An uncommon engineering task in industrial logistics: understanding physical locations where objects, individuals, vehicles, and materials are all in state of constant change. When the pallet, moving from receiving into storage; when the forklift becomes free, then is assigned a task, and proceeds to some another zone; when the trailer backs in to the yard and patiently awaits positioning on the platform; when a needed replacement part, visible in system's memory, is paradoxically invisible in physical realm-all these transformations are occurring in physical space, while the systems managing these transformations, lie within digital space. Here comes the peculiar role of AIoT, where connection between two: IoT, that supplies physical environment's status signals; AI, that gives semantic interpretation to these signals against operations' background.
The real engineering conundrum is not on more data collection; it's transforming physical actions into meaningful operations information.
- Signals are Supplied by IoT IoT allows signal transmission from the variety of sources found in industrial environments: from asset position trackers to RFID tags; from myriad sensors to computers, that visually track assets on floor; fromwarehouse management systems to manufacturing and transportation systems and so forth. While, on their own, these signals only bring us partial context of reality. For instance: Location and identifier of the asset-here, that's 'Asset A', which is in 'Zone B' at '10:42', in state of 'Available'. It may be enough to report whereabouts of the asset, but not sufficient for the operations system to reply to the query: 'Is it required to relocate the asset, if so, what would be best place for that move, what precedence would that carry, which additional resources have to be allocated for that purpose?'.
Here wherecontextualization matters, 2.
AI provides the contextual interpretation AI deals with multitude of events; rather than single record at a time. Warehouses may benefit from following contexts: storage zone, stock availability, processing tasks, crane operations, incoming orders, etc. For manufacturing: a particular production order, materials location (bin in central store, or directly on floor, if staging, as if part of the line,); available devices and number of personnel.
Automation might not be a first order requirement-often only revealing of exceptions, optimizing order priorities, and advising people will do fine. Why it's important is because an AI system doesn't just say: 'The item is located at X position. The AI system says-given the 'X', what would be the consequence?'.
- Intralogistics is ideal area of AIoT Application Warehouses, in particular have complex transport networks.
Movement of resources-parts, materials; personnel driving lift trucks, pushing containers, etc., occur among stockpiles, intermediate zones and workstations. AIoT has ample opportunities here, supporting a variety of applications, starting with: material feeding in workstations (work line) or to specific work stations(kits), or in series of routine trips in specific order(milk runs) along prescribed trajectories, or just dispatching lift trucks efficiently, or simple, repeated transfer between point A and point B. Here location information determines an asset, that has a certain requirement; that requirement can be compared against what's happening on the floor; this determines if resources are enough or can the workflow accommodate the task at that specific time without unnecessary friction? Example may be material is available but located far, lift truck nearby is assigned somewhere else...
- Warehouses also pose a similar problem One can't optimize a warehouse just because its inventory level is known.
Physical layout and all kind of physical occurrences like picking, movement, stocking or loading procedures are as important as knowing the data in the warehouse system. With AI one may analyze: picker's trajectory; shelf locations for picking activity, material stocking and replenishment routines; traffic congestion; equipment availability; order profile; dock operations, etc.; and hence uncover hidden inefficiencies, unusual long waits, delays, bad logistics paths, or abnormalities. Generally speaking, the purpose of AIoT is to bring system's logical inventory up to speed with the reality.
Namely: inventory level + asset location + movement history + demand on hand = operational context 5. Yard Operations-another complex environment Yards in industrial settings are known to be extremely dynamic. Receiving trucks and then having containers, where many are waiting; yard tractors shifting equipment and materials; appointments changed at short notice; itβs a busyplace.
Asset positioning system can tell where the item is at any point, but the 'where is my truck?'
question is not really asking of that; the more pertinent issue often comes from the state associated: is truck waiting for a pickup, is its spot of choice occupied by something else? Therefore adding operational information like 'dwell time', 'order priority', 'resource availability', 'programmed moves' or 'gate access log' can add more precise data on what's happening in yards. By combining different such sources of data it is possible not merely report asset location but make a deduction. 6. MRO Logistics bridges inventory and maintenance MRO (Maintenance, Repair, Operations) is simply another flavor of an same kind of problem.
If maintenance team is aware that required component is available in storage, the challenging question would be; where?
And is it deliverable before it's actually needed on site?
Components stored on, for example, a remote shelf, a part of a local store in another area, production floor, or designated transfer area all require a specific type of supply. Locating components based on the maintenance need leads from just information to actual readiness; there lies the real significance.
It is not the presence of a component in stock that is relevant, but its availability for maintenance activity. 7. The same type of arrangement can be useful beyond warehouses Applications that need location and tracking information extend much beyond warehouses and yard operations. Cold Chain-where temperatures are highly critical, demand monitoring for any deviation; Bulk Material Handling-where materials such as grains, aggregate and powders need to be moved from large stockpiles into hoppers or silos, from where, by trucks/railcars and finally to various processing equipment; Industrial Packaging-when assembling packed goods, products, materials and auxiliary resources for dispatch to destinations.
In these and more instances the AIoT application paradigm can help in bridging information from physical operation up to a level of an operations insight.
- Engineering challenge is data contextualization.
The complicated aspect of an AIoT application never rests solely on taking another signal, but on determining the relationship between separate entities. Say, if there is an inventory item, there is a machine needing its substitute. A given data point tells us there is 'Item A', located in 'Location X, position B', which is due to be supplied with 'Order # 1', with total weight of X Kg.
Then again, let's recall our machine; it is operating; its throughput, its load factors; all these define the required operations and priority of its maintenance.
Therefore context has to be generated from all relevant entities, linking machine and material together through physical world. This aspect alone may generate multiple engineering challenges such as synchronization or lack thereof of incoming information (data latency), accuracy of identifying an entity, connecting histories of many different phenomena to the present operation state, etc. Quality of resultant decisions heavily hinges upon quality of this contextualisation layer.
- AIoT Decisions are not always high complexity algorithms-a simple logical arrangement, or a workflow example: Signal What occurred? Forklift enters zone with finished material to wait for pick up.
Context Is there another available forklift that should pick up?
Are all other lifts currently busy?
If a resource is available then assign it the material to move.
Decision What's most logical?
Dispatch previously available lift to perform the job.
Action What happens now?
Forklift started to move.
This is a practical example and it highlights where the system operates by observing signals then inferring the best next action to take with an operational view; it may or may not require AI depending on operational logic complexities, but essentially it works.
What AIoT will evolve into?
IndustrialAI will be compelled to look after real world entities beyond business logic alone; integration with time and space-moving with physical items, knowing where they are in a certain operational setup and understanding importance/priority based on the broader system status will be key aspects of future industrial applications of data and intelligence, thus bridging 'where it is' and 'what and what to do about it'. For more info visit: apertureventurestudio.com
Top comments (0)