Pseudo-Image Sensor State Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for detecting facility states using image recognition AI are limited by image resolution and require manual algorithm creation for sensor signals, making them costly and time-consuming, especially when dealing with sensors like vibration sensors with varying phases.
Innovation Solution
A state detection apparatus that transforms digitized time-series sensor signals into spectral intensity data, generates pseudo images, and uses AI for classification, allowing for edge device implementation without being affected by sensor signal phase variations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If time-series signal analysis is performed with dedicated algorithms manually created for each sensor type, then detection accuracy is improved, but system development time and cost increase significantly
Solution Approach 1:
The patent transforms time-series sensor signals into pseudo-RGB images by changing the parameter representation from temporal domain to spatial domain. This allows the use of universal image recognition algorithms instead of sensor-specific algorithms, dramatically reducing development time while maintaining detection accuracy.
Solution Approach 2:
The patent creates a copy of the time-series signal data in the form of pseudo-RGB images, which can then be processed using existing image recognition frameworks. This copying approach enables reuse of mature image processing algorithms across different sensor types without requiring custom algorithm development for each sensor.
2Measurement precision
If deep learning is used to process time-series sensor signals, then detection capability is improved, but training complexity and time requirements increase significantly
Solution Approach 1:
The patent substitutes the complex time-series signal processing mechanism with an image processing mechanism. By converting sensor signals to pseudo-RGB images, the system can leverage well-established image recognition deep learning models that are easier to train and deploy, reducing overall system complexity.
Solution Approach 2:
The patent introduces pseudo-RGB images as an intermediary representation between raw sensor signals and final detection results. This intermediary form allows the use of image recognition algorithms that are simpler and more efficient to train compared to direct time-series signal processing, thereby reducing training complexity.
3Adaptability or versatility
If vibration sensor signals with varying phases are transformed into images, then phase independence is achieved, but image variation complexity increases
Solution Approach 1:
The patent applies parameter changes by transforming the phase-sensitive time-domain signal into the frequency domain through Fourier transform, then converts to spatial domain through image generation. This multi-parameter transformation eliminates phase sensitivity while creating a standardized image representation that reduces variation complexity.
Solution Approach 2:
The patent moves the data from one dimension (time-series signal with phase variations) to another dimension (spatial image). By representing the signal in the frequency domain and then as an image, the system achieves phase invariance while the image format provides a structured representation that manages complexity.
4Reliability
If dedicated analysis algorithms are created for each sensor type, then detection reliability is improved, but system cost and deployment difficulty increase
Solution Approach 1:
The patent creates a universal system that can process multiple types of sensor data (vibration, temperature, pressure, etc.) through a single common pipeline: transforming signals to pseudo-RGB images and using a unified image recognition model. This multi-functional approach maintains detection reliability while dramatically easing deployment compared to sensor-specific systems.
Solution Approach 2:
By copying the essential characteristics of different sensor signals into the pseudo-RGB image format, the system creates a universal representation that can be processed by a single detection model. This copying approach maintains the reliability needed for each sensor type while enabling easy system-wide deployment.
Data Source
AI summary
The time-series signal of the sensor is transformed to the spectral intensity by fast Fourier transform (FFT) or the like, and the one-dimensional data of the spectral intensity is generated. A pseudo image is generated, for example, by repeatedly arranging the one-dimensional data in the vertical direction, or by arranging the one-dimensional data for a plurality of sensors in the vertical direction. The state of the facility is identified by analyzing the pseudo image with an image analysis unit such as a convolutional neural network.