Particle Sensor Arrangement Using ML for Real-Time Shape Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing sensor arrangements require significant effort to determine the shape of particles, and existing methods are inefficient for rapid determination of particle features.
Innovation Solution
A sensor arrangement that includes an emitter, detector, measurement volume, digitizing unit, and evaluation unit with a trained machine learning algorithm to determine particle features, particularly shape, without the need for cameras and image recognition software, enabling quasi-real-time detection of particle features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional sensor arrangements are used to determine particle shape, then measurement capability is provided, but the effort and time required are significantly high
Solution Approach 1:
The patent replaces complex mechanical/image processing systems with a trained machine learning algorithm that processes detector signals directly. The algorithm has been trained on particle images and can determine shape parameters from scattering patterns without requiring actual image capture and analysis, thus substituting a time-consuming mechanical/optical process with a faster computational model.
Solution Approach 2:
The machine learning algorithm is trained in advance on a large dataset of particle images and their corresponding shape parameters. This preliminary training phase allows the algorithm to learn the relationship between scattering patterns and particle shapes, so that during actual measurement, the algorithm can quickly determine particle features without performing complex real-time image analysis.
2Measurement precision
If conventional methods are used for particle feature detection, then accurate measurements can be obtained, but the processing speed is insufficient for rapid determination
Solution Approach 1:
The system replaces slow conventional image recognition software with a pre-trained machine learning algorithm that processes detector signals directly. This substitution maintains measurement accuracy by using the same underlying physics of light scattering while achieving processing speeds of 10^6 particles per second through optimized computational algorithms.
Solution Approach 2:
The patent changes the processing approach from analyzing complete images to analyzing extracted scattering pattern parameters. By transforming the problem from image recognition to parameter-based analysis using a trained algorithm, the system achieves both high accuracy and high throughput for particle feature determination.
3Shape
If image recognition software is used for particle shape analysis, then detailed shape information can be obtained, but the system complexity and processing time increase
Solution Approach 1:
The patent extracts only the essential scattering pattern parameters needed for shape determination, eliminating the need for complete image capture and processing. The trained machine learning algorithm works with these extracted parameters to determine particle shape, thereby reducing system complexity while maintaining the ability to obtain detailed shape information.
Solution Approach 2:
The system replaces complex image recognition software with a streamlined machine learning algorithm that processes simplified detector signals. This substitution reduces computational complexity and processing time while still providing accurate particle shape information through the algorithm's training on particle images.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enables rapid determination of particle features, such as shape and size, at a rate of 106 particles per second, facilitating in-line control of chemical processes and detection of agglomerates, shape deviations, contamination, and foreign particles.
Implementation Method 1
an emitter (14) for emitting electromagnetic radiation (16)
Implementation Method 2
a detector (18a, 18b) for receiving radiation emitted from the emitter and for providing detector signals from the received radiation
Implementation Method 3
The particles scatter, reflect and/or absorb the radiation. The transmitted, scattered and/or reflected radiation is measured by the detectors
Implementation Method 4
The particles scatter, reflect and/or absorb the radiation
Data Source
AI summary
A sensor arrangement for detecting features of particles includes an emitter for emitting electromagnetic radiation, a detector for receiving the electromagnetic radiation emitted from the emitter and for providing detector signals based on the received electromagnetic radiation, and a measurement volume irradiable by the electromagnetic radiation emitted by the emitter. The measurement volume is configured for receiving particles flowing therethrough. The sensor arrangement further includes a digitizing unit for digitizing the detector signals, and an evaluation unit for evaluating the detector signals. The evaluation unit stores a trained algorithm for machine learning. The algorithm is configured for determining at least one feature of the particles based on the detector signals.