analyze-magnetic-levitation
정보
이 스킬은 패시브 자부상의 실현 가능성을 평가하고 필요한 안정화 메커니즘을 식별하기 위해 얼네쇼 정리를 적용하여 자기 부상 시스템을 분석합니다. 자기부상 열차, 베어링, 초전도 장치 등 다양한 시스템에 대한 힘의 균형과 안정성 계산을 수행합니다. 개발자는 공간 모드에 걸친 고급 안정성 분석과 마이스너 효과와 플럭스 고정 같은 현상들을 구분하기 위해 이 스킬을 사용해야 합니다.
빠른 설치
Claude Code
추천npx skills add pjt222/agent-almanac -a claude-code/plugin add https://github.com/pjt222/agent-almanacgit clone https://github.com/pjt222/agent-almanac.git ~/.claude/skills/analyze-magnetic-levitationClaude Code에서 이 명령을 복사하여 붙여넣어 스킬을 설치하세요
문서
Analyze Magnetic Levitation
Can system achieve stable lev? ID mechanism enabling/forbidding, calc force balance + stability, verify stable vs perturbations in all DOF including tilting.
Use When
- Eval proposed lev design physically viable
- Determine why perm magnet fails + ID workaround
- Analyze superconducting lev (Meissner, flux pinning, mixed-state trapping)
- Design/troubleshoot active EM feedback lev (maglev trains, magnetic bearings)
- Assess diamagnetic lev feasibility given material + field
- Understand spin-stabilized lev (Levitron) dynamics
In
- Required: Levitated object (mass, geometry, magnetic moment or susceptibility)
- Required: Field src (perm magnets, electromagnets, supercond coils, geometry)
- Optional: Op env (temp, vacuum, vibration constraints)
- Optional: Desired lev height/gap
- Optional: Stability reqs (stiffness, damping, bandwidth active)
Do
Step 1: Characterize System
Complete physical desc of object + field src pre-analysis:
- Object props: Mass m, geometry (sphere, disk, rod), magnetic moment mu (perm magnet objects), vol susceptibility chi_v (para/dia/ferromagnetic), conductivity sigma (eddy currents).
- Field src props: Config — perm magnet array (Halbach, dipole, quadrupole), electromagnet coil params (turns, current, core), or supercond coil (critical current, critical field).
- Field geometry: Spatial profile B(r). ID gradient dB/dz along lev axis + curvature d^2B/dz^2 governing stability.
- Env constraints: Temp range (cryogenic for supercond), atm (vacuum reduces damping), vibration spectrum.
## System Characterization
- **Object**: [mass, geometry, mu or chi_v, sigma]
- **Field source**: [type, configuration, key parameters]
- **Field profile**: [B(r) functional form or measured map]
- **Gradient**: [dB/dz at intended levitation point]
- **Environment**: [temperature, pressure, vibration]
→ Complete spec of object + field src → determine forces + stability no more assumptions.
If err: Susceptibility/moment unknown → measure or estimate material data tables. No quantity → force calc impossible. Composite objects → effective susceptibility vol-weighted avg.
Step 2: Apply Earnshaw
Passive static lev possible?
- State Earnshaw: Region free of currents + time-varying fields, no static arrangement of charges/perm magnets produces stable equilibrium pt for para/ferromagnetic body. Laplacian of magnetic potential energy nabla^2 U >= 0 (para/ferro) → U has no local min.
- Classify response: Object paramagnetic (chi_v > 0), diamagnetic (chi_v < 0), ferromagnetic (chi_v >> 0, nonlinear), supercond (perfect diamagnet, chi_v = -1), or perm magnet (fixed mu).
- Apply:
- Para/ferro/perm magnet in static field from perm magnets/fixed currents → Earnshaw forbids stable lev. ≥1 spatial direction unstable.
- Diamagnetic → Earnshaw does NOT forbid. nabla^2 U <= 0 allows local min. Passive static permitted.
- Supercond → Meissner = perfect diamagnetism, flux pinning → both lev + lateral stability.
- Doc verdict: Clearly state Earnshaw-forbidden or Earnshaw-permitted + which material prop determines.
## Earnshaw Analysis
- **Object magnetic classification**: [paramagnetic / diamagnetic / ferromagnetic / superconducting / permanent magnet]
- **Susceptibility**: chi_v = [value with units]
- **Earnshaw verdict**: [FORBIDDEN / PERMITTED]
- **Reasoning**: [which condition of the theorem applies or fails]
→ Definitive classification Earnshaw-forbidden or permitted + specific physical reasoning.
If err: Mixed magnetic character (ferromagnetic core + diamagnetic shell) → analyze each separately. Overall stability from net energy landscape → may need numerical field computation.
Step 3: ID Circumvention
If Earnshaw forbids passive static → ID which of 4 std circumventions:
-
Diamagnetic lev: Object itself diamagnetic (chi_v < 0). Examples: pyrolytic graphite over NdFeB, water drops + frogs in 16 T Bitter magnets. Reqs strong gradients: (chi_v / mu_0) * B * (dB/dz) >= rho * g, rho = density.
-
Supercond lev: Type-I or type-II supercond below T_c.
- Meissner lev: Complete flux expulsion → repulsive force. Stable but limited load + supercond must stay in Meissner state (B < B_c1).
- Flux pinning (type-II): Flux vortices pinned at defect sites. Provides both vertical lev force + lateral restoring. Supercond suspended below or above magnet. Locked in 3D position rel to field src.
-
Active EM feedback: Sensors measure position, controller adjusts electromagnet currents → equilibrium. Examples: EMS maglev (Transrapid), active magnetic bearings. Reqs power + sensors + control system bandwidth > mechanical resonance freq.
-
Spin-stabilized lev: Spinning perm magnet (Levitron) → gyroscopic stabilization of tilting mode Earnshaw makes unstable. Spin > critical freq omega_c → gyroscopic stiffness overcomes magnetic torque. Object must stay within narrow mass window.
## Circumvention Mechanism
- **Mechanism**: [diamagnetic / superconducting (Meissner or flux pinning) / active feedback / spin-stabilized]
- **Physical basis**: [why this mechanism evades Earnshaw's theorem]
- **Key requirements**: [material property, field strength, temperature, spin rate, or control bandwidth]
- **Limitations**: [load capacity, power consumption, cryogenics, mass window]
→ ID specific mechanism + physical basis clearly + quant reqs.
If err: Doesn't fit any → check hybrid (perm magnets primary force + eddy current damping stability, or diamagnetic stabilization of paramagnetic). Consider electrodynamic lev (moving conductors in field) → distinct via Lenz.
Step 4: Calc Lev Conditions
Force balance + quant conditions:
-
Vertical force balance: Magnetic force = gravity.
- Magnetic dipole in gradient: F_z = mu * (dB/dz) = m * g.
- Diamagnetic: F_z = (chi_v * V / mu_0) * B * (dB/dz) = m * g.
- Supercond (image method): Model as mirror + compute repulsion between magnet + image.
- Active feedback: F_z = k_coil * I(t), I(t) feedback-controlled.
-
Solve lev height: F_z(z) = m * g → equilibrium z_0. Analytic → solve algebraic. Measured/numerical → graphically or numerically.
-
Restoring force gradient (stiffness): k_z = -dF_z/dz at z_0. Stable → k_z > 0 (force decreases w/ increasing height). Vertical oscillation freq omega_z = sqrt(k_z / m).
-
Lateral stiffness: k_x = -dF_x/dx in horizontal. Earnshaw-permitted (diamagnetic, supercond) → should be positive. Feedback systems → depends on sensor-actuator geometry.
-
Load capacity: Max mass levitated → field gradient where equilibrium marginally stable (k_z → 0 at max displacement).
## Levitation Conditions
- **Force balance equation**: [F_z(z) = m*g, explicit form]
- **Equilibrium height**: z_0 = [value]
- **Vertical stiffness**: k_z = [value, units N/m]
- **Vertical natural frequency**: omega_z = [value, units rad/s]
- **Lateral stiffness**: k_x = k_y = [value, units N/m]
- **Maximum load**: m_max = [value, units kg]
→ Complete force balance + equilibrium pos + stiffness vertical + lateral + load capacity.
If err: No force balance solution (too weak for gravity) → can't levitate. Increase gradient (stronger magnets, closer spacing), reduce mass, or switch higher-susceptibility material. Neg stiffness any direction → unstable that direction → Step 3 for stabilization.
Step 5: Verify Stability All DOF
Stable all 6 rigid-body DOF (3 trans, 3 rot):
-
Translational: k_z > 0, k_x > 0, k_y > 0. Axially symmetric → k_x = k_y by sym. Compute restoring force small delta_x, delta_y, delta_z.
-
Tilting: Restoring torque small ang deflections theta_x, theta_y about horizontal axes. Magnetic dipole → torque depends on field curvature + moment of inertia. Tilting instability = primary failure passive perm magnet lev (spin stabilization Levitron addresses).
-
Spin (if applicable): Spin-stabilized → spin > critical omega > omega_c. Critical freq determined by magnetic torque / ang momentum ratio. Below omega_c → precession → tilting instability.
-
Dynamic: Active feedback → control loop phase margin (>30°) + gain margin (>6 dB) at all resonance freqs. Sensor noise no excite instability.
-
Thermal + external perturbations: Temp fluctuations (supercond near T_c), air currents (diamagnetic lev of light objects), mechanical vibration (field src mounting).
## Stability Analysis
| Degree of Freedom | Stiffness / Restoring | Stable? | Notes |
|-------------------|----------------------|---------|-------|
| Vertical (z) | k_z = [value] | [Yes/No] | [primary levitation axis] |
| Lateral (x) | k_x = [value] | [Yes/No] | |
| Lateral (y) | k_y = [value] | [Yes/No] | |
| Tilt (theta_x) | tau_x = [value] | [Yes/No] | [most common failure mode] |
| Tilt (theta_y) | tau_y = [value] | [Yes/No] | |
| Spin (theta_z) | [N/A or value] | [Yes/No] | [only relevant for spin-stabilized] |
→ All 6 DOF inherently stable or stabilized by ID'd mechanism (feedback, gyroscopic, flux pinning). System viable.
If err: Any DOF unstable + no stabilization → design not viable. Common fix: add active feedback unstable direction, diamagnetic material passive stabilization lateral mode, or increase spin → gyroscopic. Step 3 → incorporate additional mechanism.
Check
- Object props (mass, susceptibility/moment, geometry) fully specified
- Field src + profile characterized + gradients computed
- Earnshaw correctly applied to magnetic classification
- Circumvention mechanism ID'd + physical basis
- Force balance solved + equilibrium pos
- Stiffness computed all 3 translational
- Tilting stability analyzed both horizontal tilt axes
- Spin-stabilized → critical spin rate computed + verified
- Active → control bandwidth + stability margins checked
- Load capacity limits estimated
Traps
- Assume perm magnets levitate each other statically: Earnshaw forbids para/ferro, yet common misconception. Attraction/repulsion along 1 axis always → instability perp axis. Always apply theorem pre-force balance.
- Confuse Meissner w/ flux pinning: Meissner (type-I) pure repulsion, only works supercond below magnet. Flux pinning (type-II) locks supercond at fixed position rel to field → suspension any orientation. Physics + design implications fundamentally different.
- Ignore tilting modes: Many analyses check only translational + declare stable. Tilting instability = primary failure passive magnetic lev, needs separate analysis. Pos translational stiffness all directions while tilt-unstable possible.
- Underestimate diamagnetic field reqs: Diamagnetic susceptibilities very small (chi_v ~ -10^-5 most, -4.5 x 10^-4 pyrolytic graphite). Levitate mg objects → strong gradients, typically B * dB/dz > 1000 T^2/m for non-graphite.
- Neglect eddy current effects: Time-varying or moving conductors → eddy currents → forces + heating. Active feedback → eddy currents in object create phase lag → destabilize control loop.
- Treat supercond as perfect diamagnets all conditions: Type-II in mixed state (B_c1 < B < B_c2) partial flux penetration. Lev force depends on magnetization history (hysteresis), not just instantaneous field.
→
evaluate-levitation-mechanism— comparative analysis → select best approachanalyze-magnetic-field— detailed field profile computation, input to thisformulate-maxwell-equations— derive EM field eqsdesign-acoustic-levitation— alternative non-magnetic lev for compareformulate-quantum-problem— quantum treatment supercond lev (BCS, Ginzburg-Landau)
GitHub 저장소
연관 스킬
llamaguard
기타LlamaGuard는 폭력 및 혐오 발언 등 6가지 안전 범주에서 LLM 입력과 출력을 조정하기 위한 Meta의 70-80억 파라미터 모델입니다. 94-95% 정확도를 제공하며 vLLM, Hugging Face 또는 Amazon SageMaker를 사용해 배포할 수 있습니다. 이 기술을 사용하여 AI 애플리케이션에 콘텐츠 필터링 및 안전 가드레일을 손쉽게 통합하세요.
cost-optimization
기타이 Claude Skill은 리소스 적정화, 태깅 전략, 지출 분석을 통해 개발자들이 클라우드 비용을 최적화할 수 있도록 지원합니다. AWS, Azure, GCP에서 클라우드 비용을 절감하고 비용 거버넌스를 구현하기 위한 프레임워크를 제공합니다. 인프라 비용을 분석하거나, 리소스를 적정화하거나, 예산 제약을 충족해야 할 때 사용하세요.
quantizing-models-bitsandbytes
기타이 스킬은 bitsandbytes를 사용하여 LLM을 8비트 또는 4비트 정밀도로 양자화하며, 최소한의 정확도 손실로 50-75%의 메모리 감소를 달성합니다. 제한된 GPU 메모리에서 더 큰 모델을 실행하거나 추론을 가속화하는 데 이상적이며, INT8, NF4, FP4와 같은 형식을 지원합니다. 이 스킬은 HuggingFace Transformers와 통합되어 QLoRA 학습 및 8비트 옵티마이저를 가능하게 합니다.
dispatching-parallel-agents
기타이 Claude Skill은 3개 이상의 독립적인 문제를 동시에 조사하고 해결하기 위해 다중 에이전트를 배치합니다. 공유 상태나 의존성 없이 해결 가능한 무관련 장애 시나리오에 맞게 설계되었습니다. 핵심 기능은 병렬 문제 해결로, 각 독립 문제 영역마다 하나의 에이전트를 할당하여 효율성을 극대화합니다.
