directly on CPUs, eliminating the need for costly, high-end GPUs. This breakthrough allows for high-efficiency AI deployment without the heavy hardware investment. Built for the demands of modern industrial sites, PASCAL2 provides rigorous monitoring of hazardous zones and restricted areas. The system is engineered to automatically detect unauthorized intruders and fallen wor

Changdeuk Kang

Mar 181 min read

New Digital Twin in PASCAL2

PASCAL2 introduces a newly upgraded Digital Twin capability. Like PASCAL2 itself, the integrated Digital Twin operates without a database or GPU, minimizing implementation costs. By training and saving field data in PASCAL2 just once, the same model can be used simultaneously for both Anomaly Detection and Digital Twin — eliminating the need for separate training dedicated to the Digital Twin. Users can build a complete Digital Twin system simply by copying the saved training

Changdeuk Kang

Feb 241 min read

PASCAL 2...

PASCAL2, an artificial intelligence model developed exclusively for the Data Service Platform, has emerged as a next-generation industrial monitoring solution capable of performing real-time machine learning and anomaly detection. Powered by its built-in SGDEngine, the system supports continuous online learning and instant anomaly identification. It is demonstrating outstanding performance across industrial control, data analytics, and anomaly detection applications. Unlike c

Changdeuk Kang

Feb 122 min read

Unlimited Connections.

One of the key aspects of a smart factory is system integration. System integration allows the IT layer to view the hardware layer as a single virtual machine, creating an environment that makes it easier to build smart factory services. The Cnix Data Service Platform proposes a simpler way to connect data between production facilities — allowing you to integrate your equipment with just a few clicks. It supports configuration changes for data sharing between production facil

Changdeuk Kang

Oct 31, 20251 min read