Gangling Industrial Agent
The Multi-Agent OS
for the Steel Industry
A "second brain" for steel plants - perceptible, simulatable, collaborative, evolvable
01

What is Gangling?

A metallurgical industrial-agent development and operation platform for design institutes and steel enterprises

For Design Institutes

Become the "agent factory" of the AI era and a secondary development platform

For Steel Plants

Build a "second brain" covering production, quality, equipment, energy and business

For the Industry

Accumulate replicable, iterable, reusable metallurgical intelligent assets

02

Why Now?

The steel industry is entering the era of industrial agents

VS
Industry Pain Points
PAIN POINTS
  • 01流程复杂：多工序耦合强，依赖专家经验
  • 02系统割裂：MES、DCS、ERP、质量、能源数据分散
  • 03AI 碎片化：单点模型多，难复制、难协同、难运维
  • 04人才断层：老师傅经验难沉淀、难传承
  • 05经营压力：降本、提质、低碳、保交付多目标并行
Opportunities
OPPORTUNITIES
  • 01技术成熟：工业大模型与多智能体技术快速成熟
  • 02能力积淀：设计院具备深厚工艺知识与工程实施能力
  • 03需求强烈：钢厂对可落地、可复制、可进化的 AI 平台需求强烈
  • 04资产汇聚：工业数据、机理模型、知识资产加速汇聚
  • 05趋势演进：行业需要从"单点 AI"走向"全流程协同智能"
03

A Decade of Evolution

From "smart assistant" to "autonomous steel plant"

L1
Assistant
2026—2028
Q&A · Report insight · Anomaly analysis · Assisted diagnosis
L2
Decision Agent
2028—2030
Process optimization · Quality prediction · Equipment prediction · Scheduling advice
L3
Closed-loop Agent
2030—2033
Auto tuning · Cross-process collaboration · Shadow running · Controlled loop
L4
Autonomous Plant
2033—2036
Multi-agent autonomy · Plant-wide optimization · Autonomous operation · Human-in-the-loop
04

Product Architecture

One Foundation, Two Centers, Three Platforms, N Professional Agents

01

Metallurgical Data & Semantics Foundation

Industrial object model · Data governance · OT/IT connection · Unified semantics

02

Model Center + Knowledge & Tool Center

LLM · prediction/optimization/mechanism models | knowledge base · graph · rules · industrial tools · connectors

03

Agent Studio · Agent Runtime · AgentOps

Configuration development / knowledge assembly / workflow orchestration | inference / multi-agent collaboration / access control | evaluation / replay / continuous optimization

04

9 Professional Agents

Blast furnace · Converter · CCM · Rolling · Quality · Equipment · Energy · Production command · Business analysis

05

Core Capabilities

Develop industrial agents like engineering

Agent Studio

Configurable development · Modular assembly · Visual orchestration

Industrial Semantics & Knowledge

Object model · Knowledge graph · Procedure & case accumulation

Model & Tool Center

Mechanism models · Optimization models · Unified AI model registry

Multi-Agent Collaboration

Task decomposition · Capability discovery · Cross-process collaboration

Secure Execution System

Policy engine · Constraint validation · Approval loop · Access grading

AgentOps

Evaluation · Monitoring · Replay · Audit · Version management

06

Collaboration Model

Platform × Design Institute × Steel Plant - building the metallurgical agent ecosystem

Platform Company

Agent foundation · Runtime engine · Model gateway · Security system

Design Institute

Process knowledge · Engineering · Project adaptation · Continuous training

Steel Plant

Business scenarios · Site data · Expert experience · Validation environment

→
Before
  • Doing projects
  • One-off delivery
  • Engineering service provider
After
  • Accumulating agent assets
  • Continuous service & value
  • Industrial AI solution provider

We deliver platform capability, design institutes accumulate industry capability, steel plants gain business value.

07

Scenarios & Value

Start from key processes, evolve to plant-wide collaborative intelligence

01Blast Furnace Agent

Furnace diagnosis · Steady-state optimization · Anomaly alert

02Converter Agent

Endpoint prediction · Alloy optimization · Process advice

03CCM/Rolling Agent

Quality alert · Defect analysis · Parameter optimization

04Equipment Agent

Predictive maintenance · Fault analysis · Downtime alert

05Energy Agent

Gas/power/steam co-optimization · Energy-carbon synergy

Mature-stage targets
1% - 3%Cost per ton
3% - 8%Energy consumption
20% - 50%Quality defects
20% - 40%Unplanned downtime
50%+Scenario replication cycle

* Mature-stage ranges; actual results depend on scenarios, data and implementation depth.

08

Roadmap

From pilot to platform replication, step by step

1
0-90 days · Pilot

Data access · Object mapping · Knowledge import · 3 pilot agents

2
3-6 months · Plant replication

5-8 key scenarios · Shadow running · Evaluation system

3
6-12 months · Platform rollout

Multi-agent collaboration · Closed-loop advice · AgentOps governance · Continuous optimization

4
Year 2+ · Industry accumulation

Agent template library · Knowledge packs · Model packs · Asset-based operation

Local first, then plant-wide Advice first, then closed loop Replicate first, then self-govern Value first, then scale

Low-risk start · Measurable value · Continuous iteration

Build the Steel Industry Agent Ecosystem with Gangling

Joint pilot · Joint development · Ecosystem partnership - join us in building the first demonstration projects

Contact Us