SimOps
Process modelling, simulation, and circular intelligence for bio-manufacturing and cyber-physical systems
synopsis
SimOps is a process modelling and circular intelligence platform for bio-manufacturing and cyber-physical systems. A maker designs a process as a cascade of stages — each with inputs, outputs, efficiency, sidestreams, sensors, and a bill of materials — then runs simulations, compares variations, and gradually binds model fields to live sensor readings as the process moves from design into operation.
The core insight is that a process that cannot see its losses cannot improve them. Sidestreams — marc, CO₂, spent SCOBY, waste heat, bacterial cellulose — that leave a process as waste are modelled from day one as capturable resources. The Sankey diagram makes the flow visible; the agent fleet makes it actionable.
Sensor observations are normalised via the W3C SOSA/SSN ontology, making synthetic and live readings structurally identical. The model runs on synthetic data until sensors arrive — synthetic observations are training data, not placeholders. When a field is bound to a live source, the model becomes an operational digital twin.
workbench
The v3 workbench is a single-page ABW app ( open workbench →) with three sections and a persistent companion:
| process | The canonical model — stages, KPIs, Sankey flow diagram, bill of materials. Each numeric field is tagged synthetic or live; declaring a SOSA contract opens the field to sensor binding. |
| variations | Forks of the base process with named hypotheses. KPIs are computed lazily per variation. Comparison across N variations produces a ranked narrative from the comparator agent. |
| simulations | Saved evaluation runs — evaluate (single process), compare (N variations), sweep (one field across a range). Each run is a citable artifact with a decision affordance. |
| companion | Always-on agent at the bottom of the page. Knows the active process, all variations, recent events, and budget. Emits structured action blocks — edits, forks, comparisons, annotations — rather than freeform text. |
agent fleet
| simops_companion | Primary strategist. Knows the whole fleet, active process, variations, events, and budget. Issues typed action blocks: edit_process, fork_variation, compare_variations, declare_sosa_contract, invoke_agent, annotate. |
| simops_cascade | Deterministic forward/backward KPI propagation: NER, carbon delta, OpEx per stage, mass balance. Available as a direct sub-second tool for real-time feedback in the UI. |
| simops_predictor | OLS regression on sosa_observation rows. Forecasts novel parameter combinations with R² and confidence intervals. Improves with each simulation run. |
| simops_optimizer | Recommends parameter adjustments against a target objective; compares the current operating point to the Pareto frontier. |
| sidestream_miner | Per stage: identifies capturable resource flows, estimates value, capture fraction, and regulatory class with external references. (Registration pending.) |
| supply_chain_oracle | Agent-resolved BoM pricing with risk-flag surfacing and procurement provenance. |
| comparator | Narrates winner and trade-offs across N simulation results — interprets regulatory, ecological, and operational dimensions not captured by the metrics. (Registration pending.) |
| bioreactor_modeler | Fits Monod-type kinetic models (μ_max, K_s, K_i) to session data; reports R² and RMSE; flags model drift between sessions. |
architecture
process.yaml ←──────────────────────────────────┐
│ │
▼ │
simops_cascade (deterministic, <1ms) │
│ │
├─▶ KPIs (NER, carbon, OpEx, LCC) │
├─▶ Sankey flow │ edit_process
└─▶ sosa_observation rows │ (diff → accept)
│
simops_companion ──────── action blocks ─────────┘
│
├─▶ fork_variation → variations/<slug>.yaml
├─▶ compare_variations → comparator narration
├─▶ invoke_agent → sidestream_miner · supply_chain_oracle · …
├─▶ declare_sosa_contract → field binding contract
└─▶ annotate → insights / decisions / risks
All workspace state is git-backed. Every simulation run, annotation, and sensor binding is a committed file. Synthetic observations and live sensor readings share the same sosa_observation schema — the transition from design to digital twin is a field-by-field binding, not a system swap.
reference deployments
| deployment | context | sensors | status |
|---|---|---|---|
| Chlorella PBR | Photobioreactor — algae biomass, NER, LCC optimisation | OD600, PAR, pH, temp | Synthetic · binding pending |
| Ambu SCOBY | Kombucha / SCOBY fermentation batch tracking | Brix, temp | Synthetic · MCU pending |
related research
tags
SimOps Workbench — v3
The v3 workbench is a live ABW app. Spawn a workspace, design a process, run simulations, compare variations, and gradually bind model fields to live sensor readings as hardware comes online. The companion agent orchestrates the full fleet from a persistent chat interface at the bottom of the page.