Numadelic LabsResearch. Adapt. Create.
Current stage: research and concept development. Numadelic has no commercially available products and is not taking orders. What this means

VIVO: adaptive manufacturing systems

A research programme on making customised products from workpieces that differ in shape, material and behaviour. VIVO is at the research and concept stage. No machine or production cell is claimed.

The industrial challenge

Most manufacturing processes assume that each workpiece has predictable geometry and material properties. Fixtures, tool paths and quality checks are designed around that assumption.

Natural stones and other heterogeneous materials do not behave that way. They differ in outline, dimensions, internal characteristics, defects and how they respond to a process. A customer-specific product adds another layer: the design itself changes from order to order.

Manufacturing from such workpieces can therefore involve a lot of measurement, planning and judgement for every single piece. VIVO investigates whether those steps can be connected in software so that the plan adapts to the workpiece instead of the workpiece being forced to fit the plan.

Concept visualization: a wireframe ellipsoid fitted inside the irregular dashed outline of a stone, showing how a simple model approximates a real, irregular workpiece.
Concept visualization A simple geometric model (solid) approximates an irregular stone (dashed). Real stones need richer models; this is the starting point.

Proposed VIVO architecture

Five stages in a loop. The last stage feeds what it learns back into the earlier ones.

  1. Sense

    Acquire information about the workpiece.

  2. Model

    Construct a computational representation of it.

  3. Decide

    Evaluate what can be made and how.

  4. Execute

    Translate the plan into machine operations.

  5. Adapt

    Use measurement and feedback to revise the plan.

1. Sense

Potential inputs include overall dimensions, surface geometry, imaging, material characteristics and processing conditions. Which sensors are worth using is a research question: each must be justified by what it lets the later stages decide.

2. Model

The proposed Digital Material Twin is a computational representation of the relevant geometry, constraints and material information for one workpiece. It records what is known and how uncertain it is.

3. Decide

Evaluate candidate products, orientations, geometries and manufacturing sequences against feasibility constraints. This is where customer requirements are connected to what this specific piece of material allows.

4. Execute

Translate the selected plan into operations that suitable equipment, tools and fixtures can carry out. Numadelic has not built a complete production cell; this stage is a design objective for prototyping.

5. Adapt

Use measurements and process feedback to reconsider the plan when the workpiece or the process turns out to differ from the model. An accurate record of what went wrong is as valuable as a success.

What is not claimed

These stages describe a proposed architecture. They do not describe a finished product, a tested machine or a validated result.

Research concepts and terms

These terms name ideas the research may explore. They are proposed concepts, not established, validated products.

Digital Material Twin (DMT)
A per-workpiece model of geometry, constraints and material information, with its uncertainties.
Unified Manufacturing Intelligence (UMI)
The proposed decision layer that links customer requirements, material-specific possibilities and process capability.
Unified Manufacturing Graph (UMG)
A proposed data structure linking workpieces, features, operations, tools and measurements so that plans can be searched and revised.
Material-aware planning
Choosing designs and operations that take the actual material of this piece into account.
Adaptive process control
Changing process parameters or the plan during manufacture in response to measurements.
Product feasibility assessment
Deciding early whether a requested product can be made from a given workpiece. The demonstration below is a simplified geometric example.
Geometric reasoning
Computing whether shapes fit, how thick walls are and how features relate.
Workpiece characterisation
Measuring the properties of a piece of material that matter for the product and the process.
Toolpath planning
Generating the tool movements that remove material to reach the planned shape.
Inspection and feedback
Checking the result and feeding the findings back into the model and the plan.

Interactive demonstration: can this bowl fit in this stone?

Adjust the stone and the bowl. The page computes, from your numbers, whether a washbasin bowl fits inside a simplified stone with the minimum wall thickness you require.

Conceptual Geometric Screening. Not a Validated Manufacturing Tool.
Calculating…
Plan view at the rim plane.
Side section through the bowl centre.
Mathematical assumptions and limits
  1. The stone is an ellipsoid. It occupies x²/a² + y²/b² + z²/c² ≤ 1 with semi-axes a = L/2, b = W/2, c = H/2, centred at the origin with z pointing up. Real stones are irregular, so this is a coarse stand-in.
  2. The top is cut flat on a rim plane z = zr. In automatic mode the page tries 40 heights and uses the highest one (least trimming) whose thinnest wall still meets the minimum you entered. If none does, it uses the height that gives the thickest wall.
  3. The bowl is a spherical cap with opening radius r = D/2 and depth d ≤ r. Its sphere has radius R = (r² + d²) / (2d). The opening is centred on the rim plane at the offset (x, y) you set.
  4. Wall thickness is the shortest distance from a point on the bowl surface to the stone’s remaining outer surface: the ellipsoid below the rim plane, or the edge where the rim plane meets it. It is found by solving for the nearest point on the ellipsoid numerically.
  5. Sampling. Thickness is evaluated on a grid of 10 rings by 36 directions over the bowl, plus the lowest point. A thinner spot between samples could be missed.
  6. Volumes use exact formulas for the ellipsoid, the removed cap and the spherical cap. The mass uses an assumed density of 2.65 g/cm³; real stones differ.
  7. Alternatives are found by trying nearby values in this same model. They are suggestions for exploring, not recommendations.

What this tool does not do. It does not detect defects or fractures, scan a real stone, simulate machining, estimate strength or stability, check drainage or plumbing, or show that a real stone is safe to manufacture. Passing here means only that the simplified shapes fit with the wall thickness you typed in.

What is needed to move VIVO forward

Turning this architecture into something that can be tested depends on evidence and partners that Numadelic does not yet have.

  • Material characterisation. Laboratory measurements on real river stones: variation in geometry and composition, internal flaws and how they show up in sensing.
  • Sensing trials. Testing which measurements are reliable enough to drive decisions.
  • Process development. Machining trials on real workpieces, with recorded outcomes.
  • Equipment integration. Access to suitable machines, tooling and fixtures for pilots.
  • Validation criteria. Agreed, measurable definitions of success for a first product.

If you can help with any of these, see Collaborate & Invest.

Partner in the development of adaptive manufacturing intelligence.

Numadelic is looking for investors, public research institutions, industrial pilot partners and academic collaborators who want to help move a carefully defined research programme toward validated prototypes.

Collaborate & Invest Contact Numadelic