Acoustic quality assurance in machining production

Initial situation and idea

Incorrect machine settings on machining systems result in poor product quality, in an increase in recourse claims, as well as frequent and unexpected downtime due to machine breakdowns. These issues are exacerbated by low automation rates and a shortage of skilled personnel, resulting in rising production costs. SMEs often struggle to cover these costs. For this reason, both users and machine and control system manufacturers have expressed the need for reliable, automated methods of monitoring cutting machines. Currently, the analysis of faulty machine settings is carried out on an individual basis, with specialized staff using the occurring process or system sounds to draw conclusions about potential causes of malfunction. Therefore, the assessment of machine condition and process stability in machining production is partly based on acoustic emissions. Analyzing this airborne noise in combination with machine learning methods should enable the automated acoustic monitoring of cutting machines. 

Brief explanation of the project

The amoZFerg project aims to reduce machine downtime due to incorrect parameter settings by up to 50% using the new acoustic inspection method. This will lead to significant cost savings given the high cost of machines and current downtime of up to 10%. Additionally, the project aims to achieve 95% detection accuracy for various error classes on machines, reducing the reject rate to 3% for very small production batches. The results of the analyses will be output via the machine control system, which will significantly reduce the need for individual technicians to be assigned to the machines and tools by their manufacturers.

The innovation in acoustic monitoring

Conventional monitoring systems, which are mostly optical, are often expensive and require special expertise. They can also be a source of errors and are not universally applicable. Optical sensors are limited in their ability to monitor complex components and applications involving coolants or lubricants. Acoustic monitoring using microphones has already shown promising results in welding (see the “AKoS” project). This method is now also being researched for use in machining, particularly for monitoring milling and turning machines. Microphones record sound emissions that are later classified according to various predefined machine error patterns. A challenge in acoustic analysis is the constant background noise present in factory halls.

• Selection of the acoustic measurement setup based on room and process acoustics (microphone concept).
• Development of machine learning methods for detecting and classifying defined error patterns on cutting machines based on acoustic signals (robust against disturbing noise).