Sensors and Data in Laser Welding: What Gets Measured and How
Laser welding is monitored through cameras, thermal sensors, acoustic and optical emissions, and real-time equipment signals. Some quantities cannot be measured directly and require soft sensors estimated from other signals.
A laser weld happens in milliseconds to minutes. The sensor systems that monitor it must capture physical phenomena at high speed and across multiple modalities. The data they produce is the input to every ML model applied in the process.
Physical Sensors
High-speed camera systems monitor the weld pool geometry in real time. Pool shape, width, and spatter events are all visible in the image stream. CNN-based models are applied directly to this data.
Thermal sensors measure temperature dynamics across the weld zone. Temperature distribution determines the heat-affected zone (HAZ) and drives capacity fade in the material around the weld.
Acoustic sensors (Piezoelectric Transducers) listen for vibrations and stress waves beyond the range of human hearing. These signals indicate metal cracking, distortion, or other structural changes that precede weld failure.
Optical / spectroscopic sensors capture emission spectra from the plasma and weld pool. The spectroscopic signal carries information about the chemical composition of the melt and the energy state of the process, both of which are related to weld quality.
Laser line scanners measure the X, Y, Z coordinates of the joint, which is the gap between the two metal pieces that must be welded. A joint that is not properly tracked will not be welded correctly.
Hall effect sensors measure the magnetic field around the conductor carrying current to the laser. This gives an indirect measure of the electrical state of the welding system.
PLCs (Programmable Logic Controllers) record arc voltage, current, and welding speed directly from the equipment.
Soft Sensors
Some physically meaningful quantities cannot be measured directly without destructive inspection. You would have to cut the metal open to measure penetration depth. Soft sensors estimate these from the available physical sensor signals:
- Penetration depth estimator: inferred from process signals
- Weld pool width estimator: inferred from camera or thermal data
- HAZ monitor: heat-affected zone extent, critical for materials where a large HAZ causes structural problems (aerospace applications require tight HAZ control)
Multi-Sensor Fusion
No single sensor captures the full state of the weld. Multi-sensor fusion combines signals from cameras, thermal sensors, acoustic sensors, and equipment signals into a unified representation. The challenge is synchronising signals from sensors with different sampling rates and aligning them to the same physical event in the weld.