Use Cases and Value Map

July 7, 2026 · View on GitHub

SCPN Phase Orchestrator is a synchronisation-analysis toolkit with a review-only control-proposal surface. It turns cyclic, event-driven, and state-machine behaviour into phase variables, runs those variables through coupled-oscillator dynamics, and produces reviewable control proposals for systems where synchrony is useful, harmful, or diagnostic — for human review, not automated actuation.

The practical question it answers is:

Which parts of this system are locking together, which parts should not be locking together, and which safe control knob can move the system back toward the intended regime?

The Short Explanation

Most complex systems contain repeating processes: heartbeats, queues, retry storms, grid frequency, machine vibration, market cycles, sleep stages, traffic lights, robotic gaits, plasma modes, or software-service heartbeats. SPO maps those processes into three oscillator channels:

ChannelWhat it capturesExample inputs
PhysicalContinuous measured wavesvoltage, pressure, vibration, EEG, PPG, temperature
InformationalTimed events and decisionsrequests, retries, trades, alarms, messages
SymbolicDiscrete states and modessleep stage, machine state, traffic phase, workflow state

Once every relevant signal has a phase, SPO can compute order parameters, phase-locking metrics, causal coupling evidence, regime transitions, and bounded control proposals with deterministic replay evidence.

Where It Creates Value

Read the Evidence column honestly: exactly one domain — the power grid — is externally validated (SPO's growth-rate estimate tracks ANDES small-signal eigenvalues, ρ up to 0.87). Every other row is a reusable scaffold: a working binding that parses, runs, and gives you a starting point, not a validated detector for that domain. Generic early-warning detection is at chance on real data outside the grid, and SPO reports that rather than overclaiming. Use the spo audit-detector auditor to check any detector before you trust it.

Market or domainOperational problemSPO value pathEvidence
Power and energyGrid oscillations, weak damping, inverter coordinationdetect harmful coherence, test coupling policies, produce audit-backed control proposalsValidated — growth rate vs ANDES eigenvalues (ρ≤0.87)
Fusion and plasmaMHD mode coupling, transport oscillations, control timingmap multi-rate observables into coupled phase channels and review stabilising proposalsScaffold — not yet validated
Cloud and platform operationsRetry storms, queue cascades, service-heartbeat lock-indetect synchronised failure modes and desynchronise overloaded layersScaffold — not yet validated
Manufacturing and machineryVibration, process drift, cyclic defects, SPC patternsexpose phase coupling between process variables and control loopsScaffold — not yet validated
Cardiology and neurosciencerhythm coherence, seizure precursors, band couplinganalyse phase locking, PAC/PLV/ITPC, and multi-scale synchrony without hidden actuationScaffold — review-only, no clinical claim
Robotics and swarmsgait coordination, leader-follower locking, swarm consensusprototype phase policies before physical deploymentScaffold — not yet validated
Traffic and logisticssignal timing, platoon waves, queue wavessimulate coordination policies and detect unstable synchronisation patternsScaffold — not yet validated
Finance and marketsregime coupling, sector synchrony, crash precursorsreview cyclic coherence and coupling shifts as risk evidence, not trading adviceScaffold — diagnostic, not prediction
Digital twinsplant/twin drift and residual coherencebind real telemetry to phase-state twins with replayable audit recordsScaffold — not yet validated
Research platformsdifferentiable oscillator models and inverse couplingoptimise coupling matrices and infer topology from trajectory dataToolkit — numerically validated vs analytic solutions

How to Recognise a Good SPO Use Case

A good candidate has all of these properties:

QuestionGood signPoor fit
Is there repeated behaviour?waves, cycles, retries, rotations, stages, heartbeats, events, or state loopsone-off static records with no temporal phase meaning
Can each source be timestamped or ordered?samples, events, and states can be aligned to a common clock or sequencemeasurements have no reliable ordering or cadence
Does synchrony matter?some coherence is desirable, harmful, diagnostic, or causalthe domain only needs scalar threshold alerting
Can assumptions be written down?coupling, lag, forcing, and safety boundaries are reviewablethe domain requires undocumented black-box decisions
Is actuation sensitive?proposals need replay, audit, rate limits, and human reviewa simple uncontrolled script is acceptable

Why the Market Angle Matters

Many organisations already collect cyclic telemetry but keep it split across signal dashboards, incident logs, notebooks, and controller-specific tools. SPO turns those fragments into a common phase contract. That matters because the same questions reappear across sectors: which components are phase-locked, what coupling path caused it, whether the lock is good or bad, and which bounded knob could move the system without violating review policy.

The commercial value is highest where wrong control is expensive: energy, industrial operations, clinical research, cyber-physical infrastructure, robotics, aerospace, and high-value simulation. In those settings the audit surface is as important as the mathematics. A proposal that cannot be replayed, rejected, or traced should not reach production.

Core User Journeys

UserFirst useful pathProduction path
Domain expertidentify oscillators and boundariesmaintain a domainpack with cited assumptions and validation evidence
ML researcheruse differentiable Kuramoto layers and SAF lossestrain coupling/topology proposals with accelerator checks and reproducible seeds
Platform engineerrun QueueWaves or a minimal domainpackwire metrics to bounded policies, audit replay, and dashboard review
Control engineerinspect K, alpha, zeta, and Psi knobskeep actuation behind rate limits, safety tiers, and human review gates
Documentation readerstart with this page and the quickstartfollow the tutorial sequence and API reference for the selected surface
Investor or product reviewerinspect use cases, evidence boundaries, and roadmapreview validated capabilities, commercial licensing path, and open frontier items

Expected value pattern

Across domains, value is most predictable when teams keep the same sequence:

  1. define what should synchronize and what should remain decoupled,
  2. represent inputs consistently across physical, informational, and symbolic channels,
  3. test bounded control candidates in replay-first mode,
  4. keep only candidates with explicit regime, metric, and evidence gates.

This produces a consistent operating language for operators, platform owners, and product reviewers because "good coherence", "harmful coherence", and "bounded response" are handled in the same contracts for different sectors.

What SPO Is Not

SPO is not a black-box automation system. Hardware writes are adapter-scoped, review-gated, and disabled for quantum or neuromorphic compiler targets until a verified external execution pipeline exists.

SPO is not a generic time-series dashboard. It specifically models phase, coherence, coupling, synchronisation, desynchronisation, and safe control proposal boundaries.

SPO is not a claim that every cyclic system is identical. Each domainpack must state its measurement source, phase extractor, coupling assumptions, safety boundaries, and validation evidence.

Capability Map

CapabilityWhat to readWhy it matters
Binding specsBinding APImakes domain assumptions explicit and reviewable
Oscillator extractionOscillators APImaps physical, informational, and symbolic data to phase
UPDE enginesUPDE APIruns Kuramoto, Stuart-Landau, delay, stochastic, and higher-order dynamics
SupervisorSupervisor APIclassifies regimes and proposes bounded actions
Audit replayAudit APImakes runs reproducible and tamper-evident
Differentiable dynamicsDifferentiable Kuramotosupports gradient-based topology and coupling optimisation
AutotuneAutotune APIextracts phases, identifies frequencies, infers coupling, and proposes replay-only policies
Hardware adaptersAdapter Bridgeskeeps optional integrations explicit and safety-bounded
NotebooksNotebooks and Demosgives runnable learning routes for each major capability

Learning Routes

GoalRoute
Understand the productThis page -> System Overview -> Pipeline Execution
Run the first local simulationInstallation -> Quickstart -> Hello World
Build a domainpackNew Domain Checklist -> Oscillator Hunt Sheet -> Build K_nm Templates
Move from data to a runFrom Raw Sources to Run -> Deterministic Replay
Work with ML surfacesDifferentiable Kuramoto Tutorial -> nn API -> SAF guide
Review production readinessProduction Guide -> Release Hygiene -> Roadmap

Evidence Boundary

The repository contains a broad capability surface, but each claim must be read with its evidence boundary:

  • committed tests demonstrate behavioural contracts, not universal domain truth;
  • benchmark pages report dated local or CI snapshots, not guaranteed throughput on every machine;
  • hardware adapters document safety boundaries and optional dependencies;
  • roadmap items distinguish implemented surfaces from research or deferred tracks;
  • domainpacks are examples until calibrated against real domain data and reviewed by the responsible domain expert.

That boundary is intentional. SPO is designed for systems where incorrect control claims are expensive, so documentation must separate executable capability, review-only evidence, and future research scope.

Adoption sequence by role

For each use case, apply this sequence before live recommendations:

  1. define the target coherence outcome (desirable lock, avoidable lock, or neutral synchrony),
  2. map signals into physical/informational/symbolic channels,
  3. run at least one audit-backed deterministic run,
  4. validate policy and monitor settings on replay,
  5. only then promote to review cadence with bounded actuation.

This keeps the evidence boundary explicit and prevents overreach from unvalidated inference.

Use this page to select a first domain

Before opening implementation guides, define:

  • target signal class (continuous, event, state),
  • expected control direction (synchronise, desynchronise, classify),
  • and the first evidence package (minimal domainpack + replay file).

If those three items are not agreed, the remaining process should stay in evaluation mode and avoid actuator exports.

Use this page as the final business-value filter before any advanced engine choice is made.

For the three focus tracks — industrial predictive maintenance, critical infrastructure, and biosignal/clinical — with their lead domainpacks, runnable commands, validation status, and safety-tier posture, see Beachhead Verticals.