Container Inspection Control Using Minor Defect Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing container treatment systems fail to consider valuable information from containers with minor defects, leading to inadequate treatment and potential system inefficiencies.
Innovation Solution
A method involving a transport device for container parts, sensor devices for spatially resolved data acquisition, and a real-time evaluation using a machine learning container inspection model to assess defects and adjust treatment processes, including a plant inspection task to detect subtle trends and system deviations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If containers with minor defects are excluded from the container stream, then manufacturing precision is improved, but loss of information occurs as valuable data from these containers remains unconsidered
Solution Approach 1:
The system introduces feedback by feeding back the images and defect information of containers with minor defects into the training dataset for the machine learning model. This allows the model to learn from these cases and improve its inspection accuracy, transforming previously discarded information into valuable training data that enhances future defect detection capabilities.
Solution Approach 2:
The system changes the parameter of defect severity classification by introducing a gradient approach where containers are not simply classified as defective or non-defective, but rather their defect severity is evaluated on a spectrum. This allows containers with minor defects to be retained and their images utilized for training, rather than being uniformly excluded.
2Ease of operation
If a predefined threshold value is used to exclude containers, then ease of operation is improved, but measurement precision deteriorates as subtle trends and system deviations are not detected
Solution Approach 1:
The system transitions from a static threshold-based inspection approach to a dynamic machine learning-based approach. The ML model continuously learns from new data and adapts its detection criteria, enabling it to identify subtle trends and system deviations that fixed thresholds would miss, while maintaining operational simplicity through automated decision-making.
Solution Approach 2:
The machine learning model performs self-improvement by automatically learning from the feedback provided by containers with minor defects. The system refines its own inspection criteria and threshold values through continuous training, reducing the need for manual threshold adjustment while improving detection precision for subtle defects.
Data Source
Figure 1
Figure 2
AI summary
Method for operating a container treatment plant for treating a plurality of container parts for containers, plastic containers and/or bottles, wherein a transport device transports the container parts as a container part stream along a transport path from one treatment unit of the container treatment plant to another treatment unit of the container treatment plant, wherein, for carrying out a container inspection task, a sensor device, in particular spatially resolved sensor data and camera images relating to the container parts, optically acquires and a real-time evaluation device evaluates the spatially resolved sensor data in real time by means of a container inspection model of machine learning, which comprises a set of parameters which are set to values that were learned as a result of a machine learning procedure, characterized in thatthat a set of container part characteristics based on a machine learning method is specified, and the acquired, spatially resolved sensor data are evaluated with respect to a plant inspection task different from the container inspection task, based on the specified set of container part characteristics, whereby, depending on the inspection result of the plant inspection task performed, a plant inspection parameter is determined, or a similarity parameter is determined which is characteristic of the similarity of the spatially resolved sensor data to reference data, and wherein these parameters are provided for the control and/or regulation of the container handling plant.