Multisensory data fusion technique and its application to welding process monitoring

Abstract Ultrasonic metal welding is one of the key technologies in manufacturing lithium batteries, and the welding quality directly determines the battery performance. Therefore, an online welding process monitoring system is critical in identifying abnormal welding processes, detecting defects, and improving battery quality. Traditionally, the peak welding power is used to indicate abnormal process signals in welding process monitoring systems. However, since various factors have complex impacts on the electric power signals of ultrasonic welding processes, the peak power is inadequate to detect different types of welding defects. Therefore, a signal pattern matching method is proposed in this study, which is based on the electric power signal during the entire welding process and thus is capable of identifying abnormal welding processes in various conditions. The proposed method adopts isometric transformation and homogenization as signal pretreatment methods, and Euclidean distance is used to calculate the similarity metric for signal matching. The effectiveness and robustness of the proposed method are experimentally validated under different abnormal welding conditions.

Download Full-text

Towards the development of a smart fused filament fabrication system using multi-sensor data fusion for in-process monitoring

Rapid Prototyping Journal ◽

10.1108/rpj-06-2019-0167 ◽

2020 ◽

Vol 26 (7) ◽

pp. 1249-1261 ◽

Cited By ~ 3

Author(s):

Michele Moretti ◽

Federico Bianchi ◽

Nicola Senin

Keyword(s):

Data Fusion ◽

Process Monitoring ◽

Sensor Data ◽

Fused Filament Fabrication ◽

Content Type ◽

Heterogeneous Sensors ◽

Sensor Data Fusion ◽

Part Quality ◽

Multi Sensor Data Fusion ◽

Smart Machine

Purpose This paper aims to illustrate the integration of multiple heterogeneous sensors into a fused filament fabrication (FFF) system and the implementation of multi-sensor data fusion technologies to support the development of a “smart” machine capable of monitoring the manufacturing process and part quality as it is being built. Design/methodology/approach Starting from off-the-shelf FFF components, the paper discusses the issues related to how the machine architecture and the FFF process itself must be redesigned to accommodate heterogeneous sensors and how data from such sensors can be integrated. The usefulness of the approach is discussed through illustration of detectable, example defects. Findings Through aggregation of heterogeneous in-process data, a smart FFF system developed upon the architectural choices discussed in this work has the potential to recognise a number of process-related issues leading to defective parts. Research limitations/implications Although the implementation is specific to a type of FFF hardware and type of processed material, the conclusions are of general validity for material extrusion processes of polymers. Practical implications Effective in-process sensing enables timely detection of process or part quality issues, thus allowing for early process termination or application of corrective actions, leading to significant savings for high value-added parts. Originality/value While most current literature on FFF process monitoring has focused on monitoring selected process variables, in this work a wider perspective is gained by aggregation of heterogeneous sensors, with particular focus on achieving co-localisation in space and time of the sensor data acquired within the same fabrication process. This allows for the detection of issues that no sensor alone could reliably detect.

Download Full-text