Knowledge Vault 4 /75 - AI For Good 2023
Accelerating sustainable manufacturing with AI
AI FOR GOOD ML Workshop
< Resume Image >
Link to IA4Good VideoView Youtube Video

Concept Graph & Resume using Claude 3 Opus | Chat GPT4o | Llama 3:

graph LR classDef humancentric fill:#FFD700, font-weight:bold classDef smartmanufacturing fill:#FF4500, font-weight:bold classDef supplychain fill:#8A2BE2, font-weight:bold classDef optimization fill:#7FFF00, font-weight:bold classDef aiapplications fill:#00CED1, font-weight:bold classDef edgecomputing fill:#FF69B4, font-weight:bold classDef robotic fill:#BA55D3, font-weight:bold classDef agriculture fill:#8B4513, font-weight:bold classDef employment fill:#4682B4, font-weight:bold classDef sustainability fill:#20B2AA, font-weight:bold A[Accelerating sustainable manufacturing
with AI] --> B[Dr. Chun Wang:
human-centric
smart
manufacturing 1] B --> C[Intelligent manufacturing:
industrial internet,
digital twins 2] B --> D[Focuses on human
needs, machine
collaboration 3] B --> E[Challenges:
complex systems
modeling,
collaboration 4] A --> F[Malua Narayan:
AI, neural networks
for supply chains 5] F --> G[Neural supply chains:
IoT, AR/VR, AI, cloud 6] F --> H[Neural capabilities:
incremental adoption
stages 7] F --> I[AI benefits:
inventory, transportation,
manufacturing optimization 8] F --> J[Generative AI:
automated business
modeling,
materials 9] A --> K[Lei Ren:
AI frontiers
in industrial internet 10] K --> L[Neurodynamic architecture:
efficient industrial
edge intelligence 11] K --> M[Cloud-edge network:
prediction accuracy,
computing time 12] K --> N[Multi-edge approach:
adaptive fault
diagnosis 13] K --> O[Temporal attention
network: predicts
health indicators 14] K --> P[Self-supervised
learning: enhances
soft sensors 15] A --> Q[Digital twin
robotic system:
synthetic grasp
datasets 16] Q --> R[Transfer learning:
AI models
generalize
to new conditions 17] Q --> S[Big models,
knowledge graphs:
model generalization 18] Q --> T[Sylvain Calinon:
teaching robots
manipulation skills 19] Q --> U[Challenges:
diverse data,
anticipative planning,
multitask learning 20] Q --> V[Uses sensors,
interfaces to
program robots 21] Q --> W[Applications:
underwater valve
turning,
portrait drawing robot 22] A --> X[Yongchang Zhang:
coding-based task
scheduling 23] X --> Y[Mathematical models:
manufacturing,
computational tasks 24] X --> Z[Evolutionary algorithm:
sub-optimal
scheduling solutions 25] X --> AA[Parallel search:
handles 4000+
tasks quickly 26] A --> AB[Ashruf Abushadi:
AI for
smart agriculture
in Namibia 27] AB --> AC[Satellite imagery,
drones: map
invasive species 28] AB --> AD[AI: optimal
harvesting locations
for biomass 29] AB --> AE[Digital tools:
farmers at center
via app 30] A --> AF[Irmgard Nübler:
tech impact
on employment,
skills 31] AF --> AG[Tech creates,
destroys jobs
historically 32] AF --> AH[Adjustment:
reduced hours,
R&D,
education 33] AF --> AI[Robotization
impacts vary
by country 34] AF --> AJ[Tech leverage
differs between
Asia, Latin America 35] AF --> AK[Rapid change:
inequality, work
quality,
surveillance 36] A --> AL[Thomas Sobottka:
sustainable
industrial
manufacturing 37] AL --> AM[Industry:
35% CO2 emissions,
efficiency needed 38] AL --> AN[Barriers:
low renewable
energy,
electrify heat 39] AL --> AO[AI optimization:
improve energy
efficiency 40] AL --> AP[Digital twins:
optimize
real-time data,
energy 41] AL --> AQ[Demo factory:
rapid re-planning
with solar
forecasts 42] AL --> AR[Learning factories:
AI energy
optimization
education 43] AL --> AS[Key takeaway:
AI helps
industry
sustainability 44] class B,C,D,E humancentric class F,G,H,I,J supplychain class K,L,M,N,O,P edgecomputing class Q,R,S,T,U,V,W robotic class X,Y,Z,AA optimization class AB,AC,AD,AE agriculture class AF,AG,AH,AI,AJ,AK employment class AL,AM,AN,AO,AP,AQ,AR,AS sustainability

Resume:

1.- Dr. Chun Wang presented on human-centric smart manufacturing, integrating cyber, physical and human systems.

2.- Smart manufacturing involves intelligent products, production, services, supported by industrial internet and digital twins.

3.- Human-centric smart manufacturing focuses on human needs in product design, human-machine collaboration in production, and service-oriented transformation.

4.- Challenges include system modeling of uncertain complex systems, knowledge engineering across domains, and effective human-machine collaboration.

5.- Malua Narayan spoke about using AI and neural networks to optimize and de-risk supply chains for sustainable manufacturing growth.

6.- Neural supply chains integrate IoT, AR/VR, AI and cloud to be adaptive, enable autonomous operations and self-optimization.

7.- Enterprises can adopt neural capabilities incrementally from "neural-ready" to "neural-adopting" to "neural self-acting".

8.- Case studies showed benefits of using AI for inventory optimization, transportation cost reduction, and steel manufacturing optimization.

9.- Generative AI has potential for automated business modeling to identify new applications for materials.

10.- Professor Lei Ren discussed AI frontiers in industrial internet across edge computing, cloud-edge collaboration, data-driven techniques and digital twins.

11.- Neurodynamic architecture enables efficient industrial edge intelligence by adaptive sequence learning on edge devices with limited compute.

12.- Cloud-edge lightweight temporal convolutional network improves prediction accuracy and computing time for remaining useful life estimation.

13.- Label split and multi-edge approach enhances adaptive fault diagnosis by enlarging diagnosable faults and handling unlabeled data.

14.- Multi-channel temporal attention network predicts industrial health indicators considering channel contributions and reducing delay.

15.- Self-supervised learning explores diverse industrial data characteristics to enhance soft sensors for key indicator prediction.

16.- Digital twin robotic system generates synthetic grasp datasets and enables adaptive robotic grasping in industrial flexible production.

17.- Cross-domain transfer learning and meta-learning enable AI models to generalize to unseen conditions in industrial applications.

18.- Industrial big models and knowledge graphs are an emerging frontier requiring further research in model generalization and domain adaptation.

19.- Dr. Sylvain Calinon presented research on teaching robots manipulation skills from human demonstrations using machine learning.

20.- Challenges include handling diverse data types, combining anticipative planning and reactive control, and enabling multi-task learning.

21.- Uses sensors, XR interfaces and smartphones to intuitively program robots by demonstration and evaluate robot's ability to perform tasks.

22.- Showed applications in underwater valve turning leveraging virtual demonstrations and a portrait drawing robot exhibit.

23.- Dr. Yongchang Zhang introduced coding-based evolutionary operators for large-scale task scheduling in industrial edge computing.

24.- Mathematical models represent complex precedence between manufacturing and computational tasks in IoT-based manufacturing.

25.- Coding-based evolutionary algorithm generates diverse search operators to find sub-optimal scheduling solutions in reasonable time.

26.- Parallel evolutionary search and solution merging enable handling 4000+ tasks in under 10 seconds compared to 400+ seconds sequentially.

27.- Ashruf Abushadi from UNIDO shared a case study of using AI for smart agriculture in Namibia.

28.- Satellite imagery and drones helped map invasive plant species threatening farmland using AI-based predictive modeling.

29.- AI enabled calculating optimal locations and yields for harvesting invasive plants as input to a biomass processing factory.

30.- Digital tools put farmers at the center, incentivizing participation in the solution through a mobile app.

31.- Irmgard Nübler from ILO discussed the impact of technologies on employment, skills, and development goals.

32.- Historically, employment rates have grown despite fears of tech-driven job losses, as tech both destroys and creates jobs.

33.- Adjustment processes like reduced working hours, increased R&D and education, and changing consumer behavior mediate tech's employment impacts.

34.- Robotization and manufacturing employment impacts vary significantly by country based on institutions, skills, and position in value chains.

35.- Capabilities to leverage technology for development differ between Asia and Latin America, enabling Asia to build complex supply chains.

36.- Rapid tech change can worsen inequality, work quality and surveillance, requiring updated institutions and labor protections.

37.- Thomas Sobottka presented on transforming industrial manufacturing for a sustainable future and meeting climate goals.

38.- Industry accounts for 35% of CO2 emissions, requiring dramatic increases in energy efficiency and flexibility.

39.- Barriers include low renewable energy share, need to electrify heat, and volatility of solar/wind power.

40.- AI-based optimization and digital twins can improve energy efficiency in design and synchronize consumption with renewable supply.

41.- Digital twins combine real-time production data, energy forecasts, and market signals to optimize across criteria.

42.- A portable demo factory shows rapid production re-planning based on incoming solar forecasts.

43.- Learning factories can help develop and deploy these AI solutions for energy optimization at scale through education and research.

44.- Key takeaway: Energy is both a challenge and opportunity that AI can help industry navigate for sustainability.

Knowledge Vault built byDavid Vivancos 2024