Skip to content

SIGMA Engineering GmbH (product: SIGMASOFT® Virtual Molding)

Company Overview

Business Overview

SIGMA Engineering GmbH was founded in 1998 in Aachen as a 100% subsidiary of MAGMA Gießereitechnologie GmbH (founded 1988), the market leader in metal-casting simulation (MAGMASOFT). Since 2017 SIGMA has been described as a sister company of MAGMA rather than a subsidiary; both sit under the family holding Dr. Flender Holding GmbH. [Evidence] 114

Key corporate facts:

  • Registered at the District Court of Aachen, HRB 7468; share capital €256,000; 11–50 employees (registry-derived data — the company publishes no headcount). [Evidence] 15
  • Thomas Klein has been Managing Director since January 1, 2017 (successor to Dr. Hartmann, who led the company for the prior decade); Timo Gebauer is CTO. [Evidence] 167
  • International presence: subsidiaries/offices established in Chicago (2006), Singapore and São Paulo (2009), Istanbul (2010), plus regional sites (Korea, China, India) visible as localized websites. [Evidence] 1
  • Group context: Dr. Flender Holding acquired Flow Science, Inc. (FLOW-3D CFD software, Santa Fe, USA) in December 2021; the group then totalled about 320 employees worldwide. [Evidence] 14

As a private GmbH, SIGMA publishes no revenue figures. [Evidence of absence]

Automotive Business

SIGMA does not report by industry, but automotive is a visible customer segment: its public case studies include a Ford Research & Advanced Engineering (Aachen) collaboration on crash simulation of plastic parts and an automotive suspension bushing troubleshooting study (elastomer with metal insert). [Evidence] 1011

Engineering Business

SIGMA is a one-product company: it develops, sells and supports SIGMASOFT Virtual Molding, plus engineering services around it (material characterization, project consulting, the "SIGMA Academy" training arm). The simulation core is shared lineage with MAGMA's casting technology — SIGMA states the underlying 3D finite-volume technology has been "developed, validated, and continuously optimized for over 30 years" in process simulation. [Evidence] 24

Role in Automotive Moulded Parts Workflow

Workflow Stage Relevant? Notes
Requirements No —
Industrial Design No —
Concept Design Indirect Early feasibility via simulation service work
CAD Modelling No Geometry is imported; SIGMA sells no CAD
Engineering Review Indirect Simulation results inform reviews
Simulation Core strength Process simulation (flow, cure, thermal, warpage, fiber orientation)
DFM Indirect Simulation exposes manufacturability problems (weld lines, air traps, sinks); no rule-checker product
Tool Design Strong (thermal/runner layout) Full-mould model: tempering concepts, hot/cold runners, insulation, heater placement
Mould Flow Core strength Including elastomer/LSR (liquid silicone rubber)/thermoset curing — beyond classic "mould flow"
Prototype Yes "Virtual moulding trials" replace physical sampling loops
Validation Yes Side-by-side virtual-vs-real demos (K 2025)
Manufacturing Engineering Partial Process set-point selection via virtual DoE; SIGMAecon costing
Production Release No —

What makes SIGMA structurally different: it is a deep specialist occupying only the simulation/mould-thermal slice of the chain — no CAD, no PLM, no manufacturing execution — and within that slice it models more of the physical reality (whole mould, many cycles, more material classes) than its bigger rivals. [Synthesis]

AI Strategy

Public AI Vision

None found — and that is the finding. SIGMA publishes no AI vision, no AI roadmap, and no AI-branded product. Its stated philosophy is "Less Guess – More Knowledge": understanding the real process through physics. The K 2025 press release describes SIGMASOFT as "an industry benchmark for digital reality in polymer processing" (CEO Thomas Klein) — the self-image is fidelity to physics, not intelligence. [Evidence] 25

Checked for AI language and found none: the Autonomous Optimization product page, the Virtual Molding deep-dive page, the company history, the K 2025 press release, and the K 2025 trade coverage. [Evidence of absence] 34157

Investor Statements

Not applicable — private, family-held company; no investor communications exist. [Evidence of absence]

Engineering AI Strategy

The closest thing to an "optimization strategy" is Autonomous Optimization (§5.1): specify goals, and the software "finds the optimal way to achieve them" by automatically computing a defined test series — virtual DoE. This is classical automated search, and SIGMA's own description matches that mechanism precisely. [Evidence] 3 Whether any ML work exists internally is unknown; nothing has been announced through the 6.x releases (2022–2026). [Evidence of absence] 15

Timeline of Product (not AI) Evolution

Date Event
1998 Company founded; first SIGMASOFT release (full 3D)1
2003 First true 3D fiber-orientation calculation1
2017 Thomas Klein becomes MD (Jan 1)16; v5.2 introduces Autonomous Optimization + virtual DoE1
Oct 2018 Fakuma: live LSR demo — 60-simulation autonomous optimization case12
2021 v5.3.1 adds compression moulding (elastomers)19
2022 v6.0: new user interface; "Virtual Thermoplastics" material-fingerprint service18
2024 v6.1 with SIGMAecon functionality (per company history)1
Sep 9, 2025 v6.2 announced for K 2025: up to 4× faster; SIGMAecon + SIGMA Rubber Designer premiere5
Oct 2025 K 2025: side-by-side virtual-vs-physical moulding demos with Momentive, Maplan, Engel, Nexus7
2026 v6.2 delivered (per company history)1

No entry in this timeline is a machine-learning feature. [Evidence of absence]

Products Relevant to Engineering

SIGMASOFT Virtual Molding — the entire product line is one simulation platform sold in material-specific configurations.

Purpose: simulate the injection-moulding process as it actually runs in production — the plastic and the mould. SIGMA's stated differentiator is that the model includes "plates, tempering systems, air gaps, and gating systems (hot/cold)" and computes inhomogeneous mould temperatures over multiple consecutive cycles, where competitors typically model the cavity with simplified boundary conditions. It claims to be "the only software that takes the complexity of injection molding seriously and maps it" — a vendor superlative with no third-party benchmark attached. [Evidence + Marketing] 4

Material breadth: thermoplastics, elastomers/rubber, silicone (LSR), thermosets, MIM/CIM, plus compression moulding (since 2021). For LSR, SIGMA calls itself "the market leader in the simulation of liquid silicone rubber applications" [Marketing] 8; independent academic groups do use SIGMASOFT in published LSR research, which at least confirms real adoption in that niche. [Evidence] 9

AI capabilities: none claimed by the vendor. The optimization layer is virtual DoE (§5.1). [Evidence] 3

Relevance to moulded parts: total — this is the whole company.

AI Capabilities

Strictly, this section should be empty: no SIGMA capability found in this research learns from data. The capabilities below are what a buyer might mistake for AI, described honestly.

1. Autonomous Optimization (virtual DoE). Description: the user defines goals and a set of variable parameters (gate positions and counts, filling times, temperatures, heater layouts…); the software then "automatically" computes "a defined test series" — one full physics simulation per candidate — and presents all results in a parallel-coordinate diagram with sliders for the user to locate the best trade-off. SIGMA's own page says it plainly: Autonomous Optimization "is based on a virtual DoE." [Evidence] 3 The honest test: does it learn from data or compute from equations? It computes from equations, many times over, and searches the results. It is deterministic, carries nothing from one job to the next, and improves nothing with use. This is exactly the automated-search-relabelled case analysed in the study's concept note — see What is Optimization technology. The name "Autonomous" invites an AI reading the mechanism does not support. [Evidence + Marketing] 322 Concrete scale (best documented case): at Fakuma 2018, SIGMA's live LSR demo ran 60 simulations (3 filling times × 20 gate positions) over about two days, selecting a two-gate configuration that gave a 55-second cycle with no flash, short shot or air entrapment, and cut cavity temperature spread from 40 °C to under 10 °C. Useful, real — and transparently brute force. [Evidence] 12 Cost framing: trade coverage quotes roughly $50 per virtual trial vs ~$1,000 for an 8-hour physical DoE trial. [Evidence — trade-press figure] 13

2. SIGMAecon (2024/2025). Computes part cost, energy use and CO₂ footprint for the simulated process and for the variants generated by Autonomous Optimization (e.g. hot vs cold runner, insulation concepts). Arithmetic on simulation outputs — no learning. [Evidence] 5

3. SIGMA Rubber Designer (2025). A generic elastomer database that approximates a compound's behaviour without physical measurement. This is a curated material database with interpolation, solving a real pain point (rubber compounds are rarely characterized) — not ML. [Evidence] 5

4. Virtual Thermoplastics (2022). A service that takes a "digital fingerprint" of a customer's polymer to identify unknown material properties and sharpen predictions. Measurement plus model fitting — classical parameter identification. [Evidence] 18

5. Integrative simulation hand-off (fiber orientation → crash). SIGMASOFT exports fiber-orientation tensors and weld-line data into MATFEM's MF-GenYld+CrachFEM material model for LS-DYNA/RADIOSS crash simulation (the Ford collaboration, §7). Physics-to-physics data mapping. [Evidence] 10

Engineering Workflow Contribution

SIGMA enters the automotive moulded-parts workflow after CAD and before (or instead of) physical mould trials:

part + full mould geometry imported → multi-cycle process simulation (fill/cure/thermal/warpage) → virtual DoE across gates, times, temperatures, heater layouts → SIGMAecon cost/energy comparison of variants → chosen set-up transferred to the real machine; optionally fiber-orientation data exported to structural/crash simulation.

The pitch is to replace iterative physical mould trials ("guess") with one virtual campaign ("knowledge"). At K 2025 SIGMA ran production cells on the show floor next to their simulations specifically to demonstrate that the predictions match reality (CTO Timo Gebauer: "The K show gives us the opportunity, together with our partners, to showcase how our results compare with reality."). [Evidence] 7

Everything in this loop is physics simulation plus automated search plus human judgment. No stage learns from accumulated data. [Synthesis]

Public Customer Evidence

Case Studies

  • Ford Research & Advanced Engineering (Aachen) + MATFEM — integrative simulation: SIGMASOFT's fiber-orientation tensors and weld-line data feed MATFEM's MF-GenYld+CrachFEM material model to make crash simulation of highly loaded plastic parts anisotropy-aware (LS-DYNA, RADIOSS). Strongest automotive name on SIGMA's books. Caveats: the case study is undated on the site and reports no quantified outcome ("improve component quality and reduce over dimensioning"). [Evidence — undated, unquantified] 10
  • Automotive suspension bushing (elastomer) — troubleshooting crack formation via insert-preheating analysis; customer unnamed. [Evidence] 11
  • Fakuma 2018 LSR showcase — the 60-simulation autonomous-optimization case (§5.1), run publicly with machine and material partners; the best-quantified SIGMASOFT result found. [Evidence] 12
  • Günther Heisskanaltechnik (hot-runner maker) — reported >2,500 heat and filling simulations with SIGMASOFT since adopting it after a 2010 benchmark. Caveat: source page could not be fetched directly; figures come from search extracts — treat as unconfirmed (Appendix C). [Evidence — thin] 20

Conference Demonstrations

  • K 2025 (Düsseldorf, Oct 8–15): SIGMASOFT 6.2 premiere; live virtual-vs-real comparisons with partners — a hard-soft frisbee (Momentive), a folding cup in a four-cavity mould (Maplan), needle-free connectors (Nexus), and a large fuel-cell seal moulded on an Engel cell — the latter the most automotive-relevant demo (fuel-cell stack sealing). [Evidence] 76
  • Fakuma appearances are a regular fixture (2018 LSR demo above). [Evidence] 12

White Papers / Testimonials

SIGMA maintains a press-release and technical-article archive; most items are application stories without named customers or hard numbers. No customer story combining SIGMA + automotive + AI exists — consistent with the vendor claiming no AI. [Evidence of absence]

Technical Architecture (Inferred)

Evidence

  • 3D finite-volume simulation core shared with/derived from MAGMA's casting technology, "developed, validated, and continuously optimized for over 30 years." [Evidence] 2
  • Full-mould, multi-cycle thermal modelling as the architectural signature. [Evidence] 4
  • v6.2 (2025/26): "innovation of the simulation engine, together with new computational approaches, delivers higher-quality results up to four times faster." No detail on what changed. [Evidence + Marketing] 5

Synthesis

The product is a classical HPC-style desktop/workstation solver with a virtual-DoE orchestration layer on top: mesh the part and the whole mould, run coupled thermal-flow-cure physics over several cycles until the mould reaches thermal steady state, repeat per DoE candidate, aggregate into the parallel-coordinate explorer. [Synthesis]

Inference

The casting heritage explains the architecture: in metal casting the mould's thermal history dominates, so MAGMA's technology always modelled the whole tool — SIGMA inherited that bias, which is why "the mould is in the model" comes naturally to SIGMASOFT and is bolted on later elsewhere. It also explains the strength in thermally-driven processes (elastomer/LSR/thermoset curing, MIM). No cloud/SaaS offering was found; the 4× speed-up in 6.2 suggests solver-level optimization, not a move to ML surrogates. [Inference]

AI Technologies

Inventory against the usual checklist: LLMs — none. RAG — none. Knowledge graphs — none. CAD automation — none (no CAD product). Generative design — none. Simulation AI / ML surrogates — none. AI agents — none. Vision AI — none. Digital twin — arguably yes in spirit: "Virtual Molding" is a physics twin of the production cell, and the K 2025 "digital reality" framing leans that way, but SIGMA does not market the term and there is no data feedback loop from real production into the model. [Evidence of absence + Synthesis] 54

Research Publications

Papers

SIGMA staff publish application-oriented technical articles in trade media (Plastics Technology, Kunststoffe) rather than peer-reviewed ML research. No SIGMA-authored machine-learning paper was found. [Evidence of absence] Independent academics publish peer-reviewed work using SIGMASOFT — e.g. LSR material-characterization and moulding studies — which validates the tool's standing in the LSR niche without involving AI. [Evidence] 9

Patents

A US patent titled "Method of simulating a shaping process" (US10520917) surfaced in searches alongside SIGMA material, but its assignee could not be verified in this research; a similar-sounding Austrian filing (AT516632A2) turned out to belong to Engel Austria, not SIGMA. No patent is attributed to SIGMA in this report. [Evidence — unresolved; see Appendix C]

Standards

No standards-body activity found. [Evidence of absence]

Open Source

Nothing found. No open-source projects, GitHub organisation, Hugging Face presence, published models or datasets surfaced anywhere in this research. For a 50-person proprietary-solver vendor this is unsurprising and consistent with the closed, licence-based business model. [Evidence of absence — from search coverage, not an exhaustive audit]

Engineering Service / Platform Mapping

SIGMASOFT is a standalone specialist tool: geometry comes in from any CAD system (CATIA, NX, Creo, SolidWorks) via neutral formats; results go out to structural/crash codes (LS-DYNA, RADIOSS via MATFEM's model) and, at trade shows, to live production cells (Engel, Maplan machines). There is no platform play, no PLM integration story, and no cloud service. In study terms: SIGMA plugs into everyone's workflow and owns none of it. [Synthesis] 107

Engineering Intelligence Stack Mapping

  1. Intent — minimal: the user's optimization goals entered into the virtual DoE ("specify your goals"). No requirements capture. [Evidence] 3
  2. Knowledge — strong but encoded, not learned: 30 years of process physics, material databases (mould steels, insulation, heater cartridges, the new generic elastomer database). [Evidence] 45
  3. Reasoning — brute-force search over simulations (virtual DoE) plus human reading of parallel-coordinate plots. Deterministic; no inference from data. [Evidence] 3
  4. Execution — none: SIGMA stops at the recommendation; humans transfer set-points to the machine. [Synthesis]
  5. Feedback — none in software: the K 2025 virtual-vs-real comparisons are marketing-grade validation events, not a closed data loop; nothing from production flows back to improve the models automatically. [Evidence of absence]7

Strengths

  • Deepest full-mould, multi-cycle thermal simulation in the segment — the architectural inheritance from casting simulation. [Evidence] 42
  • Material breadth beyond thermoplastics — elastomers, LSR, thermosets, MIM/CIM, compression moulding — where Moldflow/Moldex3D coverage is weakest; self-claimed LSR market leadership is at least consistent with independent academic use. [Evidence + Marketing] 89
  • Honest positioning — sells physics fidelity ("digital reality," "Less Guess – More Knowledge"), does not slap "AI" on its DoE. Rare and creditable. [Evidence] 53
  • Willingness to be tested publicly — live virtual-vs-real moulding at K 2025. [Evidence] 7
  • Stable family ownership (Dr. Flender group) with adjacent simulation businesses (MAGMA, Flow Science). [Evidence] 14

Weaknesses

  • No machine learning anywhere — no surrogate models, no learned defect prediction, no data products — at the exact moment a direct competitor (SIMCON/Cadmould) launched a public research preview of a trained neural solver (March 2026) claiming 1,000× speed-ups. Brute-force DoE at ~hours per candidate cannot match a seconds-per-variant surrogate for design-space exploration. [Evidence + Synthesis] 1712
  • Small company, one product — 11–50 people; R&D bandwidth for an ML pivot is structurally limited. [Evidence] 15
  • Narrow workflow slice — no CAD, DFM rule-checking, PLM, or manufacturing execution; always an import/export island in the customer's chain. [Synthesis]
  • Thin public evidence culture — undated, unquantified case studies; no named automotive customer with hard numbers except trade-show demos. [Evidence of absence]10
  • The name "Autonomous Optimization" invites misreading as AI even though the vendor's own text does not claim it. [Inference] 3

Current Gaps (largely manual today)

Choosing which parameters and ranges to include in the virtual DoE is expert manual work; reading the parallel-coordinate plots is manual; transferring set-points to the real machine is manual; and nothing learns from the company's accumulated simulation archive — every project starts from physics scratch. The mountain of "synthetic data" SIGMASOFT generates (every DoE run is a labelled simulation outcome) is exactly the training corpus a surrogate model would need, and today it is discarded as exhaust. [Synthesis + Inference]

Future Direction

Public roadmap signals are modest and physics-flavoured: v6.2's computational efficiency push, SIGMAecon's cost/CO₂ angle (sustainability reporting pressure), and the Rubber Designer's attack on the material-data bottleneck. [Evidence] 5

The strategic question is imposed from outside. SIMCON's Cadmould AI Solver (March 18, 2026 — transformer-based, trained on "hundreds of terabytes" of simulation data, up to 1,000× faster, positioned explicitly as explore with AI, verify with the classical solver) defines the ML-surrogate playbook for this segment. SIGMA's assets for a response are real — a trusted solver to generate training data, and the deepest thermal physics in the niche — but no public sign of such a move exists as of August 2026. Watch the 6.3/7.0 cycle and Fakuma 2026 messaging. [Evidence] 17 [Inference]

Relevance to Automotive Moulded Parts

Capability Strength Notes
Plastic Part Design Indirect Simulation feedback only; no design tools
Surface Design (Class A) None —
CAD Automation None No CAD product
Mould Flow / Process Simulation Core — differentiated Full mould, multi-cycle, thermoplastics + elastomer/LSR/thermoset/MIM; weld lines, air traps, sinks, warpage, fiber orientation
DFM Indirect Problems surface through simulation, not rule checks
Tool Design Strong (thermal) Tempering concepts, runner systems, heater layout, insulation — evaluated in the model
Manufacturing Engineering Partial Virtual DoE for set-points; SIGMAecon for cost/energy
Quality Partial Defect prediction from physics, not from data
Engineering Knowledge Reuse Weak No retrieval, no learning from past projects

Critical finding. SIGMASOFT's "Autonomous Optimization" — the feature most likely to be bought as "AI" — is virtual design-of-experiments: automated brute-force search over physics simulations. It is not machine learning, and no machine learning was found anywhere in the product through the 6.x releases (2022–2026). The vendor itself, to its credit, does not use the words "artificial intelligence" or "machine learning" in any material examined. Where third parties or buyers describe it as AI, the study's test applies: it computes from equations; it does not learn from data.22 [Evidence] 35

Net. For an automotive moulded-parts organisation, SIGMA is the specialist you bring in when the problem is thermally hard (LSR seals and grommets, thermoset e-motor components, rubber bushings, thick-walled optics, MIM) — and a calibration reference for this study: this is what the segment looks like without AI marketing. [Synthesis]

Key Takeaways

  1. SIGMA Engineering is a small (11–50 person), family-held Aachen specialist; sister company of MAGMA under Dr. Flender Holding. [Evidence] 1514
  2. SIGMASOFT Virtual Molding's differentiator is the whole mould in the model, over many cycles — an inheritance from casting simulation. [Evidence] 42
  3. Its material breadth (elastomer, LSR, thermoset, MIM/CIM) is where it beats Moldflow/Moldex3D, and where automotive electrification work (seals, e-motor thermosets) lands. [Evidence] 85
  4. "Autonomous Optimization" = virtual DoE = automated brute-force simulation search. Not machine learning. The documented flagship case is 60 simulations over two days.22 [Evidence] 312
  5. SIGMA itself never says "AI" in any material examined — the honest calibration point of this study. [Evidence of absence] 35
  6. No ML appeared in SIGMASOFT 6.0–6.2 (2022–2026). [Evidence of absence] 15
  7. Automotive evidence is real but thin: Ford crash-simulation collaboration (undated, unquantified), an elastomer bushing case, a fuel-cell seal demo at K 2025. [Evidence] 107
  8. Competitive alarm: SIMCON launched a public research preview of a trained neural solver (March 2026) in this exact segment; SIGMA has shown no public answer. [Evidence] 17
  9. Beware name collisions: sigmasoft.ai and sigmasoftai.com are unrelated companies. [Evidence] 21

References

Primary Sources — SIGMA Engineering

  • Company history — https://www.sigmasoft.de/en/about-us/history/
  • About us — https://www.sigmasoft.de/en/about-us/
  • Autonomous Optimization (virtual DoE) product page — https://www.sigmasoft.de/en/applications/autonomous-optimization/
  • Virtual Molding deep dive — https://www.sigmasoft.de/en/applications/sigmasoft-Deep_Dive_Virtual_Molding/
  • SIGMASOFT Elastomer/LSR — https://www.sigmasoft.de/en/applications/sigmasoft-lsr/
  • SIGMASOFT MIM and CIM — https://www.sigmasoft.de/en/applications/sigmasoft-mim-and-cim/
  • K 2025 press release (SIGMASOFT 6.2), Sep 9 2025 — https://www.sigmasoft.de/en/about-us/press/pressreleases/SIGMASOFT-at-K-Show-2025
  • K 2025 press release PDF — https://sigmasoft.com.sg/shared/.galleries/press-releases/K-show/2025/PR_SIGMA_ANNOUNCEMENT_K_2025_EN.pdf
  • Fakuma 2018 press release PDF ("From Virtual DoE to Hands-On Virtual Molding") — https://sigmasoft.de/shared/.galleries/press-releases/sigmasoft_press-release_fakuma-2018_en.pdf
  • Case studies index — https://www.sigmasoft.de/en/applications/case-studies/
  • Ford/MATFEM crash-simulation case study — https://www.sigmasoft.de/en/applications/case-studies/details/Injection-Molding-Simulation-improves-the-Results-of-Crash-Simulation/
  • Virtual Thermoplastics press release — https://www.sigmasoft.de/en/about-us/press/pressreleases/Virtual-Thermoplastics-Understand-the-injection-molding-process-in-detail-and-predict-it-more-precisely/
  • Virtual DoE part-properties press release — https://www.sigmasoft.de/en/press/pressreleases/Improved-Part-Properties-via-Virtual-DoE/

Corporate / Registry

  • North Data registry extract (HRB 7468, Aachen) — https://www.northdata.com/SIGMA+Engineering+GmbH,+Aachen/HRB+7468
  • Dr. Flender Holding acquires Flow Science, Dec 2021 — https://www.flow3d.com/dr-flender-holding-gmbh-acquires-flow-science-inc/ and https://www.prweb.com/releases/dr_flender_holding_gmbh_acquired_flow_science_inc_in_december_2021/prweb18470126.htm
  • Thomas Klein appointed MD, plasticker, Mar 16 2017 — https://plasticker.de/Kunststoff_News_29829_Sigma_Engineering_Thomas_Klein_ist_neuer_Geschaeftsfuehrer
  • LinkedIn — https://www.linkedin.com/company/sigma-engineering-gmbh

Trade Press

  • Plastics Technology: "Autonomous Optimization Proves Its Worth in LSR Molding" (Fakuma 2018 case) — https://www.ptonline.com/news/autonomous-optimization-proves-its-worth-in-lsr-molding
  • Plastics Technology: "Automate Injection Molding Simulation With Autonomous Optimization" — https://www.ptonline.com/articles/automate-injection-molding-simulation-with-autonomous-optimization
  • Plastics Technology supplier showroom (SIGMASOFT) — https://www.ptonline.com/suppliers/sigmasoft-virtual-molding
  • Plastech: "At K 2025, Sigma unveils Sigmasoft 6.2" — https://www.plastech.biz/en/news/At-K-2025-Sigma-unveils-Sigmasoft-6-2-and-new-features-21123
  • Plastech: "Sigmasoft to demonstrate simulation fidelity at K 2025" — https://www.plastech.biz/en/news/Sigmasoft-to-demonstrate-simulation-fidelity-at-K-2025-21265
  • CompositesWorld: compression moulding added to SIGMASOFT — https://www.compositesworld.com/products/sigma-engineering-adds-compression-molding-simulation-to-sigmasoft-virtual-molding
  • PolyForm NEXT (Günther Heisskanaltechnik + SIGMASOFT + KI, German; not directly fetchable) — https://www.polyformnext.de/heisskanal/produktentwicklung-wie-simulation-und-ki-den-spritzguss-effizienter-machen.htm

Competitive Context

  • SIMCON Cadmould AI Solver launch (Business Wire), Mar 18 2026 — https://www.businesswire.com/news/home/20260318680159/en/SIMCON-Unveils-Worlds-First-Large-Engineering-Model-for-Plastic-Injection-Moulding
  • Plastics Technology on Cadmould AI Solver — https://www.ptonline.com/products/simcon-introduces-injection-molding-simulation-tool-with-ai
  • SIMCON product page — https://www.simcon.ai/en-us/solutions/cadmould-ai-solver-injection-molding-simulation

Research Papers (third-party use of SIGMASOFT)

  • LSR material-characterisation study using SIGMASOFT (PMC) — https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12656677/
  • LSR fisheye-lens mould study (PMC) — https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10346709/

Name-collision (unrelated companies)

  • https://www.sigmasoft.ai/ and https://sigmasoftai.com/ — unrelated to SIGMA Engineering GmbH.

Appendix A — Timeline

1988 MAGMA founded → 1998 SIGMA founded (100% MAGMA subsidiary), first SIGMASOFT → 2003 3D fiber orientation → 2006–2010 offices (Chicago, Singapore, São Paulo, Istanbul) → Jan 2017 Thomas Klein MD; SIGMA becomes MAGMA's sister company; v5.2 Autonomous Optimization / virtual DoE → Oct 2018 Fakuma 60-simulation LSR demo → 2021 compression moulding (v5.3.1); Dec 2021 group buys Flow Science → 2022 v6.0 (new UI, Virtual Thermoplastics) → 2024 v6.1 (SIGMAecon) → Sep 2025 v6.2 announced (4× faster; SIGMAecon + Rubber Designer premiere at K 2025) → Oct 2025 K-show virtual-vs-real demos → 2026 v6.2 delivered. No AI/ML event exists to put on this timeline. (Context: Mar 2026, competitor SIMCON launches a public research preview of a trained neural solver.)

Appendix B — Glossary

  • Virtual Molding — SIGMA's brand for simulating the entire moulding process: plastic, full mould steel, cooling/heating channels and runners, over many consecutive cycles — not just the melt filling the cavity.
  • Virtual DoE (design of experiments) — running a structured series of simulations across combinations of candidate settings (gates, times, temperatures) instead of physical trials; a search method, not a learning one.
  • Autonomous Optimization — SIGMA's name for its virtual DoE layer: define goals, the software computes the test series and charts the results.
  • Parallel-coordinate diagram — a chart where each simulation run is a line crossing one vertical axis per parameter/result; sliders filter runs to find the best trade-off.
  • LSR (liquid silicone rubber) — a two-component silicone injected cold into a hot mould that cures (vulcanizes) it — thermally the reverse of thermoplastic moulding, which is why mould-thermal simulation matters so much.
  • Thermoset — a polymer that cures irreversibly in the hot mould (e.g. epoxy for e-motor components); cannot be re-melted.
  • MIM / CIM — metal / ceramic injection moulding: metal or ceramic powder in a polymer binder is moulded, then debound and sintered.
  • Fiber-orientation tensor — a mathematical description of which way short glass fibers point at each spot in a moulded part; controls local stiffness and is passed to crash simulation.
  • Surrogate model — a neural network trained on many simulations that then predicts outcomes near-instantly; what SIMCON built and SIGMA has not.
  • Hot/cold runner — the channel system delivering material to the cavity; heated for thermoplastics, cooled for reactive materials (elastomers/LSR).

Appendix C — Notes (thinnest-evidence areas to revisit)

  1. SIGMAecon's debut version: the company history credits it to v6.1 (2024), while the K 2025 press release calls it "brand-new" and premiering with 6.2 — likely introduced in 6.1 and expanded/showcased at K 2025, but unresolved.
  2. v6.2 dates: announced Sep 2025 for K 2025, but the history page lists delivery in 2026 — treat "announced 2025, shipped 2026" as the reading.
  3. Ford/MATFEM case study is undated — could be considerably older than it appears on the current website; do not present as recent without checking.
  4. The $50-per-virtual-trial vs $1,000-per-physical-trial figure comes from Plastics Technology summaries, not a fetched primary page.
  5. Günther Heisskanaltechnik figures (2,500+ simulations; interest in combining simulation's synthetic data with KI/AI) come from search extracts of a page that returned 403 — unconfirmed, and the AI musing appears to be the customer's, not SIGMA's.
  6. Employee count (11–50) is registry-aggregator data, not vendor-published.
  7. Patent US10520917 assignee unverified — do not attribute to SIGMA.
  8. "Market leader in LSR simulation" is a vendor self-claim with no independent market-share data.
  9. Negative findings ("SIGMA never says AI"; "no ML in 6.x") rest on the pages examined here — a broad but not exhaustive sweep; re-verify at Fakuma 2026 and any 6.3/7.0 announcement. Checked 2026-08-02: Fakuma 2026 (Oct 12–16, Friedrichshafen) — SIGMA is listed as exhibitor on its own events page and on fakuma-messe.de, but has published no preview, press release or AI positioning change yet; news page has no 2026 items; history page says v6.2 is being delivered to the global customer base in 2026 (the plausible Fakuma centerpiece). Re-check ~Sep 2026 when exhibitor previews publish.

Executive Summary

  • Who they are. SIGMA Engineering GmbH is a small (11–50 employees), privately held simulation specialist in Aachen, Germany — the sister company of MAGMA Gießereitechnologie GmbH, the casting-simulation leader. Both belong to the family holding Dr. Flender Holding GmbH, which also bought the CFD company Flow Science (FLOW-3D) in December 2021. The whole group employed about 320 people at that time. [Evidence] 15114
  • Why they matter for moulded parts. Their single product, SIGMASOFT Virtual Molding, simulates the injection-moulding process with the entire mould in the model — steel plates, cooling/heating channels, hot and cold runners, inserts — over many consecutive production cycles, not just the plastic in the cavity. It is unusually strong in elastomers, liquid silicone rubber (LSR), thermosets and metal/ceramic injection moulding (MIM/CIM) as well as thermoplastics — a genuine differentiator against Moldflow and Moldex3D. [Evidence] 48
  • The central honest finding. SIGMASOFT's flagship "Autonomous Optimization" (introduced 2017) is virtual design-of-experiments (DoE): the software automatically runs a large, user-defined series of physics simulations across candidate settings and visualizes the results so the best one can be picked. It is automated search by brute-force simulation — not machine learning. Nothing in it learns from data.22 [Evidence] 3
  • The surprising twist: SIGMA does not call it AI. Across every vendor page and press release examined for this report (product pages, the K 2025 release, the company history), the words "artificial intelligence" and "machine learning" never appear. SIGMA's slogan is "Less Guess – More Knowledge" and its pitch is physics fidelity, not AI. In a study full of AI-washing, SIGMA is the calibration point: a vendor whose honest story really is "physics + automated search, no ML" — and who mostly markets it that way. The one connotation risk is the word "Autonomous" itself. [Evidence] 352
  • The competitive clock is ticking. In March 2026, fellow German vendor SIMCON launched the Cadmould AI Solver — a transformer neural network trained on simulation data, claiming results up to 1,000× faster. Genuine ML has now arrived in SIGMA's exact market segment, and SIGMA has shown no public answer. [Evidence] 17

  1. SIGMA Engineering company history — https://www.sigmasoft.de/en/about-us/history/ ↩↩↩↩↩↩↩↩↩↩↩

  2. SIGMA Engineering "About us" — https://www.sigmasoft.de/en/about-us/ ↩↩↩↩↩↩

  3. Autonomous Optimization product page ("based on a virtual DoE"; no AI/ML language) — https://www.sigmasoft.de/en/applications/autonomous-optimization/ ↩↩↩↩↩↩↩↩↩↩↩↩↩↩

  4. SIGMASOFT Virtual Molding deep dive (full-mould, multi-cycle claims; "only software" superlative) — https://www.sigmasoft.de/en/applications/sigmasoft-Deep_Dive_Virtual_Molding/ ↩↩↩↩↩↩↩↩↩

  5. "SIGMASOFT at K-Show 2025" press release, Sep 9 2025 (v6.2, SIGMAecon, Rubber Designer; Klein "digital reality" quote; no AI/ML language) — https://www.sigmasoft.de/en/about-us/press/pressreleases/SIGMASOFT-at-K-Show-2025 ↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩

  6. Plastech, "At K 2025, Sigma unveils Sigmasoft 6.2 and new features" — https://www.plastech.biz/en/news/At-K-2025-Sigma-unveils-Sigmasoft-6-2-and-new-features-21123 ↩

  7. Plastech, "Sigmasoft to demonstrate simulation fidelity at K 2025" (Gebauer quote; Momentive/Maplan/Engel/Nexus demos incl. fuel-cell seal) — https://www.plastech.biz/en/news/Sigmasoft-to-demonstrate-simulation-fidelity-at-K-2025-21265 ↩↩↩↩↩↩↩↩↩

  8. SIGMASOFT Elastomer/LSR page (incl. "market leader" self-claim, [Marketing]) — https://www.sigmasoft.de/en/applications/sigmasoft-lsr/ ↩↩↩↩

  9. Third-party peer-reviewed LSR studies using SIGMASOFT — https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12656677/ and https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10346709/ ↩↩↩

  10. Case study: "Injection Molding Simulation improves the Results of Crash Simulation" (Ford Research & Advanced Engineering Aachen, MATFEM; undated, unquantified) — https://www.sigmasoft.de/en/applications/case-studies/details/Injection-Molding-Simulation-improves-the-Results-of-Crash-Simulation/ ↩↩↩↩↩↩

  11. SIGMA case-studies index (incl. automotive suspension-bushing elastomer study) — https://www.sigmasoft.de/en/applications/case-studies/ ↩↩

  12. Plastics Technology, "Autonomous Optimization Proves Its Worth in LSR Molding" (Fakuma 2018: 60 simulations = 3 filling times × 20 gate positions, ~2 days; 55-s cycle; 40 °C → ≤10 °C) — https://www.ptonline.com/news/autonomous-optimization-proves-its-worth-in-lsr-molding; SIGMA's Fakuma 2018 press release — https://sigmasoft.de/shared/.galleries/press-releases/sigmasoft_press-release_fakuma-2018_en.pdf ↩↩↩↩↩↩

  13. Plastics Technology SIGMASOFT supplier showroom ($50 virtual vs ~$1,000 physical trial; trade-press figure) — https://www.ptonline.com/suppliers/sigmasoft-virtual-molding ↩

  14. Dr. Flender Holding acquires Flow Science, Dec 2021 (group ~320 employees; owner of MAGMA and SIGMA) — https://www.flow3d.com/dr-flender-holding-gmbh-acquires-flow-science-inc/; PRWeb version — https://www.prweb.com/releases/dr_flender_holding_gmbh_acquired_flow_science_inc_in_december_2021/prweb18470126.htm ↩↩↩↩↩

  15. North Data registry extract, SIGMA Engineering GmbH, Aachen HRB 7468 (capital €256k; 11–50 employees) — https://www.northdata.com/SIGMA+Engineering+GmbH,+Aachen/HRB+7468 ↩↩↩↩

  16. plasticker, "Thomas Klein ist neuer Geschäftsführer," Mar 16 2017 — https://plasticker.de/Kunststoff_News_29829_Sigma_Engineering_Thomas_Klein_ist_neuer_Geschaeftsfuehrer ↩↩

  17. SIMCON Cadmould AI Solver launch (with Emmi AI; transformer "Large Engineering Model"; up to 1,000× faster; explore-with-AI, verify-with-solver), Business Wire, Mar 18 2026 — https://www.businesswire.com/news/home/20260318680159/en/SIMCON-Unveils-Worlds-First-Large-Engineering-Model-for-Plastic-Injection-Moulding; Plastics Technology coverage — https://www.ptonline.com/products/simcon-introduces-injection-molding-simulation-tool-with-ai ↩↩↩↩

  18. "Virtual Thermoplastics" press release (digital polymer fingerprint service) — https://www.sigmasoft.de/en/about-us/press/pressreleases/Virtual-Thermoplastics-Understand-the-injection-molding-process-in-detail-and-predict-it-more-precisely/ ↩↩

  19. CompositesWorld, SIGMA adds compression-moulding simulation (v5.3.1 era) — https://www.compositesworld.com/products/sigma-engineering-adds-compression-molding-simulation-to-sigmasoft-virtual-molding ↩

  20. PolyForm NEXT article on Günther Heisskanaltechnik, SIGMASOFT and KI (page returned 403; details from search extracts — see Appendix C) — https://www.polyformnext.de/heisskanal/produktentwicklung-wie-simulation-und-ki-den-spritzguss-effizienter-machen.htm ↩

  21. Unrelated namesakes: https://www.sigmasoft.ai/ and https://sigmasoftai.com/ — no connection to SIGMA Engineering GmbH, Aachen. ↩

  22. Why virtual DoE / "Autonomous Optimization" is classical automated search, not AI — see the concept note What is Optimization technology, which analyses this exact SIGMASOFT case: many deterministic physics simulations across a grid of candidate settings, best performer picked; nothing is learned from data. ↩↩↩↩