Prototype testing & end-of-line inspection

Automated Error Detection through Good/Bad Analysis with Artificial Intelligence

In the following project, our aim is to classify components using a good/bad analysis.

For this purpose, the AI model is trained with a learning set of good and bad parts of the specific test sample. The significant features for the separation between good and reject are determined by the model. No expert knowledge or manual adjustments are required. The resulting model is validated by the internal division of the learning set. The model can be extended at any time through relearning. It therefore offers flexible learning of different test components and extension of existing models.

With the automated defect detection system QAIros, it should be possible to optimise components as early as the development stage using AI-supported classification. The results should then be transferred to the production process in order to carry out an automated 100% inspection.

quality-testsystem-qairos-and-impulse-hammer

Development of a Portable Test System

In co-operation with the company Miba Sinter Austria GmbH, we developed a portable test system for 100% crack testing of sintered parts. These parts are manufactured with high precision and in large quantities and are used in the automotive sector, e.g. in gearbox construction. Sintering is a complex process that is influenced by numerous factors both at material (e.g. grain size, density) and at process level (e.g. pressure, temperature).

The Project

Our portable system makes it possible to test prototypes under near-series conditions at the development phase. This results in process and parameter optimisation, which ensures that the reject rate in subsequent series production remains minimal. This project represents a significant advance in quality assurance and increased efficiency in the area of sintered part production.

System Structure and Components

Hardware

The main hardware components include:

Rotatable unit: This component is used to optimise the positioning of the test parts. It is able to rotate test parts by 120°, which enables three different test positions.
Component-specific bearing: This is attached to the rotatable unit. It is customised using 3D printing technology to ensure an optimal fit and stability.
Automatic impulse hammer WaveHit^MAX, the orientation of which can be adjusted in the X and Y directions with a position indicator to ensure precise application of the test impacts.
Microphone with variable orientation so that it can be optimally positioned to capture the sound response.

This hardware setup allows reproducible positioning and testing of test parts with diameters of up to 200 mm and a weight of up to 1 kg.

Software

Training

The AI models are trained in the training module of the software. It requires a data set consisting of good and reject parts of the specific test object which are provided by the customer. The test parts are excited at the appropriate points with the automatic impulse hammer WaveHit^MAX while the microphone records the sound response. Using the data set the software trains the AI model automatically:

Model training: The AI model analyses the recorded sound responses, identifies significant features to separate pass and fail parts and creates the model without manual intervention.
Automatic validation: The model is validated by an internal division of the learning set, which increases accuracy and reliability.
Flexible relearning: The model can be extended at any time by adding new data to integrate additional test parts and variations.

This approach does not require expert knowledge, as no manual adjustments such as threshold determinations or peak definitions are necessary.

Classifiation

The software's classification module enables new test parts to be tested using the trained model:

Project file: A specific project file is created for each model, containing all relevant test settings (e.g. hammer force, position) to ensure consistent and reproducible measurements.
Result documentation: The classification results are documented and can be seamlessly integrated into a database or exported in formats such as Excel and PowerPoint to facilitate further processing and analysis.

Summary and Advantages

The QAIros testing system is an innovative solution for process-integrated and automated quality assurance (QA) that offers maximum reliability. In addition to the option of integrating the system directly into the production line, it can also be used as a portable system, which is particularly advantageous in prototype development. This enables process and parameter optimisation as early as the development phase.

The advantages of the system include:

Fast verification of the suitability of the test parts for the acoustic resonance method (ART), which enables rapid quality assessment.
Reliable and fast teach-in and classification of test parts, facilitating the transition from development to series production.
Reproducible setup, which ensures that test results are consistent and comparable.
Quick adaptation for new test parts, allowing flexible adjustment to changing production requirements.
Optimisation of the reject rate already during the development of the prototype, which leads to a significant reduction in production waste.
Seamless connection of test results to existing database structures, for example via CSV formats, to ensure efficient data management.

This comprehensive system offers an innovative solution for quality assurance in sintered part production and ensures a significant reduction in the reject rate thanks to the combination of advanced hardware and intelligent software.