内容来源：https://aibusiness.com/generative-ai/siemens-nvidia-industrial-ai-reduce-engineering-effort

内容总结：

英伟达与西门子正深化合作，在德国埃尔朗根西门子电子工厂部署数字孪生、AI智能体及合成训练数据等系统，以加速生产并降低成本。该项目是双方今年1月公布的“全球首个完全由AI驱动的制造工厂”计划的一部分，被视为西门子在自身工厂中验证工业AI实际价值的重要一步，而非仅仅展示技术潜力。

在具体应用方面，其中一项技术利用AI从数字孪生中生成印刷电路板组件的逼真图像，可在实物原型完成前用于训练检测系统。双方声称，该方法可将工程工作量减少最多85%，上市时间缩短最多40%。另一项应用则使用AI智能体对生产场景进行建模和预测，并通过工厂数字孪生在实际实施前验证拟议方案。英伟达提供Cosmos世界模型用于生成质量检测的合成训练数据，其工厂运营蓝图支持生产规划中的AI智能体，Omniverse库则用于结合真实工厂数据对方案进行模拟验证。

目前，埃尔朗根工厂已在生产、物流、质量检测和维护等环节使用超过100种AI算法。西门子计划到2030年在该工厂转型上投资约5.59亿美元，并将在未来数月推出更多应用。

分析人士认为，西门子在自有工厂部署AI具有示范意义，有助于推动更多终端用户和系统集成商采纳工业AI。所披露的效率提升数据——如人工重排产工作量减少最多90%、PCB检测工程工作量节省最多85%——为与客户进行严肃的投资回报率讨论提供了具体依据。不过，分析人士也指出，这些节省数据描述的是西门子自有高混合工厂的潜在收益，尚未在多供应商的既有工厂环境中得到验证，规模化证明仍待观察。此外，升级现有基础设施的成本与复杂性、数据就绪度、安全及治理等问题也是推广障碍。从一次性项目转向可支撑多类AI用例的共享工业AI平台，被视为行业发展趋势，但能否在不同生产环境中复制承诺的效率收益，将是下一阶段的考验。

除西门子外，欧姆龙、ABB和罗克韦尔自动化等企业也已与英伟达合作，利用其Omniverse技术推进工业应用，反映出制造业对AI与仿真技术日益增长的兴趣。

中文翻译：

赞助内容由Google Cloud提供
选择你的首批生成式AI用例
要开始使用生成式AI，首先应聚焦于那些能够改善人类信息体验的领域。
合作伙伴正在利用数字孪生和AI智能体，作为打造全球首个AI驱动制造工厂路线图的一部分。
英伟达正在德国西门子的一家工厂推出一系列AI系统，而德国正寻求巩固其在全球工业AI竞赛中的地位。
这一进展于本周公布，属于英伟达在德国AI路线图的一部分，建立在两家AI与制造业巨头今年1月公布的计划基础之上——双方计划开发他们所描述的全球首批完全由AI驱动的制造工厂。
根据协议，英伟达正在西门子位于德国埃尔兰根的电子工厂部署包括数字孪生、AI生成训练数据和AI智能体在内的系统，以加速生产并降低成本。
这些技术中包括一款应用，利用AI从数字孪生中生成印刷电路板（PCB）组件的照片级逼真图像。
这些虚拟副本可在物理原型就绪之前用于训练检测系统。两家公司声称，该方法可将工程工作量减少最多85%，将上市时间缩短最多40%。
第二款应用利用AI智能体对生产场景进行建模和预测。随后利用工厂的数字孪生在生产计划实施前对其进行验证。
英伟达正在提供其Cosmos世界模型来生成用于质量检测的合成训练数据，而其工厂运营蓝图则支持AI智能体进行生产规划。英伟达Omniverse库用于将拟议方案与真实工厂数据进行模拟比对。
据合作伙伴称，埃尔兰根工厂已在生产、物流、质量检测和维护环节中使用超过100种AI算法。西门子表示，计划到2030年在该工厂转型方面投入约5.59亿美元。
未来几个月还计划推出更多应用。
Interact Analysis分析师Samantha Mou表示，该联合项目的意义在于西门子在自己的运营中展示了这项技术，而非仅仅宣传其潜力。
“在我看来，这是朝着鼓励更多终端用户和系统集成商在工厂中采用工业AI迈出的有意义的一步，”Mou说。“通过在自己的工厂部署AI，西门子能够展示其实际价值，并与客户分享第一手经验。”
Interact的另一位分析师Matthieu Kulezak表示，所报道的效率提升为潜在客户提供了具体数字，便于他们在评估PCB检测等项目的投资回报率时加以参考。
“他们所指出的结果——人工重新规划工作量减少最多90%、PCB检测工程工作量节省最多85%——正是那种具体数据，能够让与客户进行认真的ROI对话变得更加容易，而不是停留在愿景幻灯片的层面，”他说。
Gartner分析师Jonathan Davenport认为，更重要的是从一次性项目向共享平台的转变。
“需要理解的一个重要点是，部署单独的工业AI点解决方案无法实现规模化，”他说。
“我们从西门子和英伟达身上看到的是工业AI平台的开发，在该平台上可以部署多种AI用例，并由数字孪生、仿真能力和AI模型提供支撑，从而实现从孤立应用向更集成、更可扩展的自动化迈进，”他继续说道。
该平台能否推广到埃尔兰根以外，则是更难回答的问题。
“我想提醒一点：标题中的节省数字描述的是西门子自有高混合工厂的潜在收益，而非在multi-vendor棕地站点中经过验证的节省，因此规模化的证明仍有待时日，”Interact的Kulezak说。
其他障碍包括升级现有基础设施的成本和复杂性、数据就绪度、安全性以及治理。制造商必须证明预期的效率提升足以证明投资合理，同时确保AI应用能够在现有系统和安全要求范围内运行。
“因此，扩展一个用例往往更多的不是重新训练模型，而是为每个站点重建集成和上下文数据集，”Gartner分析师Davenport补充道。
尽管如此，西门子并非唯一一家利用英伟达技术推进工业自动化的企业。
欧姆龙、ABB和罗克韦尔自动化也已与英伟达合作，利用其Omniverse技术开发工业应用，反映出业界对将AI和仿真应用于制造业日益浓厚的兴趣。
下一个考验是，制造商能否在不同生产环境中复现所承诺的效率提升，同时避免集成每款应用的成本和复杂性侵蚀收益。

英文来源：

Sponsored by Google Cloud
Choosing Your First Generative AI Use Cases
To get started with generative AI, first focus on areas that can improve human experiences with information.
The partners are using digital twins and AI agents as part of a roadmap to create world-first AI-powered manufacturing facilities.
Nvidia is launching a range of AI systems at a Siemens factory in Germany, as the country looks to bolster its standing in the global industrial AI race.
The development came this week as part of Nvidia’s AI roadmap in Germany, building on a plan the AI and manufacturing giants revealed in January to develop what the companies described as the world's first fully AI-powered manufacturing plants.
Under the deal, Nvidia is rolling out systems including digital twins, AI-generated training data and AI agents at Siemens' Electronics Factory in Erlangen, Germany, to accelerate production and cut costs.
Among the technologies is an application using AI to generate photorealistic representations of printed circuit board (PCB) assemblies from digital twins.
These virtual copies can be used to train inspection systems before physical prototypes are available. The companies claim the approach can cut engineering effort by up to 85% and time-to-market by up to 40%.
A second application uses AI agents to model and predict production scenarios. A digital twin of the factory is then used to validate proposed plans before implementation.
Nvidia is providing its Cosmos world models to generate synthetic training data for quality inspection, while its Factory Operations Blueprint supports AI agents in production planning. Nvidia Omniverse libraries are used to simulate proposed plans against real factory data.
According to the partners, more than 100 AI algorithms are already in use at the Erlangen site across production, logistics, quality inspection and maintenance. In total, Siemens said it plans to invest about $559 million in the site's transformation through 2030.
Further applications are planned in the coming months.
The significance of the joint project lies in Siemens demonstrating the technology in its own operations, rather than simply promoting its potential, said Samantha Mou, an analyst at Interact Analysis.
“To me, this is a meaningful step towards encouraging more end users and integrators to adopt industrial AI in factories,” Mou said. “By deploying AI in its own factories, Siemens can demonstrate its practical value and share first-hand experience with customers.”
Matthieu Kulezak, another analyst at Interact, said the reported efficiency gains give prospective customers concrete figures to consider when assessing ROI of projects such as PCB inspections.
“The results they point to of up to 90% less manual replanning effort and up to 85% savings in engineering effort for PCB inspection are the kind of specifics that make it easier to have a serious ROI conversation with customers, rather than staying at the level of vision slides,” he said.
The bigger story is the shift from one-off projects to a shared foundation, according to Jonathan Davenport, a Gartner analyst.
“One important point to understand is that deploying individual industrial AI point solutions will not scale,” he said.
“What we see from Siemens and Nvidia is the development of an industrial AI platform on which multiple AI use cases can be deployed, supported by the digital twins, simulation capabilities and AI models needed to move from isolated applications to more integrated and scalable automation,” he continued.
Whether that platform can travel beyond Erlangen is the harder question.
“One caveat I would keep in mind: the headline savings figures describe potential benefits at a Siemens-owned, high-mix plant, not savings proven across multi-vendor brownfield sites, so the scaling proof is still to come,” Kulezak, of Interact, said.
Other barriers include the cost and complexity of upgrading existing infrastructure; data readiness; security; and governance. Manufacturers must demonstrate that the expected efficiency gains justify the investment, while ensuring AI applications can operate within their existing systems and security requirements.
“Scaling a use case is therefore often less about retraining the model and more about rebuilding the integrations and contextual data sets for each site,” Davenport, the Gartner analyst, added.
Still, Siemens is not alone in using Nvidia’s tech to pursue industrial automation.
Omron, ABB and Rockwell Automation have also collaborated with Nvidia on industrial applications using its Omniverse technology, reflecting a burgeoning industry interest in applying AI and simulation to manufacturing.
The next test will be whether manufacturers can reproduce the promised efficiency gains across different production environments, without the cost and complexity of integrating each application eroding the benefits.