跳至主要内容

Master AI Manufacturing Company Certification · 第 1 级

AI Foundations for Manufacturing Professionals

为生产、车间运营、程序、质量、维护、工程支持、供应链、库存、安全、员工、财务、工业系统、数据、自动化和制造业领导工作,建立负责任的 AI 基础。

级别
1
模块
16
预计时间
40 小时
课程
版本 2026.4

六级路径

MAMC 计划的每个级别

各级别按顺序完成。每一级都有自己的模块、知识检测、期末考试和证书;第 3 至第 5 级增加经审核的结业项目,第 6 级增加讲授式作业。

  1. 第 1 级 · 基础

    AI Foundations for Manufacturing Professionals

    40 小时 · 16 个模块 · 开放报名

    • Modern Artificial Intelligence in a Manufacturing Company
    • What AI Can and Cannot Reliably Do With Plant Information
    • Export-Controlled Technical Data, Deemed Exports and AI Tools
    • Trade Secrets, Process Knowledge and What Never Leaves the Plant
    • Who Owns the Decision: Engineering, Quality, Safety and Production Accountability
    • Law, Standard, Certification Scheme or Contract: Telling Them Apart
    • 以及另外 10 个
  2. 第 2 级 · 认证专业人员

    Certified AI Manufacturing Professional

    40 小时 · 14 个模块 · 开放报名

    • From Foundations to Practice: The Level 2 Operating Model
    • Prompting Architecture for Manufacturing Tasks
    • The Plant Prompt Library: Ownership, Versioning and Retest Triggers
    • Controlled Drafting From the Released Document Set
    • Work Instructions People Can Actually Follow at the Machine
    • Nonconformance, Containment and Corrective Action Drafting
    • 以及另外 8 个
  3. 第 3 级 · 高级运营

    Advanced AI Manufacturing Operations

    40 小时 · 14 个模块 · 开放报名

    • From Task to System: The Operations View of Plant AI
    • Mapping a Plant Workflow Before Automating Anything
    • Risk-Tiering Plant AI Tasks by Consequence and Reversibility
    • Writing an SOP a Supervisor Can Enforce on Every Shift
    • Building a Retrieval Corpus From Controlled Plant Documents
    • Retrieval Failure Modes: Superseded Revisions, Wrong Part, Wrong Customer
    • 以及另外 8 个
  4. 第 4 级 · 大师认证

    Master AI Manufacturing Company Certification

    40 小时 · 14 个模块 · 开放报名

    • What a Manufacturer Owes: Accountability for AI Across the Company
    • Writing the Company AI Governance Programme
    • The AI Decision Record and the Evidence Pack
    • Mapping Every Obligation That Binds the Plant
    • Export Control and Controlled Information as a Governance Function
    • Defence Flowdowns, the Certification Programme and Cloud AI Services
    • 以及另外 8 个
  5. 第 5 级 · 自动化专家

    AI Manufacturing Automation Specialist

    40 小时 · 14 个模块 · 开放报名

    • What an Agent May Do in a Plant: Authority and Its Limits
    • The Agent Authority Register and What Never Goes to an Agent
    • Orchestrating a Plant Workflow Step by Step
    • The Integration Map and Data Contracts
    • Read-First Integration With ERP, MES, QMS and CMMS
    • The OT Boundary Nothing Crosses
    • 以及另外 8 个
  6. 第 6 级 · 认证讲师

    Certified AI Manufacturing Instructor

    40 小时 · 14 个模块 · 开放报名

    • Teaching Experienced Plant People: Who Is in the Room
    • Designing a Session Backwards From the Behaviour
    • Curriculum Accuracy When Standards, Rules and Tools Change
    • Facilitating Sceptics, and Running the Same Session on Every Shift
    • Running a Demonstration Built Around a Deliberate Failure
    • Designing Hands-On Labs on Fictional Plants and Data
    • 以及另外 8 个

课程大纲

第 1 级详解:课程涵盖的内容

每个模块都将实用的 AI 能力与工人安全、合格人员审阅、知识产权保护、网络安全、经授权的系统访问、经批准的程序、制造商说明、可审计性、上报和可问责的决策联系起来。

第 01 模块

4 个课时

Modern Artificial Intelligence in a Manufacturing Company

What the systems sold to a plant as "AI" actually are, and why that changes how you check them. This module separates fixed automation, rules, statistical process control, machine learning and generative AI, explains how a language model builds an answer, opens up the retrieval, tools and agents inside plant products, and maps the categories a small manufacturer meets.

  • Fixed Automation, Rules, Machine Learning and Generative AI 35 分钟
  • How a Language Model Builds an Answer 40 分钟
  • Retrieval, Tools and Agents Inside the Products You Are Sold 40 分钟
  • A Field Guide to Plant AI Categories and the Public Help You Can Reach 45 分钟

第 02 模块

4 个课时

What AI Can and Cannot Reliably Do With Plant Information

Fluent output is not checked output. This module shows how a tool invents a standard, a revision or a part number, how tolerance and unit arithmetic fails silently, how to read an accuracy figure next to the recall it is hiding, and why the strongest countermeasure to over-trust is a habit rather than a warning label.

  • Fabricated Standards, Revisions and References 35 分钟
  • Tolerances, Units and the Limits of Context 40 分钟
  • Reading Accuracy, Recall and F1 Honestly 45 分钟
  • Automation Bias and the Discipline of Doubt 40 分钟

第 03 模块

4 个课时

Export-Controlled Technical Data, Deemed Exports and AI Tools

The sharpest AI trap in manufacturing. This module explains what technical data is, why releasing it to a foreign person inside the United States can itself be an export, what the published-information and end-to-end-encryption provisions really cover, who must register, and how to decide whether a drawing may enter a system at all.

  • What Counts as Controlled Technical Data 40 分钟
  • The Deemed Export Rule and What a Tool Does With a File 45 分钟
  • Published Information and Encryption: What They Do Not Cover 40 分钟
  • Screening a File Before It Reaches Any System 45 分钟

第 04 模块

4 个课时

Trade Secrets, Process Knowledge and What Never Leaves the Plant

Your parameters, your heat-treat recipe, your yields and your cost model are the plant's crown jewels, and federal law protects them only while you keep them secret. This module covers the statutory definition, the reasonable-measures element a pasted prompt can quietly weaken, what a federal civil action involves, the notice employers owe, and how to classify plant data.

  • What Makes Process Knowledge a Trade Secret 35 分钟
  • Reasonable Measures and the Outside Tool 45 分钟
  • When a Secret Walks: Misappropriation, Remedies and the Notice Employers Owe 40 分钟
  • Classifying Plant Data Before a Tool Sees It 40 分钟

第 05 模块

4 个课时

Who Owns the Decision: Engineering, Quality, Safety and Production Accountability

A tool can draft, extract, summarise and flag. It never dispositions a part, releases a lot, seals a drawing or clears a machine. This module covers the duties that do not move when AI helps: the employer's duty under the general duty clause, the named quality authority, the state question of whether a document needs a seal, and a code that binds only its members.

  • The Employer Duty That Does Not Move 35 分钟
  • The Named Quality Authority and What AI Never Signs 40 分钟
  • Does This Engineering Document Need a Seal? Two State Examples 40 分钟
  • A Professional Code Binds Its Members, Not the Plant 35 分钟

第 06 模块

4 个课时

Law, Standard, Certification Scheme or Contract: Telling Them Apart

An OSHA rule is law. A quality standard is a document you buy. Certification is a scheme sold by a registrar or demanded by a customer. And an automotive or aerospace requirement binds you through a purchase agreement, not a statute. This module teaches the four apart, places the quality standards accurately, and shows where the voluntary AI management standard fits.

  • Four Kinds of Obligation in a Plant 40 分钟
  • Where the Quality Standards Actually Sit 40 分钟
  • A Customer Requirement Is a Contract, Not a Law 40 分钟
  • The AI Management Standard and What a Vendor Claim Is Worth 35 分钟

第 07 模块

4 个课时

OSHA Duties and the AI Standard That Does Not Exist

There is no OSHA artificial intelligence standard, directive or guidance document. This module teaches the duties that do bind a manufacturing employer — the general duty clause, machine guarding, the control of hazardous energy, personal protective equipment, hazard communication and injury recordkeeping — and shows how an AI-related hazard is reached through them.

  • The Employer Duty and the OSHA AI Standard That Does Not Exist 35 分钟
  • Machine Guarding and Hazardous Energy When Equipment Gets Smarter 40 分钟
  • PPE Hazard Assessments and Chemical Information Drafts 35 分钟
  • Injury Recordkeeping, State Plans and What National Figures Can Say 35 分钟

第 08 模块

4 个课时

Robots, Cobots and Machine Safety When Perception Is Learned

What changes when a stop, a speed limit or a separation distance depends on a model that can be wrong. This module separates a safety function from a perception feature, works through the robot-system failure modes OSHA itself documents, names the robot and machinery safety standards with their status and edition, and explains why a vision system is never the safeguard.

  • A Safety Function Is Not a Perception Feature 40 分钟
  • How Robot Systems Actually Hurt People 40 分钟
  • Robot Safety Standards, and Which Edition Applies 40 分钟
  • Collaborative Applications, Exoskeletons and the Integration Contract 40 分钟

第 09 模块

4 个课时

Prompt Engineering for Manufacturing Professionals

How to ask for plant work in a way you can check: the data screen that happens before you type, a nine-part prompt built around a supplied released document rather than a recalled one, disciplined iteration, and patterns for work instructions, nonconformance narratives, maintenance summaries and supplier email. Every output is a draft until a named person owns it.

  • The Screen That Happens Before the Prompt 35 分钟
  • A Structured Prompt for Plant Work 40 分钟
  • Iterating Without Drifting From the Source 35 分钟
  • Prompt Patterns for Plant Documents and Supplier Correspondence 40 分钟

第 10 模块

4 个课时

Verifying a Standard, a Regulation and a Revision Level

An assistant that names a standard, a regulation or an edition is giving you a lead to check, not a citation to quote. This module builds a research workflow that confirms a reference exists, is current, and says what is claimed for this product and this customer, using a withdrawn quality standard edition and a superseded medical-device provision as the worked examples.

  • The Assistant Is Not the Authority 35 分钟
  • Withdrawn, Superseded or In Revision: Reading a Status Field 40 分钟
  • Does This Requirement Reach My Part, My Plant and My State? 40 分钟
  • The Verification Record That Survives an Audit 35 分钟

第 11 模块

4 个课时

Drafting Work Instructions, Quality Records and Supplier Communication

How to get a first draft without a first mistake. This module covers drafting work instructions, maintenance procedures, nonconformance narratives and supplier correspondence from a verified set of released documents, what the record must show about preparer, reviewer and evidence, and why no AI tool drafts a disposition, a release or a conformance statement.

  • First Draft Without First Mistake: Working From a Verified Set 40 分钟
  • Where Drafting Stops: Safety Content in Instructions and Procedures 40 分钟
  • Nonconformance Narratives and Corrective Action: Drafted, Never Dispositioned 40 分钟
  • Electronic Records, Audit Trails and What You Send a Customer 40 分钟

第 12 模块

4 个课时

Machine Vision Inspection and Predictive Maintenance: What the Evidence Actually Shows

Read a detection claim like an engineer. This module takes apart accuracy, precision, recall, F1 and mean average precision, shows what the most-used public datasets actually contain, converts a published miss rate into escaped defects on your own volume, and sets out what a pilot on your own material has to look like before anyone believes a number.

  • Accuracy, Precision, Recall, F1 and mAP: What Each One Hides 45 分钟
  • What the Public Datasets Actually Are 40 分钟
  • From a Miss Rate to Escaped Defects on Your Own Volume 45 分钟
  • A Pilot on Your Own Material, and the Trust Problem Behind It 45 分钟

第 13 模块

4 个课时

Building Repeatable Plant AI Workflows

How a one-off chat becomes a controlled plant process: a six-stage workflow canvas with a trigger, a data-class screen, a named AI step, verification, a named approval and a record. It also covers designing so a wrong output cannot become a wrong action, where an agent's authority must stop, and how to keep the review step honest.

  • From a Chat Window to a Controlled Plant Workflow 40 分钟
  • Designing So a Wrong Output Cannot Become a Wrong Action 40 分钟
  • Agents in the Plant and Where Their Authority Stops 40 分钟
  • Keeping the Review Step Honest and Recording the Approval 40 分钟

第 14 模块

4 个课时

Plant Cybersecurity, Vendor Claims and the OT Boundary

Where an AI tool may sit in a plant and where it may not. This module draws the operational-technology boundary, sets the asset inventory and access basics a plant without security staff can run, teaches you to read a vendor's data-handling claims for what they avoid saying, and shows how a federal contract clause turns a recommendation into an obligation.

  • The OT Boundary and Why No AI Tool in This Level Crosses It 40 分钟
  • Asset Inventory and Access Basics Without a Security Team 40 分钟
  • Reading a Vendor's Security and Data-Handling Claims 45 分钟
  • When a Federal Contract Clause Sets Your Security Floor 40 分钟

第 15 模块

4 个课时

Practical Manufacturing AI Lab

Four hands-on labs on fictional plants. Find the planted errors in an AI-drafted work instruction and nonconformance report; catch a fabricated standard citation and a wrong drawing revision, then rebuild the weak prompt; screen a quote package for controlled technical data and trade secrets; and take apart a vision-inspection claim before drafting a plant AI use policy.

  • Lab One: The AI-Drafted Work Instruction and Nonconformance Report 45 分钟
  • Lab Two: The Fabricated Citation and the Wrong Revision 45 分钟
  • Lab Three: Screening a Quote Package and Designing the Workflow 45 分钟
  • Lab Four: The Vendor Claim and the Plant AI Use Policy 45 分钟

第 16 模块

4 个课时

Capstone: The Plant AI Implementation Plan

The plan you will actually write: choose and scope one low-risk problem, classify every data class it touches and set the controls, name the duties that do not move to a tool, ask the vendor the questions that matter, and define the baseline your own plant measures, the failure indicators, the stop conditions and the improvement loop.

  • Choosing and Scoping a Low-Risk First Problem 40 分钟
  • The Data Classes in Your Plan and the Controls They Require 40 分钟
  • Duties That Do Not Move, the Vendor Question and the Documentation 40 分钟
  • The Baseline, the Stop Conditions and the Improvement Loop 45 分钟

学习成果

您将能够做到什么

  • 评估 AI 工具在制造用例中的适用性、局限性、工人安全、数据安全、知识产权、工业系统风险和专业审阅界限。
  • 为生产、质量、维护、供应链、库存、培训、报告和客户沟通创建可靠的提示词和核实程序。
  • 改善运营文档和分析,同时不把机械控制、安全、工程、维护、质量、网络安全或监管方面的决定委托给模型。
  • 以明确的数据访问、权限、审批、例外处理、审计轨迹、输出检查和上报路径,自动化重复性的办公室和工厂工作流程。
  • 围绕经授权的制造系统、可信的运营数据、经批准的程序和负责任的责任人,规划安全的集成和范围明确的智能体。
  • 制定包含治理、员工发展、网络安全控制、运营指标、实施优先事项和可衡量投资回报的制造业 AI 采用路线图。

评估

Level 1 Final Comprehensive Examination

期末评估衡量您将课程内容应用于真实的生产、质量、维护、供应链、员工、安全文档、工业数据、网络安全、财务和领导场景的能力。

题目
50
时间限制
90 分钟
及格分数
80%
最多尝试次数
3

引领制造业负责任地采用 AI

从完整的第 1 级课程所涵盖的、不偏向任何厂商的基础、安全界限、核实控制措施和制造工作流程开始。

独立的专业教育

MAMC 是一个独立的专业教育计划。它不是工程执照、安全认证、行业执照、监管批准或设备资质,也不能取代合格的工程、维护、质量、安全、网络安全、法律、会计或监管审查。AI 生成的输出需要人工核实。AI 不得独立控制工业机械、绕过经批准的安全系统、授权不安全的设备操作,或使未经授权的人员访问工业系统。设备操作必须遵循经批准的程序、制造商说明、适用的要求和现场特定的控制措施。该计划未获得 OSHA、NIST、ISO、ASME、IEEE、NFPA、任何工程委员会、制造业协会、劳工组织、认证机构、监管机构或政府机构的认可。不声称提供任何专业教育学时或批准,也不保证任何生产率、质量、安全、正常运行时间、合规、盈利或投资回报结果。