AI Development Toolkit for Inspections on the Edge
SPECTRO CORE is an inspection AI SDK, designed for embedding machine vision and time series data analysis applications directly into manufacturing machines. It improves production yields, enables higher processing accuracy, and reduces defects rates by detecting errors and defects for a wide range of materials.
Thanks to the lightweight Sparse Modeling based technology powering SPECTRO, it is ready for integration directly with manufacturing machines and production equipment, without the need for additional hardware, and is compatible with machinery such as laser cutters, drilling machines, welding machines and industrial / assembly line robots.
HACARUS was awarded the “inVISION TOP INNOVATION 2020” prize for its work to bring smarter AI to the computer vision field.
SPECTRO CORE Benefits
SPECTRO CORE Functions
Detects defects in various materials such as wood, metal, and fiber. AI models can be created from only good samples, in case there is no defective product data.
Example of use: Detection of scratches, dents, defects, and damage.
Missing parts & Misalignment Detection
Detects defects such as missing parts and misaligned parts, identifying the parts that should not be present on the inspection target.
Example of use: Detection of missing parts, misalignment of seal position, and potential defects
Recognizes common patterns found on inspection target and detects areas that differ from other patterns. Identifies areas that should be uniform and defects non-uniform area.
Example of use: Soldering inspection, identification of line breaks on circuit boards.
Predictive Maintenance & Component Health
Capable of analyzing time-series data, such as sound and vibrations produced by machinery to provide insights about component health, enabling preemptive maintenance.
Example of use: Ensure continuous operation, predict component failure.
Proven Superior Performance
Time Series Analysis of Vibration Data
Utilizing SENSPIDER, MACNICA’s AI-ready IoT sensor hub combined with SPECTRO CORE’s versatile algorithms, the joint solution can detect anomalies in vibration data. In the case study highlighted on the right, using data from industrial fans, the solution can detect anomalies over 200 times faster compared with a common k-Nearest Neighbors (KNN) approach, with 100% accuracy.
Solar Cell Visual Inspection
When comparing SPECTRO’s performance with Classifier (SVM) and Deep Learning (CNN) techniques for detection of defects on Solar Cells, SPECTRO far outperforms the competition. Not only is accuracy higher, but SPECTRO also creates AI models faster – even when using a far smaller dataset.