「英中对照」霍尼韦尔联合NVIDIA与Dell推出Forge Cognition设施智能体运维与数字孪生平台

smartbuildingmag.com

15 {{ "2026-10-15T00:00:00+00:00" | date "MMM" }} '26

Written on {{ "2026-10-15T00:00:00+00:00" | date "longDate" }} Modified on {{ "2026-10-15T00:00:00+00:00" | date "longDate" }}撰写于 {{ "2026-10-15T00:00:00+00:00" | date "longDate" }} 修改于 {{ "2026-10-15T00:00:00+00:00" | date "longDate" }}

Honeywell Technologies introduces Forge Cognition with NVIDIA and Dell Technologies to resolve facility alarm fatigue, predictive maintenance, and energy inefficiencies.霍尼韦尔技术公司携手 NVIDIA 与 Dell Technologies 推出 Forge Cognition，以解决设施告警疲劳、预测性维护和能源效率低下问题。

  www.honeywell.com

Facility engineering teams overseeing commercial campuses, high-density real estate, and industrial plants face severe operational bottlenecks caused by fragmented building management systems, recurring false-positive alarms, and uncoordinated equipment control loops. Sifting manually through thousands of disparate telemetry data streams from heating, ventilation, and air conditioning (HVAC) systems, electrical distribution panels, and physical access controls slows root-cause isolation and drives up utility costs. Resolving these operational challenges requires converging information technology with operational technology (IT/OT) through physical AI models, continuous digital twin simulations, and autonomous multi-agent orchestration.负责商业园区、高密度地产和工业工厂的设施工程团队面临着严重的运营瓶颈，其成因包括彼此割裂的楼宇管理系统、反复出现的误报警以及缺乏协调的设备控制回路。人工筛选来自供暖、通风与空调（HVAC）系统、配电盘和物理门禁的数千条互不相关的遥测数据流，会拖慢根因定位并推高公用事业成本。要解决这些运营挑战，需要借助物理 AI 模型、持续的数字孪生仿真和自主多智能体编排，将信息技术与运营技术（IT/OT）相融合。

Agentic Architecture and Physical World Context
Honeywell Technologies launched Forge Cognition as a self-contained agentic artificial intelligence and simulation platform engineered for commercial buildings, university campuses, and industrial operations. Rather than functioning as a passive reporting dashboard, the platform overlays existing building automation infrastructure to contextualize equipment behaviors and map complex dependencies across multi-vendor assets.智能体架构与物理世界上下文
霍尼韦尔技术公司推出 Forge Cognition，这是一个为商业楼宇、大学校园和工业运营打造的一体化智能体人工智能与仿真平台。该平台并非一个被动式报告仪表盘，而是叠加于现有楼宇自动化基础设施之上，将设备行为置于上下文中加以理解，并映射多厂商资产之间的复杂依赖关系。

The software foundation integrates domain expertise with telemetry streams into a structured operational knowledge base. Autonomous software agents query this operational repository to discover systemic anomalies, interpret shifting ambient conditions, and coordinate adjustments across connected supervisory control systems. Designed to support human-in-the-loop governance, the software offers manual, semi-autonomous, and fully autonomous operational modes, allowing facility managers to establish supervisory guardrails and override automated decisions at any time.该软件底座将领域专业知识与遥测数据流整合为结构化的运营知识库。自主软件智能体查询这一运营知识库，以发现系统性异常、解读不断变化的环境条件，并协调各互联监控系统的调整。该软件专为支持人在回路治理而设计，提供手动、半自主和全自主三种运行模式，使设施管理人员能够建立监督护栏，并随时推翻自动化决策。

Operational Capabilities Across Facilities
The platform targets three primary operational bottlenecks across facility management environments: 跨设施的运营能力
该平台针对设施管理环境中的三大主要运营瓶颈：

• Alarm Management and Triage: Specialized software agents classify and verify incoming building management alerts, filtering transient sensor spikes and correlating related faults to reduce nuisance alarms and minimize control room alert fatigue.告警管理与分诊：专用软件智能体对传入的楼宇管理告警进行分类与核实，过滤瞬时传感器尖峰并关联相关故障，从而减少滋扰告警，最大限度缓解控制室的告警疲劳。

• Predictive Asset Health: Autonomous models continuously monitor real-time thermodynamic and electrical metrics across HVAC chillers, air handlers, and pumps to identify mechanical degradation, guiding condition-based maintenance schedules prior to catastrophic component failure.预测性资产健康：自主模型持续监测 HVAC 冷水机组、空气处理机组和水泵的实时热力学与电气指标，以识别机械性能退化，并在部件发生灾难性故障之前指导基于状态的维护计划。

• Energy Intelligence: Real-time digital twin simulations assess proposed operational adjustments, modeling thermodynamic building loads to pinpoint opportunities for energy reduction without compromising indoor air quality or occupant thermal comfort standards.能源智能：实时数字孪生仿真对拟议的运行调整进行评估，通过对建筑热力学负荷建模，在不影响室内空气质量或人员热舒适标准的前提下，精准定位节能机会。

Underpinning AI Compute and Simulation Infrastructure
Forge Cognition incorporates hardware and physical AI architectures developed in collaboration with NVIDIA and Dell Technologies. On-premises and edge computing tasks are driven by the Dell AI Factory with NVIDIA, deploying Dell PowerEdge XE servers alongside Dell AI Data Platform storage engines and integrated cyber resiliency frameworks.支撑 AI 计算与仿真基础设施
Forge Cognition 融合了与 NVIDIA 和 Dell Technologies 合作开发的硬件与物理 AI 架构。本地部署与边缘计算任务由 Dell AI Factory with NVIDIA 驱动，部署 Dell PowerEdge XE 服务器，并搭配 Dell AI Data Platform 存储引擎和集成的网络弹性框架。

The AI execution stack combines multiple NVIDIA technologies:该 AI 执行栈结合了多项 NVIDIA 技术：

NVIDIA Nemotron open models power reasoning agents, including a dedicated True Alarm agent that classifies building events and refines inference accuracy based on facility operator feedback.NVIDIA Nemotron 开放模型为推理智能体提供支持，其中包括专用的 True Alarm 智能体，可对楼宇事件进行分类，并根据设施操作员的反馈优化推理准确性。

NVIDIA Omniverse libraries provide physically accurate digital twin simulations, allowing facility engineers to simulate thermodynamic changes and test control sequences virtually prior to shop-floor deployment.NVIDIA Omniverse 库提供物理精确的数字孪生仿真，使设施工程师能够在车间部署之前以虚拟方式模拟热力学变化并测试控制序列。

NVIDIA Cosmos open models and the NVIDIA Metropolis Visual System Solutions (VSS) Blueprint enable computer vision agents to parse, index, and summarize visual and sensor telemetry feeds.NVIDIA Cosmos 开放模型与 NVIDIA Metropolis Visual System Solutions（VSS）Blueprint 使计算机视觉智能体能够解析、索引并汇总视觉与传感器遥测数据流。

Local inference workloads execute on NVIDIA RTX PRO GPUs, with planned architectural expansion to NVIDIA DGX Spark hardware to accelerate far-edge processing directly within distributed plant and mechanical rooms.本地推理工作负载在 NVIDIA RTX PRO GPU 上执行，并计划将架构扩展至 NVIDIA DGX Spark 硬件，以便在分布式厂房与机房内直接加速远边缘处理。

Additional Context
This section details technical specifications and competitive benchmarking not included in the original product announcement.

Traditional building management systems (BMS) adhere to ANSI/ASHRAE Standard 135 (BACnet), Modbus, or LonWorks communication protocols, capturing point-to-point telemetry at predefined intervals without semantic reasoning. In standard commercial facilities, alarm flooding regularly exceeds the ISA-18.2 / IEC 62682 industrial alarm management benchmark, which recommends fewer than one alarm every ten minutes per operator console during normal operation. Independent field pilot deployments of Forge Cognition—including validation at NVIDIA’s 750,000-square-foot Voyager campus in Santa Clara, California—demonstrated false-alert reductions of up to 90 percent and identified potential utility consumption decreases of up to 10 percent while maintaining occupant thermal comfort scores above 95 percent.补充背景
本节详细介绍了原始产品公告中未包含的技术规格和竞争性基准测试。

传统楼宇管理系统（BMS）遵循 ANSI/ASHRAE 标准 135 （BACnet）、Modbus 或 LonWorks 通信协议，以预定义的时间间隔采集点对点遥测数据，但不具备语义推理能力。在标准商业设施中，告警泛滥经常超出 ISA-18.2 / IEC 62682 工业告警管理基准，该基准建议在正常运行期间每个操作员控制台的告警频率低于每十分钟一次。Forge Cognition 的独立现场试点部署——包括在 NVIDIA 位于加利福尼亚州圣克拉拉、占地 750,000 平方英尺的 Voyager 园区进行的验证——表明误报率最多可降低 90 %，并发现公用事业能耗最多可降低 10 %，同时将 occupants（ occupants 热舒适度评分）维持在 95 % 以上。

Within the smart building and industrial digital twin market, Honeywell Forge Cognition competes with platforms such as Siemens Desigo CC with Building X, Johnson Controls OpenBlue, and Schneider Electric EcoStruxure. While conventional supervisory platforms rely on static rule-based fault detection and diagnostics (FDD) engines hosted primarily in centralized cloud environments, Forge Cognition differentiates by deploying world foundation models (WFMs) and local agentic reasoning on-premises via high-density GPU server clusters. This edge-native approach reduces dependency on continuous cloud connectivity, satisfies strict operational security requirements, and minimizes latency during critical HVAC and life-safety fault interventions.在智慧楼宇与工业数字孪生市场中，Honeywell Forge Cognition 与 Siemens Desigo CC with Building X、Johnson Controls OpenBlue 和 Schneider Electric EcoStruxure 等平台竞争。传统的监控平台依赖于主要托管在集中式云环境中的静态规则化故障检测与诊断（FDD）引擎，而 Forge Cognition 的差异化优势在于通过高密度 GPU 服务器集群在本地部署世界基础模型（WFM）与本地化智能体推理。这种边缘原生方法减少了对持续云端连接的依赖，满足严格的运营安全要求，并在关键 HVAC 与生命安全故障干预期间最大限度地降低延迟。

Edited by Romila DSilva, Induportals Editor, with AI assistance.由 Induportals 编辑 Romila DSilva 编辑，AI 辅助完成。

www.honeywell.com

  Ask For More Information…  索取更多信息…