en.Wedoany.com Reported - At the 8th Simulation Technology Application Conference, Pera Global Senior Vice President Tian Feng delivered a keynote report titled "Industrial Software Agents and Simulation Practices," systematically elaborating on the evolution logic of industrial software from tool-based and platform-based approaches to agent-based approaches, and showcased the independently developed industrial simulation agent development platform PERA AgentX.
Industrial software is undergoing a profound restructuring from auxiliary tools to autonomous agents. The goal is not to add a dialogue window, but to enable the software to understand engineering objectives, autonomously plan execution paths, invoke computational kernels, monitor operational processes, correct anomalies, and ultimately deliver results. General-purpose large models excel at language and experience summarization but lack adherence to physical laws and engineering standards, making them prone to "logically coherent but computationally erroneous" physical hallucinations. The core of industrial agent development lies in achieving closed-loop autonomous operation of software under knowledge constraints, respecting physical laws while possessing logical reasoning capabilities.
In his report, Tian Feng outlined the evolution roadmap of industrial software from 1.0 tool-based, 2.0 platform-based, to 3.0 agent-based, and proposed a three-step construction methodology for simulation scenarios. First, feeding the brain: inject positive and negative knowledge such as enterprise standards, material constitutive models, historical error logs, and failure cases into the model to reduce hallucinations at the source. Second, reshaping the body: transform traditional software from GUI-driven to Headless components callable in the background, enabling AI to directly dispatch kernels such as geometry, meshing, and solvers. Third, full-chain control: cover geometry cleanup, in-situ self-healing of distorted meshes, dynamic halting of numerical crises, de-rendered data feature insights, and automatic generation of standardized reports containing conclusions and improvement recommendations. Based on this, Tian Feng constructed an industrial software agent maturity model from L1 to L5, and proposed that the business model will shift from License tool authorization to RaaS (Results as a Service), where enterprises no longer pay for software buttons but for validated engineering results.
Focusing on CAE simulation as the core practice scenario, Tian Feng highlighted Pera Global's independently developed industrial simulation agent development platform PERA AgentX. The platform possesses five closed-loop capabilities: task understanding, physical reasoning, configuration decision-making, execution correction, and result interpretation, achieving full-chain unattended operation from natural language instructions to automated simulation reports. In cases such as motorcycle external flow fields, battery pack multi-physics coupling, and automotive crash safety analysis, the agent can autonomously complete modeling, solving, anomaly self-healing, and report output. At the conclusion of his report, Tian Feng stated that Pera Global has achieved a leap from "industrial simulation" to "physical AI."
Pera Global focuses on CAE autonomous simulation R&D and deepens virtual simulation and digital twin technologies based on AI large models, committed to enhancing the digital R&D level of high-end manufacturing. The company has launched the "Jingzhi iGPT Industrial Large Model" and the "GPToIKE" new industrial knowledge engineering solution centered on large models, accelerating the application of AI technology in industrial enterprises.
This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com
