v0.4 · RCPSP solver for batch processes

Run a scheduling what-if
in 30 seconds.

A modern MILP-backed scheduler for batch processes — built so you can sweep parameters in a meeting instead of waiting on SchedulePro for a quarter.

No install Bring a TSV Solve in < 60s Optimality reported
batchwise / solver — campaign_q3.tsv
solved · 0 conflictsmakespan 27.0 h
resource view3 batches · 5 resources · 16 phases0.0 h
00h
04h
08h
12h
16h
20h
24h
28h
32h
R-101
Reactor · 200 L
ChargeB-A
React 1B-A
QuenchB-A
FilterB-A
DryB-A
R-102
Reactor · 500 L
ChargeB-B
React 2B-B
QuenchB-B
FilterB-B
R-103
Reactor · 200 L
ChargeB-C
React 3B-C
FilterB-C
CIP
Clean-in-place
CIP R-101
CIP R-102
UTIL
Steam
Steam load
Steam load
prep react separation dry CIP utility
What it is

An RCPSP solver for batch manufacturing.

Recipes as unit procedures → operations → phases; renewable and non-renewable resources; sequence-dependent setups; utility envelopes; mass balance on intermediates. We solve as MILP with reported optimality gaps — not a heuristic dressed up as one.

Who it's for

Process engineers and MSAT teams.

Process engineers, manufacturing-science engineers, and MSAT groups doing capacity studies, debottlenecking, and tech-transfer modelling. Designed to be complementary to SuperPro and Aspen — not a process simulator replacement.

How it works

Three steps from a structured input to an optimal schedule.

The data shape is the same one you already keep in your simulation models. We don't ask you to re-author your recipe.

STEP / 01

Define procedures, batches, equipment

Structured input — TSV, JSON, or schema-bound table. Procedures decompose into operations and phases; resources are renewable (equipment, operators) or non-renewable (utilities, materials).

batch procedure cycles eq_pool B-AALK-013R-101 B-BALK-022R-102
STEP / 02

Set objectives and constraint envelope

Makespan, throughput, or weighted multi-objective. Sequence-dependent setups, CIP windows, hold times, mass balance on intermediates, utility limits — first-class, not bolted on.

objectivemin makespan
seq-dep setup (R-101)2.5 h CIP
utility envelopesteam ≤ 1.2 t/h
STEP / 03

Solve, inspect, sweep

Solve time, lower bound, and optimality gap reported. Infeasibility cores when constraints conflict. Re-run from a saved input file — reproducible, versionable.

solver / cloud-run
● solved 4.7s
makespan: 27.0 h
lower bound: 26.8 h
optimality gap: 0.74%
Capabilities

Four solver behaviors that change how you work.

  1. 01

    Conflict-free by construction

    Every schedule Batchwise generates is guaranteed conflict-free. If your constraints can't be satisfied, it tells you exactly which one is binding — never an invalid Gantt with overlapping equipment or broken precedence.

  2. 02

    Target-driven, not trial-and-error

  3. 03

    Inventory-aware scheduling

  4. 04

    Fast enough to iterate live

step 1 of 4·0% scrolled
batchwise · conflict-free by construction
solving
R-101R-102R-103R-1040h6h12h18h24hB-AB-BB-CB-D4 conflicts
R-101R-102R-103R-1040h6h12h18h24hP1P2P1·²P2·²4 reactors used
100%safety0d0d4d7d11d14stockoutreorder · 3 days earlynaivebatchwise
solver / cloud-run · 3 in parallelwhat-if A · +1 reactorsolving… 0%~ 40.0hwhat-if B · CIP 2hsolving… 0%~ 40.0hwhat-if C · split campaignsolving… 0%~ 40.0helapsed: 0.0s
Comparison

What we expose that the alternatives don't.

SchedulePro is paired with a process simulator and pays for it in iteration speed. Excel is honest about being a spreadsheet. Batchwise is the missing third option: a real solver, surfaced as a tool you can iterate in.

Capability
Batchwise
Spreadsheets
Intelligen SchedulePro
Target-driven optimization
✓ Yes
— No
◐ Partial
Conflict-free by construction
✓ Yes
— No
◐ Partial
Sequence-dependent setups
✓ Yes
◐ Partial
✓ Yes
Non-renewable resources / utilities
✓ Yes
◐ Partial
◐ Partial
Optimality gap reported
✓ Yes
— No
— No
Infeasibility cores / constraint introspection
✓ Yes
— No
— No
Reproducible from input file
✓ Yes
◐ Partial
◐ Partial
Re-solves in seconds for parameter sweeps
✓ Yes
— No
— No
Runs in browser, no install / dongle
✓ Yes
✓ Yes
— No
Methodology & trust

Reproducible inputs, reported optimality, transparent benchmarks.

Every solve is reproducible from a saved input file. We report lower bounds and optimality gaps rather than just a feasible schedule. Benchmark methodology is published with each performance claim — if you want to replicate, ask and we'll send the instance spec. Recipe data is encrypted in transit and at rest, never used for model training.

Formulation MILP · CP fallback
Reporting LB · UB · gap · core
Reproducibility Saved input file
Benchmarks Methodology on request
Data Encrypted; never trained on
FAQ

The questions process engineers ask first.

A MILP formulation as the primary engine, with a CP fallback for instances where MILP struggles. Commercial backend (Gurobi-class) under the hood; the formulation, not the backend, is the interesting part.

Optimal within a reported gap. We expose the lower bound, upper bound, and optimality gap on every solve. If you set a time budget that doesn't close the gap, we tell you exactly how far we are.

First-class. Sequence-dependent setups (e.g. CIP that depends on the previous batch's product), utility envelopes, and consumable inventory are part of the base formulation — not an afterthought.

Yes for batch splitting across compatible equipment in the same pool. Blending of intermediates is supported through mass-balance constraints; ask us about your specific case.

We track the standard PSPLib-style benchmarks internally and can share results on request. Most academic instances don't model the real-world constraints (CIP, utilities, mass balance) that make pharma scheduling hard — so we benchmark on industrial instances too.

An API is available for design-partner customers. The same input file you'd paste into the UI works against the API. Reproducibility is a first-class goal.

Get started

Paste your data. See the solve. Then decide.

Try the solver with your own TSV — feasibility, makespan, optimality gap, all reported. If you want to dig in on a specific debottlenecking question, book a call.