Bobbin Mounting Detection Using Machine Learning Imaging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting the mounting state of bobbins on bobbin holders, such as those using photoelectric sensors, are prone to errors due to unevenness, gaps, and foreign matter, leading to inaccurate determination of bobbin orientation and spacing.
Innovation Solution
A mounting state detection device employing machine learning to create an estimation model from training data, which images the bobbin and determines its mounting state by analyzing the positional relationship between the bobbin and slit ranges, enabling accurate assessment of bobbin orientation and spacing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a photoelectric sensor is used to detect the slit position and determine bobbin mounting state, then the detection process is simple and fast, but the detection accuracy deteriorates due to erroneous detection from unevenness, gaps, and foreign matter
Solution Approach 1:
The patent replaces the photoelectric sensor-based detection system with an imaging device (camera) that captures images of the bobbin and slit. This substitution allows for more accurate detection by visually capturing the actual positions and states of components, eliminating erroneous detections caused by unevenness, gaps, and foreign matter that affect photoelectric sensors.
Solution Approach 2:
The imaging device creates a visual copy (image) of the bobbin and slit arrangement. By capturing and analyzing this visual representation, the system can accurately determine mounting states without direct physical contact or reliance on optical sensors that are susceptible to environmental interference.
2Device complexity
If traditional detection methods are used, then the device structure remains simple, but the reliability of bobbin mounting state determination deteriorates
Solution Approach 1:
The patent replaces simple photoelectric sensor detection with an imaging-based detection system that provides more reliable determination of mounting states. The imaging device captures visual evidence that can be analyzed to confirm proper bobbin installation, significantly reducing erroneous detections and improving overall system reliability.
3Ease of operation
If photoelectric sensors are used for slit detection, then the detection method is straightforward, but the ability to accurately determine bobbin orientation and spacing deteriorates
Solution Approach 1:
The patent transitions from one-dimensional photoelectric sensor detection to two-dimensional imaging detection. The imaging device captures spatial relationships in multiple dimensions, enabling accurate determination of both bobbin orientation (angular position) and spacing (distance between bobbins) simultaneously, providing comprehensive mounting state verification.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Provided is a mounting state detection device (10) capable of accurately determining whether or not a mounting state of a bobbin is appropriate. The mounting state detection device (10) includes a control device (50) and an imaging device (61) configured to be able to image a bobbin mounted on a bobbin holder of a thread winding machine. The control device (50) executes processing for acquiring an estimation model (124) generated by performing machine learning of a plurality of pieces of training data (123). Each of the plurality of pieces of training data (123) associates a training bobbin image showing the bobbin with mounting state information representing a mounting state of a bobbin, as a label. The control device (50) further executes processing for acquiring a determination bobbin image from the imaging device (61) by imaging the bobbin mounted on the bobbin holder, and processing for determining whether or not the bobbin is mounted normally on the bobbin holder based on the mounting state information output from the estimation model (124) by inputting the determination bobbin image to the estimation model (124).