GM's predictive future for manufacturing
General Motors is using AI, predictive analytics and digital twin technology to improve manufacturing efficiency, reduce downtime and enhance vehicle quality
GM is using AI-driven predictive analytics to help its manufacturing teams optimise vehicle output through historic performance and real-time data analytics of robotics and conveyor systems. It is also using the architecture of its fourth generation EVs help in the assembly process.
The company says the digital technology it is using for maintenance is primarily an internal development that leverages cloud-based platforms and internal data infrastructure, but it is using several partner-developed AI and predictive analytics tools to monitor equipment, inspect quality and optimise vehicle output across its assembly plants. That includes NVIDIA accelerated compute platforms, including Omniverse and Cosmos to train AI manufacturing models for optimising GM’s factory planning and robotics.
We have seen several measurable benefits from applying AI-driven predictive analytics in our manufacturing operations... Recent examples include helping prevent downtime at our Fort Wayne facility and Flint Assembly truck plants
The AI models GM is using are able to learn and discern abnormal patterns that indicate potential equipment malfunctions, according to the company. These early-detection algorithms enable swift, proactive interventions, which can help avert unscheduled downtimes and ensure a seamless vehicle supply chain. Stopping an assembly line is an expensive last resort and can cost anything between $10,000 and $50,000 per minute. Unplanned downtime can also lead to overtime as companies make up for lost output, which adds to labour costs.
“We have seen several measurable benefits from applying AI-driven predictive analytics in our manufacturing operations,” says GM’s spokesperson. “Recent examples include helping prevent downtime at our Fort Wayne facility and Flint Assembly truck plants. These insights enable teams to identify potential issues earlier and act before they affect vehicle production.”
GM uses data to evaluate critical-path assets – those pieces of equipment that can cause the most serious, unplanned downtime. As one example, it leverages vibration sensors to monitor conveyors, which can threaten production. GM has also developed conveyor machine learning models to monitor the conveyors. Early indications that a piece of assembly equipment will need maintenance to prevent shutdown includes creased motion duration, drive parameters and vibration.
GM processes data on machine performance to provide instant alerts transmitted to a centralised cloud platform. Different levels of edge computing are pulled into the cloud to create a centralised manufacturing data model for further analysis.
Equipment data
GM is also using Device Level Analytics (DLA), a system that collects operational data on equipment in real time through sensors scattered throughout an assembly plant – the eyes across the plant. These sensors monitor variables such as vibration patterns, electrical current, and other information. By analysing this data, GM can anticipate potential equipment failures before they occur, allowing plant teams to schedule preventive maintenance rather than simply reacting to breakdowns.
These tools give our employees earlier, more complete information so they can act before a defect becomes a larger rework or throughput issue... For example, At Factory ZERO, WeldBrAIn assesses every resistance spot weld on every vehicle in real time. The system supports our operators with earlier insights into the weld quality for containment of any issues
“Using DLA, we monitor hundreds of thousands of devices across the build process, and when sensors detect that something may be wrong, our teams can address it before it becomes an equipment failure that stops production,” says GM’s spokesperson.
GM is also using digital twins to help plan assembly equipment lifecycles and maintenance schedules. “Digital twins are allowing us to test manufacturing lines virtually before we build them, helping us avoid costly teardowns,” says GM’s spokesperson.
The carmaker has been using the NVIDIA’s Omniverse platform to create digital twins of assembly lines, allowing for virtual testing and production simulations to reduce downtime. The application of the technology includes training robotics platforms already in use for operations such as material handling and transport, along with precision welding, to increase manufacturing safety and efficiency.
Quality control
Separate to the assembly equipment, smart tools such as DLA, WeldBrAIn and Finesse Paint Repair help GM operatives to quickly identify any quality issues on the vehicles themselves early on. This contributes to productivity and prevents problems further down the assembly line, and where problems are related to the equipment itself, GM can quicky attend to it.
“These tools give our employees earlier, more complete information so they can act before a defect becomes a larger rework or throughput issue,” says GM’s spokesperson. “For example, At Factory ZERO, WeldBrAIn assesses every resistance spot weld on every vehicle in real time. The system supports our operators with earlier insights into the weld quality for containment of any issues.”
WeldBrAIn is an AI-powered system that has expanded quality check capabilities from manual inspection of four parts per shift to checking every single weld, on every single body, in real time.
GM is also using Finesse Paint Repair vision cameras to detect issues with paint quality on vehicles moving down the assembly line. The cameras can detect microlevel imperfections that can be remedied for a better paint finish. The tool also provides the defect data to GM paint process engineers, who can use the information to check equipment and improve the paint process to prevent defects in the first place.
PlantBrAIn platform
As GM works toward its fourth-generation EV platform, which is based on lithium manganese rich (LMR) battery drive and more efficient electric propulsion, digital models are being developed to enable manufacturing teams to test production steps, review equipment placement, and identify safety or ergonomic concerns before new hardware is installed.
GM is also developing a factory intelligence platform called PlantBrAIn to combine plant-floor data with computer models to help teams improve how the new electric vehicles and battery packs are built.
The electrical architecture in the fourth generation EVs will also help in the assembly process because it can test electrical connections as the vehicle moves down the assembly line, eliminating the final validation check once the vehicle has been built.
At this October’s second annual GM Forward event the carmaker outlined what it is doing to halve traditional vehicle development times through advances in enterprise AI and virtualization. GM’s approach is to bring together design, engineering and manufacturing from day one. With regard to the LMR battery, it reported having tested more than 300,000 battery designs virtually before building a single physical prototype, moving from pack concept to formal programme approval in one month.
From preventing costly equipment failures and improving weld and paint quality, to optimising production planning and accelerating EV development, GM is using digital technology for predictive maintenance to make its manufacturing more efficient while reducing costs. As GM advances its next-generation electric vehicle programmes digital platforms and virtual manufacturing models will play a crucial role in helping the company build the next generation of smart electric vehicles with equipment that is every bit as sophisticated.
