What kinds of data are present in a factory environment?
There are many ways to break down or categorize production data. At a high level, there are four common types of data produced in an EMS factory during the production process — data generated by the manufacturing execution system (MES), supply chain data, operator entered data, and machine data.
The MES generates a lot of important data as it runs the factory. The type of data created by an MES can vary. Typically, MES data provides traceability in the form of bar codes or component lists for individual products, visibility into the factory production schedule data, detailed information about each product or process, and recordings of serialized transactions for each step taken to produce a single product.
Supply chain data is ingested from sources outside of the factory environment. It includes information about individual types of material, vendor lists for components, delivery schedules, material data sheets, and measures of individual supplier performance.
Operator-entered logs and downtime tracking represent manually captured data demonstrating reasons for why production stopped or slowed. It is usually collected as part of an overall equipment effectiveness (OEE) monitoring and improvement program.
All of this data is valuable for manufacturers. It can help establish trends, identify the need for additional training, and indicate which suppliers are causing production delays. But collecting and analyzing it requires cumbersome manual tasks and domain knowledge provided by individual personnel, so manufacturers often choose to collect this type of production data only when something goes wrong.
Machine data is different. Its collection does not disrupt the manufacturing process with manual tasks. Machine data is directly captured in an automated fashion from machines or machine vendor-supplied factory monitoring systems. Types of machine data include:
- Performance data that illustrates if and how the machine is doing its job.
- Process data, such as measurements or materials used during the course of the machine doing its work.
- Inspection data capturing the results from tests by equipment such as X-ray or solder-based inspection machines.
- Error or fault data in the form of codes that can be used to set maintenance schedules or take measures to limit downtime.
Regardless of whether or not a manufacturer is collecting all MES, supply chain, or operator entered data, machine data can be collected and used to improve performance across the factory environment.
Machine data characteristics: high data volume, baseline establishment, and automated collection
Machine data differs from other types of production data in several critical ways. First and foremost, the data volume is high. Machine data can be collected from every piece of equipment in the factory environment for every production cycle. For example, if there is a surface mount technology (SMT) line assembling boards at the rate of one every thirty seconds, hundreds of data points can be accurately collected for every second of the process.
Secondly, machines collect data on both normal and abnormal operations in a standardized and repeatable manner. Information about each product is generated and presented in the same way, and the data is complete. Manufacturers capture data from both production cycles that run smoothly and those where a problem occurs. This creates baselines that aid with interpreting data and enable manufacturers to identify abnormalities earlier in the production, helping to avoid issues and minimize downtime.
Finally, the process is automated. Operators do not spend time on data entry and there are no additional capture steps required to make the data usable, so there is no operational burden associated with machine data collection. Removal of these obstacles means all machine data is collected, all the time, without disruption of the manufacturing process.
The Power of Machine Data
All manufacturers track performance. This data is often collected and manually entered by an operator, or the MES generates a report.
Anonymized data from a real-world production line during a timeframe with no changeovers or process disruptions.
Production volume varies from hour to hour, sometimes substantially, but there is very little insight as to why. Machine data from the same production line and cycle can shed light on root causes.
Machine data for the same production cycle identifies takt time spikes that help explain output variations and indicate where issues require resolution.
Instead of simply measuring production volume, the more granular machine data illustrates gaps in production that could be the result of maintenance issues or operator error. It also reveals intervals where production slows, perhaps because of a bottleneck, as well as spans where production runs faster and smoother.
With the help of machine data, advanced machine learning, and statistical analysis, manufacturers can take measures to eliminate gaps in production and optimize output for longer intervals throughout the cycle.
How much machine data is in an EMS factory, and can it all be collected?
The short answer to the first question is: a lot, but it is very manageable. Machines produce data on everything from production defects to sensor readings, parts shortages, and maintenance issues. All of this information is highly detailed, which is why the volume of machine data is so high in almost any factory environment.
So is it feasible to collect all machine data? Yes. With modern data compression and the low cost of cloud data, the price of storage is surprisingly minimal.
Adding up all machine event data, it is a lot. Consider an EMS factory with 10 SMT lines running at high volume 24×7. Each panel produced has 6 boards with 500 total components and 5,000 total pads. For each unit, over 4MB of data are generated. Assuming a consistent 30 second takt time, almost 3,000 boards are produced each day, creating 12GB of data per day, 3.5TB each month.
It wouldn’t take long before even an enterprise data center would be bursting at the seams. Storing that amount of data as files would be a costly non-starter, but modern data compression tools can transform that mountain of data into just 55KB of real measurement data per day. That equates to 8GB for an entire month of high volume, 24×7 production, storable for about $3 per month at current cloud storage prices.
Consider an EMS factory running 10 lines 24×7. The cost of storage for all machine data generated during one month would be three dollars a month.
What are important considerations when setting up a data collection system for analyzing machine data?
As more manufacturers develop and deploy machine data collection systems, industry best practices are surfacing, and systems often adopt similar structures in response to common needs in the factory. Most architectures include these key features:
- There is usually some type of streaming event broker (often called a pub/sub architecture) that receives complex files and reports from production equipment to enable advanced analytics, holistic dashboards and visualization, automated action management, and system monitoring.
- Systems should be able to integrate data from both advanced machines and legacy equipment, such as PLCs.
- They use specialized databases and data lakes for storage.
- Dedicated telemetry and monitoring are deployed to ensure data quality.
There is a great opportunity available for manufacturers who put an emphasis on machine data. Once a data collection system is set up, both its capabilities and its uses can mature over time. Capabilities can expand to include PLCs as well as advanced machines. With a streaming architecture and a specialized database, a new culture of data use can arise among analysts and managers. New applications and business process can access the data, building the informed intelligence of the organization.