… by Sudhanshu Mittal, Head & Director, Technical Solutions, Meity Nasscom CoE
When we talk about the precision manufacturing, we are talking about a production system that is capable of producing parts and assemblies within a very tight tolerance and repeatedly across batches, shifts and operating conditions. Industries like Automotive and Aerospace components, medical devices etc are heavily dependent upon precision manufacturing. For an automotive, even the deviation of few microns can lead to increased vibrations and early component failures. In aerospace sector it becomes even more crucial – the level of tolerance is much lower and impact of failure much higher. An aircraft engine may have tolerance level of 1-2 microns. A rocket system flying at speed exceeding multitude of sound speed, may have tolerance level of less than 1 micron. In medical devices minor error may adversely impact the device reliability, cause regulatory failure and lead to loss of brand trust.
It is crucial to ensure that the manufacturing processes can deliver the required tolerance consistently. On paper the machines are accurate but in practice there are minor variations due to various causes – operator fatigue, minor setup deviation, hasty shift changeovers and the basic fact that people are people and they work differently from each other. Any slippage in precision would lead to rework, delayed deliveries and unhappy customers.
Robotics is a growing field addressing the challenges in repeated tasks. Robots don’t face the shift change problem, operator fatigue or change in behaviour and can execute same task with same speed, force and movement path. Once correctly programmed, robots can deliver sustained quality in batches and shifts. Tasks like screw tightening, presses, material placement and inspection are repeated tasks where robots are able to improve the first pass yield and eliminate rework.
Robotic solutions are primarily two categories – Industrial robots and Cobots. While both are robotic solutions aimed for precision, there are fundamental differences also. Industrial robot design assumes minimal human interaction, low variability and overall controlled working environment. Their feedback mechanisms are based on external sensing system like camera, laser or sensors mounted at end and will achieve precision using the preprogrammed and deterministic algorithm. Cobot designer on the other hand, assumes that a Cobot has to work alongside the humans and is designed accordingly. For instance, while Cobots may also use the vision camera, those are also equipped with the contact sensors (maybe 6 axis) in the joints and ends, which enable them to “feel” the contact. The data from these sensors is analysed in real time and used to make corrections, thereby ensuring that Cobot is able to employ encoders and other capabilities to address any misalignment and maintain stable contact with the parts and tools. This allows the Cobots to be effective in working in human proximity, address the part variation and drift, while meeting the required tolerance limits.
Since Cobots are designed to work in proximity with humans, safety is the most critical aspect. Cobots have design guardrails to detect the human presence and slowdown or stop to prevent any collision. The design allows easy programming / reprogramming for different parts, different tasks and different kind of production runs. Humans are expected to handle the cognitive tasks like position delicate component in press and insert operation, post which the Cobot will take over to apply controlled force and monitor displacement. In case Cobot observes the resistance beyond its expectation, it may alert the human operator to address the issue. Over time the Cobot is expected to learn from these and build expertise in its control logic.
While Cobots have seen adoption in various sectors like agriculture, cleaning, restaurant, retails and others, those have also seen strong adoption in industries like Automotive (gluing, welding, engine oil injection, quality check etc), Electronics (precision assembly, soldering, inspection of circuit boards etc), F&B (packaging, labelling, handling especially the delicate items), Pharmaceuticals (handling the delicate and hazardous materials). Some of the most common areas for adoption of Cobots are:
- Loading and unloading the parts for CNC machines, injection moulding and stamping.
- Applying adhesive, screws tightening, soldering, precise component placement.
- Pick-and-place, moving items between workstations.
- Part inspection, defect removal
- Polishing, grinding, sand blasting, paint and sealant applications.
- Stacking in warehouse.
The Cobot deployments usually start with pilot which focuses on addressing specified bottleneck. The plant manager will identify an operation having measurable quality loss, repeatability and manual variability requirement and deploy a Cobot. Once the pilot works successfully (higher first pass yield / lower rework), the focus will shift to the robustness. These experiments would bring out other kind of issues – calibration requirements, operator effect, tool variability issues and others. Once these kinds of environmental issues are identified and addressed, only then the scaling will be undertaken – the scaling by replication. Additional stations will be brought up with Cobot and operation will start shifting from engineering to shop floor operators. Precision across different lines will be measured continuously to identify any new environmental variable that comes up. As the deployment starts maturing, it will be reproduced across different products and plants.
While Cobots promise great capabilities, many challenges need to be addressed before successful deployment can happen. There needs to be repeatable and stable tasks that are to be offloaded to Cobot. Appropriate coordination with the operator along with operator skilling in reprogramming the Cobot is critical. The deployments like any other technology deployments, will suffer from initial hiccups so having unrealistic expectations about Cobot capability can hinder the adoption. Ultimately, we need to ensure that we have stable operating processes, strong tooling and calibration protocols and human engagement to ensure the success of Cobot deployments.
Cobots and AI
While Cobots in Precision Manufacturing don’t use AI because precision manufacturing requires determinism, repeatability, auditability, explainability, there are other areas where AI may become useful. These include vision-based part identification in high mix production when parts may vary visually and with unpredictable orientation. There may also be the requirement for identifying the early tool wear & tear, which would get detected through vibration signature anomaly (AI would be used only for the anomaly detection and not for motion control). In some cases, Cobot may have capability to use the activity database (insertion, tightening) and start working out the optimization (perhaps increased motor speed in beginning and then slowing to stop), however this is still optimizing the process and not the actual task. The actual task requires high precision and AI is not suitable for that purpose.
Going forward the Cobot usage will be aided by improved capabilities in sensors, motor control and selective usage of AI capabilities focused on vision, anomaly and predictive maintenance. With the increase in deployments, standardized Cobots will become normal part of the plant operation and address the scaling challenges in repeatable and automatable tasks.
