AIoT and Smart Factory Integrated Manufacturing: Creating a More Connected Automotive Production Factory
How integrated production data can contribute to enhance traceability, transparency of operations, and manufacturing decision-making
An automotive production line can only be efficient if all the equipment is running as they should. Sometimes, a station is prepared to begin working on the next car, but the part isn't yet in stock. Inventory levels register it as available and production records report increasing delay. To find the reason, engineers have to check all the sources of data.
A pertinent example of this is in manufacturing where a situation arises where individual systems may be able to perform their task but the knowledge as to how to co-ordinate the various processes is disjointed.
Introducing the Artificial Intelligence of Things (AIoT): industrial connectivity, data collection and AI. AIoT enables manufacturers to connect operational data and facilitate more informed decision making. When adopted in the right way, it can increase visibility into production, facilitate traceability and enable teams to proactively identify and investigate issues.
Integrating a smart factory goes well beyond putting something on the machines, or deploying an AI model. It must have data, connectivity, system architecture and clarity of intended function.
- Understanding AIoT in Automotive Manufacturing
AIoT is the integration of the Internet of Things (IoT), which collects data and connects devices, with AI, which can analyze the data, detect patterns or draw conclusions.
Within a car plant, the data sources can comprise robotic stations, conveyors, manufacturing tools, inspection hardware, inventory system, and manufacturing software.
Think of a situation at a particular car assembly plant where there is a bottleneck waiting for parts at a workstation. Machine data might reveal that the equipment was ready, inventory information might show that the parts are in stock, and production data might even show a number of late orders. Combined, all of these sources might lead engineers to investigate whether there is a problem with material flow, scheduling, or equipment.
AIoT is about enabling relevant information to be more easily understood across operational boundaries, not about automating every decision. No matter how intelligent the AIoT may be, the operator, engineer and manager will still need to assess the information and decide what action to take.
- Connecting Existing Manufacturing Systems
One of the biggest hurdles in smart factory connectivity is interfacing equipment and applications that aren't always meant to be compatible.
Automotive plants might have PLCs, SCADA systems, MES, ERP, or other specialized quality applications. Equipment from different generations might use different protocols and data formats.
After capturing the operational problem, the integration effort begins by mapping the information needed to solve it.
For instance, analyzing a repeated delay in assembly may require machine-state events, work-order status, component availability and production timestamps. Gathering all sensors' readings might be unnecessary since it complicates data analysis without providing engineers a better insight.
Other suitable techniques for supporting data exchange are industrial communication techniques, like OPC UA and MQTT. APIs are suitable to integrate applications software, and edge computing is useful to process some data as close to the equipment as possible, if there is a need for speed or a need to save network bandwidth.
The design should be driven by the process needs. Some delay in historic can be tolerated in one section of the process but other elements of information may need to be much faster for immediate operational response. Machine control critical to safety must still be subject to engineering and safety standards.
Prior to scaling an integration, teams need to normalize equipment IDs, timestamps, and event definitions for the relevant systems. If not, information from discrete systems may appear linked, but doesn't accurately reflect the same production event.
- Improving Vehicle and Component Traceability
Traceability enables manufacturers to trace back quality problems, recognize products that might be impacted and determine what occurred during the build process.
Once a vehicle is completed, it has numerous parts, and these parts are subjected to various manufacturing and inspection processes. Suppose a supplier reports one batch of components having a possible defect; the automaker must identify where those components were used and which vehicles might need additional inspection.
Related traceability systems can link component identifiers to work orders, assembly events, inspection results, and VINs. As a result, the history of activity becomes more comprehensive.
Various technologies address different needs: Bar codes can be used to read component identifiers at specific defined points Radio frequency identification (RFID) can provide automated identification in suitable environments Machine vision can be used to read identifiers or monitor existing physical attributes of the products A real-time location system (RTLS) can be used to identify tagged assets within a specified area.
For example, if a batch of parts is under a quality hold, when the receiving records, component scans and vehicle build records are linked, the quality team can more easily identify production records relevant to that batch. Inspection results can be used to determine the extent of further inspection required.
This process may lessen the effort involved in locating relevant record; however, it will not negate the need for physical verification or a quality process. One wrongly marked barcode, a missing scan, or a gap in the production record can threaten the traceability even in the best software.
It is the manufacturers duty to therefore consider the connectivity of their system as well as the integrity of the data it will be recording.
- Using AI to Improve Production Visibility
A connected factory produces significant data on operations, but data collection does not necessarily reveal the reason for different production outcomes.
This is one of the advantages of AI analytics. It can be used to analyze cycle times, machine states, downtime logs, and quality results, for instance. These analytics can assist engineers in diagnosing systemic disruptions, abnormal operating conditions, or production delays.
Some of the other example applications include: Customers will be able to use an equipment's condition, maintenance history, and condition to predict whether or not they need to repair the machine.
Predictive model usefulness is limited by: data quality, the applicability of the model to the operating conditions, and continued validation. Gaps in the data such as missing maintenance history, sensors that give inconsistent readings, and transitions to new conditions can affect model performance.
For other use cases, an alert that is based on a threshold is easier to validate and maintain than an AI model. Which to use depends on whether the extra complexity provides quantifiable value.
Manufacturers can evaluate performance using measures such as:
- Downtime and duration of production stoppages.
- Maintenance workload and recurring equipment failures.
- First-pass yield and quality-related rework.
- Cycle-time variation. And OEE.
- False alarms, missed events and the investigation time for alarms.
They assist teams in deciding if the system is going to help in decision-making instead of just providing more information.
- Coordinating Inventory, Material Flow, and Production
Automotive manufacturing requires the synchronization of assembly lines, warehousing, sequencing stations and internal logistics. Manufacturing may be halted due to lack of parts or poorly timed parts delivery even if the equipment itself is in working order.
Inventory visibility systems, location tracking and production schedule can all contribute to providing teams with a sense of whether, where and when parts and components are available.
Consider how a manufacturer might be able to analyze future assembly specifications against recorded inventory on hand to identify product shortfalls. Location information could point personnel to tagged containers, and production records could reveal the workstation impacted by a late part.
These signals are monitored together. There is a location system, for example, that will tell where a container was last detected, but it doesn't guarantee that container has the right component, or that the material inside passed inspection.
Thus, clear responsibilities to update the inventory, record of material movements and reconcile the contradictory records are required.
Location data about workforce also needs proper governance. When gathering it, manufacturers must specify a valid purpose, limit who has access to it and enforce proper retention and privacy policies.
https://oemnexai .com/
Top comments (0)