Verification of assembly processes is common practice across manufacturing to help guarantee quality. The current methodology is to inspect after something is applied or assembled. This practice works well for many inspection applications, for example, if the process is automated with motion control and applicators. In an automated solution there are various reasons the system may not execute a task properly. For example, in a pick and place application, a gripper can drop an object due to air leaks. Or an applicator applying sealant could have dry sealant in the tooling tip, causing improper application of the sealant. In these scenarios, and multiple others, a post process inspection system can catch these defects and alert the right team members. Not all factories, processes, or product assemblies are fully automated, and additionally, in some cases automation technologies are not cost effective. This results in manual assembly operations.
Quality & Manual Operations
Manual operations have benefits, which is why many factories are not highly automated. One reason is manual operations are less costly due to the flexibility in dynamic environments, especially compared to adding automation. Even though manual operations tend to be more dynamic and less expensive, they have their own breadth of challenges. If an assembly operation is not fully automated there are some challenges from a quality perspective to overcome. If something is missing from an assembly it is a bit easier to point to the fact an operator did not apply a component. What if the assembly operation requires specific steps or a sequence of actions? Manual assemblies can be challenging to monitor. Typically, inspection systems are applied to manual operations once the assembly steps are completed, but since the processes are manual there is no automation to understand if assembly steps or sequences were completed correctly. It is a lot harder to get to the root cause of production defects without this type of analysis.
AI-based Computer Vision Advances Quality Management
Quality management is essential for all assembly operations, but especially in manual assembly. In some applications, timing or a specific sequence of actions matter. Many processes struggle with implementing quality assurance protocols because traditional methods reduce throughput by requiring people to modify the operations to compliment the inspection solution, rather than complimenting the operation. How can manual assembly operations be monitored and tracked without disrupting line productivity? The answer is AI-based computer vision technology. This core technology leverages deep-learning algorithms that are capable of understanding human actions and events from video streams. Computer vision has made tremendous advancements in the most recent years. Now production teams within manufacturing can apply computer vision such as Matroid to their quality management systems.
Computer vision can monitor assembly sequences in near real-time, provide feedback to automation systems, give instant feedback to operators, collect valuable analytics on the operations, and much more. In the video sample provided by Matroid, a critical torquing sequence is being monitored. When the engine is assembled it is important to confirm the bolts were properly torqued and the assembly was torqued in the right sequence. Sequence is essential for proper sealing of the gaskets and alignment of internal shafts and bearings. An improper torquing sequence could lead to catastrophic failure to the engine and potential recalls for the OEM. In this application the serialized engine is tracked through the production processes. The smart tool sends feedback to the MES, ensuring proper torque was achieved. Matroid's computer vision solution is inspecting the human action with the specific tooling confirming the right sequence of torquing was completed. The results of the inference are passed to the MES to help provide the highest levels of quality management, tracking and traceability of the manual assembly processes. Operators are provided visual feedback so mistakes are caught faster, if improper sequencing is detected alerts are sent and data is made available. AI-based computer vision opens up an entirely new toolbox of solutions for quality management. Now companies are able to advance their quality management strategies across an enterprise.