A factory may have hundreds of machines connected to each other, thousands of sensors, dozens of software systems and yet the factory still isn't able to answer:
What's currently taking place on the production floor and the reason?
That is one of the less obvious issues with a smart factory.
These sensor points give today's manufacturing systems more information in real time than all of the sensory data that was possible a decade ago. Production equipment sensors. Production execution system (MES) touchpoints.
Enterprise resource planning (ERP) touchpoints and PC.
Internet of Things (IoT) sensors for data collection and visualization. RFID/real-time location systems (RTLS). A sensor-based production system.
The issue is that such systems are not creating a single operational picture in an automatic way.
Thus a factory can be extensively linked but not really integrated.
More Interconnections Are Not Always Better How to Get It as a PDF More Interconnections Do Not Always Mean Better Integration More interconnections on a network can mean better integration, but not necessarily true for social networks. Researchers have found that for social networks, an increase in the number of links could have an opposite effect of the connection, such as a loss of a social activity or training. A social network with fewer links, all of which are properly orientated, might be better for the network.
More Data Sometimes helps with a visibility issue.
A manufacturer could use RFID to trace materials, UWB to find assets, other sensors to monitor machines and yet another dashboard to display the results.
All of these technologies have its own advantages.
But what about a car that is late in a production line? Simply knowing the whereabouts of the vehicle wouldn't necessarily be enough to account for the delay.
The team that is producing the part might also need to understand if the needed material can be acquired, if the specific location is working correctly, if an AGV has performed its function, if a fixture is in the position it should be, and if upstream processes had caused the issue.
We need relationships between systems not more systems.
And integration is not about connecting everything without any reason. It's about connecting the right information which needs to be considered together.
Manufacturing Data Has Different Contexts
The same physical process is viewed in different ways by various systems.
Plc could be used to show if the machine is in operation.
An MES could display the active production order.
An RTLS platform could display a map of the location of a tool, a vehicle, or a worker.
An ERP system might indicate inventory availability.
A quality system might contain inspection results.
All of these could be right. They are just not enough on their own.
The Issue is creating a communal manner of working.
Suppose you learn that a machine was halted. That knowledge is more actionable if you can link that event to the part being machined, the production order, the raw material and downstream effects.
That is what makes connected manufacturing more than just tracking.
Time is Required to Project, Receive, and Use 40 Years of Information ## The Physical Universe Has to BE Addressed (REPRESENTED) - properly 1. We need BE capacity 2. 40 years of Time Be 3. Next 40 years of Time/Information for 4. BE ground Truth - what we send to that impact.
A factory does not work on software alone. It works with cars, parts, fixtures, tools, conveyors, robots, operators, racks, AGVs and machines.
If we cannot trust in the identification of the material elements, then digital data becomes un-readable.
These are exactly the kind of technology, for example, RFID, UWB, BLE, machine vision or industrial sensors, that can help facilitate this. They are a means of linking digital events back to physical things.
That's where VIN-level identity comes into play. In car manufacturing, there may be a requirement to relate production information to a particular vehicle, part, station or event during the process.
An effective manufacturing event ought to response extra than what occurred.
It may also need to answer:
Where did it happen?
When did it happen?
Which vehicle or component was involved?
Which process was running?
• What occurred before and after the event?
It's this relationship of location to process and time, which makes operational data more relevant.
An automotive manufacturing example using the integration of these technologies can be viewed here: automotive manufacturing applications.
Integration Does Not Mean Replacing Everything
The misconception that one single platform can replace all current systems sometimes exists with the conception of a smart factory.
This is seldom feasible in a mature manufacturing setting.
Factory systems can be installed in years of equipment and software. They are unique, or complex and tightly integrated with production. If you replace them to try to have a single architecture you're going to add cost and operational risk.
A more realistic approach is often interoperability.
Existing PLCs can continue controlling machines.
MES can continue managing production execution.
ERP will still be able to process planning and business activities.
SCADA can continue providing industrial monitoring.
These layers can then share and provide context for the information.
Other methods include the use of OEM-specific or vendor-agnostic protocols like OPC UA, MQTT, APIs and industrial networking protocols.
The important question is not necessarily:
"How do we replace this system?"
It may be:
"This system needs to leave..."
It can be easier to approach integrations with that change in thought.
Edge Computing Adds Another Layer
Not all manufacturing decisions need all of the raw data to be uploaded to a remote cloud.
Production systems produce a lot of information. Some events require low-latency responses.
Edge computing enables some of the processing to be done nearby to the equipment producing the data.
An industrial gateway can capture information from the equipment, do some filtering of unwanted events, determine critical conditions locally, and send up some information.
It will eliminate bottlenecks in data transport while also bringing operationally critical data closer to the physical process.
It certainly need not be a choice between edge and cloud. A hybrid architecture would leverage both, based on the needs of latency, bandwidth, security, reliability and analysis.
AI Needs Context to Be Useful
AI could just be the 4th and final layer of a smart factory.
But AI doesn't do away with the integration problem.
An algorithm can detect a rare machine behavior, but what is that worth in terms of what the machine is being used for?
Was it a different item variant the machine was running?
Was the line recently reconfigured?
Was material flow already delayed?
Did maintenance occur earlier in the shift?
Was what was found really "off" for that production?
Missing this context, the output of an AI system can result in nigh-technical ideas that are tough for manufacturing groups to operationalize.
Hence, the need for Data Integration and AI to be kept relatively separate as initiatives.
An AI-based manufacturing decision will only be as good as the available data – accurate, reliable, complete and meaningful.
How to Conceptualize Integration in a Practical Way
Not starting with the question, "What technologies do we connect?" Instead, manufacturers can start with the decisions they want to optimize.
A simple framework is:
- Identify
Identify the physical assets, assets, products, machinery or processes that require true identity.
- Connect
Identify the systems that provide data related to the operational problem.
- Contextualize
Link the event with the time, place, production mode, product identity, and process parameters.
- Act
Send the generated information to the users or systems that are in charge of taking the subsequent decision.
In doing so, it stops integration from being a technology-hoarding exercise.
The objective is not to connect everything.
It is to connect what matters.
The Smart Factory Actually Is a Coordination Problem
The greatest change in smart manufacturing might not be the amount of connected equipment.
It is the capacity to coordinate information that flows within the digital/physical dichotomy of manufacturing.
Thus a good architecture should address at least four questions in order of increasing value:
What is happening?
Where is it happening?
Why is it happening?
What should happen next?
The first two rely on accurate recognition and sensing. The third needs contextualized data and insights.
source : https://oemnexai.com/
Top comments (0)