MES & machine learning
Last updated on October 25th, 2022
Manufacturing is all about efficiency.
The trouble is, in order to become more efficient, you need to find ways to make more with less. This sounds counterintuitive (and not at all considerate of the law of conservation of mass), but it’s really a matter of making the most of the least. The same operation can vary dramatically between or even within factories, depending on how well operations run. In other words, your production output is only as efficient as your process.
Draw curtains, cue spotlight, and enter the manufacturing execution system, or MES. MES monitors the process of transforming raw materials into finished products by connecting and collecting data from machines and work stations on the factory floor. Used correctly, a MES can ensure effective execution of manufacturing operations and improve production efficiency.
To understand why MES is key to manufacturing efficiency, and process improvement let’s start with the data collection process itself: How does data from the line make it to a MES?
Manufacturing Data Collection Process
Before data makes it into the MES, it first passes through:
- Measuring instruments and actuators (like sensors and valves)
- Control Systems (like PLCs)
- Supervisory Control Systems (like a SCADA)
On the line, data is gathered straight from the source. Flow meters, level switches, temperature sensors, and other instruments measure key values as components are being processed. Valves and pumps, among other actuators regulate your measuring instruments, enabling engineers to control variables on the line. In other words, actuators are used to change a measured parameter, like flow or pressure.
Programmable Logic Controllers, like PLCs, interact with sensors (and sometimes actuators). They can be thought of as “tellers” and “reporters”; they tell a machine or tool on the line what to do, and then report whether the command was executed. If it wasn’t, the PLC reports why. In both cases, it aggregates the data and sends it to MES or other systems for reporting.
Let’s look at a cylinder head rundown station. An engine block comes into the station with a newly located cylinder head sitting on top. Once the switch is made in the station saying the part is in position, the serial number and model number are read from one of any number of devices (cameras for barcodes, Radio frequency ID tags that ride on the part, etc). That information is then sent to the PLC. The PLC then sends that data to the DC fastening tool controllers, and the controllers acknowledge the model type and select the pre-configured program they will run. The PLC tells the carrier assembly holding the tools to lower into position and to slowly rotate the tools. This allows the socket head to engage the bolt it will fasten. The DC tool then spins and tightens the fasteners.
This produces a torque vs time and angle vs time process signature and records final torque and final angle values for each bolt. Those values, as well as the pass/fail status for each bolt, are sent to the PLC.
The PLC will then append that data to the serial number and model number, and send it up to a SCADA or a MES. The process signatures may be stored in other locations as well (several options) for reprocessing later.
MES are used to manage, track, control, document and improve operations or processes; however, the core capabilities of MES and SCADAs sometimes overlap.
On its own, a SCADA can acquire information on traceability, data acquisition, reporting, secure user management, maintenance management, or other process-level tasks. Operators or technicians can visualize their assembly lines as a whole, and make changes to processes when needed. A SCADA with extended MES functionality might be able to store real-time data at a granular level, and track production processes through its user interface. On the other hand, a MES system with custom extensions can allow operators to visualize production units in a way that is similar to a SCADA. Ultimately, SCADA and MES can be used separately or in tandem, and a manufacturer would opt for the system (or combination of systems) that best suits their needs.
MES is often connected to an organization’s enterprise resource planning (ERP), with the ERP system taking data from one or more manufacturing execution systems , centralizing all of the information about a business. Everything from procurement, to accounting, to maintenance—even scrap and rework rates—is all available in one place.
Through this hierarchy and complete MES integration, engineers are able to observe the entire product life cycle. Ideally, a component’s complete history could be followed from when the raw material enters the plant to final end-of-line test, through each and every operation and critical step in the process.
It’s here that machine learning can combine with MES to unlock relationships between different sources and signals, providing complete insight into a facility’s operations.
In a perfect world, data would be passed through this hierarchy flawlessly. However in reality, the data that makes it to MES varies from plant to plant. And so, the question arises: What data does a MES need to collect in order to apply machine learning?
Data Traceability and Completeness for Machine Learning
Even within a single facility, older production lines might rely on manual labour to aggregate data. On the other hand, newer lines might only gather basic data from a few available sources, collecting little more than downtime reports and scrap rates. Ideally, a line should have a fully integrated MES, meaning that the available industrial data is traceable and complete.
Although it’s possible to attain the levels of data traceability and completeness necessary for machine learning without a MES, those who have a MES typically are able to automate more steps, and launch a successful machine learning project with greater ease.
Tracebility, in this context, enables units to be identified on an assembly line. If an error occurs partway through the assembly process, and the unit is scrapped or reworked, traceability is what decides whether or not it’s still the same unit. Ultimately, traceability depends on a manufacturer’s ability to track every activity related to a part and product from the second when raw materials enter the factory to when completed products are shipped.
Traceability is also essential for machine learning applications in manufacturing. If you want to use machine learning to determine how OP10 is affecting OP20, you need traceability to link the data coming from those two operations.
Data completeness is a matter of granularity. Think of measuring the distance between two planar features on a part. In most cases, a line worker would use three probes for each surface, but only record one number for the distance. That may have been good enough in the past, but it’s not sufficient if you want to leverage machine learning to improve manufacturing quality.
Machine learning models need all three measurements in order to represent the part’s planar features as accurately as possible. Without them, the model will operate as if both planes are flat, even if they aren’t. In essence, the more granular your data is, the more opportunities you have to gain insights from it.
Data traceability and completeness are often the result of a fully integrated MES. With these systems in place, machine learning can be easily integrated into and leveraged within a process.
MES Integration For Machine Learning
As the manufacturing sector continues to embrace digitalization, fully integrated manufacturing execution systems will become more and more useful for managing facilities. However, it is expensive for a plant to fully revamp their IT infrastructure. Manufacturers with partially integrated or non-existent MES won’t upgrade unless there are benefits that outweigh the costs, and returns that can be realized.
Incorporating a MES and subsequent machine learning platform into a facility’s or organization’s infrastructure reduces the cost of manual data processing. Tasks that have traditionally taken hours of manual labor, such as aggregating line data to identify trends, can be automated and completed in minutes or less. In this case, machine learning isn’t competing with statistical process control (SPC) or other traditional quality methods; it’s augmenting them so that engineers spend less time to get better insights into their operations.
A similar point applies for diagnosing product or production issues. Root cause analysis can involve dealing with hundreds or thousands of signals from the production line, depending on how much data is collected. Using automated signal pruning, machine learning can reduce the number of signals requiring manual investigation by more than 99%, cutting the time for root cause analysis from weeks to hours.
In addition to providing visibility and reducing the need for manual data analysis, machine learning can also benefit manufacturing by enhancing end-of-line testing, particularly for complex assemblies, such as engines or transmissions. By generating predictions of whether and how likely units are to pass or fail an end-of-line test, machine learning removes the need to test every unit at the end of the line. Instead, manufacturers can focus on the subset of units where predictions fall below a specified threshold.
Finally, by combining the insights gained by incorporating machine learning into a production line, manufacturers can use those insights not only to solve production problems, but actually avoid them all together. By identifying the key decision points during assembly which contribute to product failures, a machine learning platform can recommend which parts to mate together in order to minimize the potential for quality escapes from the finished unit. This approach has been shown to reduce the rework rate for axle assemblies by 65%.
Ultimately, the benefits of machine learning in manufacturing come down to the application and the data being collected from it. With the right set of information and the right machine learning approach, manufacturers can reduce their scrap and rework rates, improve product quality, and increase production throughput.
Share on social: