From magazine articles to trade show signage, the Industrial Internet of Things (IIoT) has saturated the manufacturing culture. Graphics are everywhere depicting interconnected factories that showcase how robots, devices and other equipment can be synchronized to provide real-time operational data.
Images like this are helpful but tend to strike a level of fear in most company leaders, as managing risk in an age of connected production is a paramount concern.1 On top of worries about network security, the complexity and expense of monitoring an entire enterprise of factory equipment is quite daunting. More challenging, perhaps, is trying to comprehend what data should be captured and how it will be used to optimize operations.
From free options with basic functionality to pay-per-month alternatives with advanced capability – and everything in between – many management tools capable of collecting data and managing intelligent devices exist today. These options either connect to factory equipment directly or are networked together on a plant-wide local area network (LAN). Some tools can even connect to devices outside of the facility or monitor everything from “the cloud.”
The technological change of pace is already causing traditional machine monitoring methods requiring multiple data collection tools to give way to single point of consolidation solutions, such as Yaskawa’s Cockpit. These emerging machine management platforms are extensible throughout an entire enterprise and incorporate heterogeneous devices, alleviating the problem of a growing population of IIoT devices and incompatible monitoring tools on the factory floor.
Regardless of the implementation method, the main questions to address are: What robots and devices should be monitored and managed and what type of data warrants collection?
Why and how to collect data
For various reasons, there is still a considerable amount of skepticism surrounding widespread use of IIoT and Industry 4.0 technologies and the subsequent management of device connectivity.2 This aside, effective data collection and usage offers undeniable benefits, such as maximized efficiency, increased throughput and reduced costs.
Just as calculating miles per gallon can positively impact driving habits, so can the impact of real-time data collection and visualization for operational equipment have a positive impact on valuable time and resources.
Understanding what devices and technologies can produce viable data can be challenging, but it isn’t impossible. Take a robotic welding workcell, for example. The mechanical pieces doing work are prime data sources that can be analyzed separately or in correlation with other information. On a basic level, if this workcell were connected to a digital network, it could easily provide valuable data.
Real-time data from the robot can include:
- Speed per axis (i.e., top speed and average speed), usually measured in degrees per second
- Time the robot is moving versus time the robot is idle
- An oscilloscope for graphical representation of voltages applied to axes over time
Real-time data from the controller can include:
- Cycle time of the job or part
- Number of cycles complete (including side A or B designation for two-sided postioners or dual workstations)
- Arc starts or weld counts or how many times the torch sparked an arc
- Arc-on time based on the job and active torch travel time
- Average voltage and current of the welding power supply
- Number of tip changes
- Hood filter hours remaining
Real-time data from the power supply and wire feeder can include:
- Arc starts or weld counts or how many times the torch sparked an arc
- Arc-on time
- Wire feed speed and, thus, how much wire is consumed
- Average current
- Average voltage
- Weld
Real-time data from the cycle start, E-stop and door interlocks can include:
- Number of cycle starts or stops initiated
- Number of cell entries
More complex workcells could include data capture of advanced sensors for environment variabilities, such as pressure, temperature and humidity, or mechanical variabilities, such as vibration and acceleration.
How to use it
In this age of digital data, every machine company is developing an interface (or partnering with a company that already has one) to allow customers to visualize and distill data. Capable of sending alerts if a preprogrammed threshold is met, such as a robot sitting idle over 50 percent of the time or average cycle time increases over a given amount, these interfaces enable decision makers to draw correlations about operational activity.
Artificial intelligence can also use this data to make automated changes to a system. For example, the Yaskawa controller can reduce programming and process time by calculating paths based on optimized air-cut angles and directions of the gripper. Each pick on a material handling robot is stored, and like the human mind, the robot learns from experience.
If a specific object is most successfully gripped by a particular approach angle, the robot remembers this and uses this approach first on its next try. Over time, the cycle times and errors are collectively reduced.
Outside of artificial intelligence – which is costly and still in development for greater utilization – one should ask: What should be learned about this workcell and/or factory, and what are the action items to complete from these lessons? A general goal for every company is to operate more efficiently, and digging deeper into the metrics helps determine how to measure and optimize future operations.
Robotic welding workcell data that is simple to collect and easily analyzed on its own includes:
- Alarm log: What errors are occurring, and what is causing them? Does this correlate with cycle times and production? Can this system send real-time alerts for certain alarms when immediate action needs to be taken?
- Cycle time: What is the average cycle time? If a job or process is changed, does this increase or decrease? Does this number fluctuate over a day if charted?
- Arc-on time: Is the most possible processing time being achieved by the workcell? How does this compare to a manual welder? Does this fluctuate over time and match increases in cycle count?
- Cycle count: This is also known as part count, and is usually correlated with cycle time. How many parts are made per shift, per day and per week? Do some operators complete more cycles than others? Can individual goals be set and can rewards be given to operators that exceed goals?
- Consumables metrics: How much gas is being used? Is the weld wire getting low, and should it be replenished before sacrificing a part or adding downtime? Can the cost of each weld be calculated based on wire feed and gas consumption? And, will a different wire, gas or process save money or increase weld quality?
- Maintenance log: Is proper maintenance being completed on the workcell at a frequency that ensures everything is running at a steady level, as to not become too excessive or wasteful? Are the minimum requirements being met to ensure warranty coverage? Are bad welds occurring or is additional downtime an issue if a tip or nozzle is left on for a longer period of time?
Data that can be collected and correlated with other information includes:
- Arc starts versus part count: If arc starts decrease, it would be natural to expect the part count to decrease. However, if part count does not decrease, there is likely a faulty weld. Usually, this needs to be remedied by adjusting weld parameters or reprogramming the part, ensuring that the part being welded meets quality standards. Although rare, a steady part count with decreased arc starts indicates a problem with the power supply.
- Cycle time versus arc-on time: If cycle time is not reasonably close to arc-on time, it is likely one of two issues: additional idle time by the robot(s) caused by poor programming or slow loading/unloading by the operator. Efficient programming should reduce air-cut time, reduce winds/unwinds of the torch and balance the workload for multiple robots. If programming is good, an operator may be taking longer to load and unload parts before starting the next cycle. With large or complex parts, this may be inevitable, but checking with the operator to see if the tooling is cumbersome or if the parts are deformed is a good idea.
- Idle versus teach versus uptime: A robot aids in maximizing productivity, and if it is sitting idle longer than it is being used, achieving ROI over the designated payback period will be a struggle. If a robot is sitting idle more than it is profitable, an investment in teaching time to build more jobs may be advantageous, especially if uptime could greatly increase. Subsequently, more teaching could contribute to faster cycle times, as well.
- Alarm occurrences versus cell entries versus maintenance: There are various reasons for an increased number of alarms. If maintenance is skipped or delayed, chances are it will trigger an alarm. Similarly, if cell entries occur, this could correlate with teaching time. So, if collision detection alarms keep occurring, an operator may need to touch up a job due to a change in parts or tooling. If alarms are being ignored, re-establish goals for uptime and cycle count with the operator or workcell champion. Incentivizing workers to fix issues may also increase productivity.
Data can solve the mysteries that many companies fail to see. Visualizing the data with interfaces provided by the robot OEM or other equipment manufacturers can help track performance and maximize the full value of investment. Goals should be set around metrics and increased incrementally to meet continuous improvement initiatives and boost production. As with any piece of equipment or software platform, best practices should be followed, ensuring maximum performance and ROI.
References
1. Industry 4.0 and Cybersecurity: Managing risk in an age of connected production, Deloitte, 2017
2. PMMI Vision 2025 Report, The Association for Packaging and Processing Technologies (PMMI), 2017
