One record for the whole plant.
Foundry Twin is a cloud platform that brings a foundry's records and measurements into one place, connects them to the process, and turns them into findings, reports and improvement work.
Why it's called a twin.
Each foundry is described as a set of equipment joined in the order material moves through it: furnaces, ladles, sand mixers, molding machines and lab instruments. Every reading attaches to the equipment and the quantity it describes. That model, kept in step with the plant by data, is the twin.
It's what lets the platform answer questions that cross departments, such as which sand conditions a defective batch was molded with.
How data gets in.
The first data is usually the records you already keep. Live connections come later, where they're worth having.
- Files you already keep
- Excel, CSV and machine database files. The first time a new kind of file arrives, a reader is prepared for it and a person checks it. After that, every file of that kind is read the same way, and an import can be undone completely.
- Entry on the floor
- Rejections, spot readings and shift registers on the web or in the phone apps, which keep working without a network. Paper registers can be photographed and read by text recognition.
- Machines and instruments
- A gateway on the plant network reads controllers over Modbus TCP and RTU, OPC UA, Siemens S7, EtherNet/IP or Mitsubishi MELSEC, and sends their values over an encrypted connection.
- The PC agent
- A small Windows program on the computer beside a machine. It uploads new files from folders the plant chooses, at a cadence the plant sets, and holds them safely until the network returns.
What each kind of data adds.
The platform is useful from the first day you send scrap records. Each Source Level keeps what the one before it provides and adds to it.
- 1. Scrap and rejection records
- Scrap rate with control limits, defect Pareto charts by part, line and shift, briefs and reports, targets and an action list.
- 2. Process and laboratory records
- Correlations between conditions and defects, tests of known defect mechanisms, process windows, control charts, capability, traceability and the melt shop balance.
- 3. Live equipment and instruments
- Alarms on live values, machine pages, dashboards, sharper links between process and outcome, and the power balance from energy meters.
- 4. Connected products
- The movement of metal, the structure of the metal and batch-by-batch sand results, joined to everything else.
How the twin's role grows.
Each stage answers a harder question than the one before, and needs more data and more time with your plant.
| Twin Stage | What the plant gets | Data it needs | When it starts |
|---|---|---|---|
| 1. Monitor | One place for scrap, process and machine data: dashboards, reports, Pareto charts, control charts, alarms, traceability, and the power and material balance. | Level 1 is enough to start. | From the first upload. |
| 2. Diagnose and prescribe | Correlations and mechanism tests, findings, recommended actions with an effect check, A3 and Six Sigma projects, and the Consultant. A person decides what to change. | Level 2. With Level 1 alone, only scrap and defects. | Once process records are joined to outcomes. |
| 3. Predict | A model of the plant that has learned its behavior: the risk of a defect before a pour, estimates of what isn't measured, and what-if tests of a change. | Levels 2 and 3. Level 4 makes predictions sharper. | Typically after a few months of data. |
| 4. Control | Closed-loop correction of chosen setpoints, inside limits the plant sets. It moves from advisory, to approved, to automatic within limits. | Level 3 for the points being controlled. | After predictions have proven reliable, and only by agreement. |
What your team works from.
- The Today brief
- A short written account of a shift, day, week, month or year: what was cast and scrapped, what changed most, and what moved with it. Every number must match the plant's data, or it isn't shown.
- Reports
- A statistical report for any period, heat by heat or by day, week or month: key figures, tables by shift and by grade, the worst heats, the defect Pareto, the strongest correlations, and capability where a specification exists. In PDF, Excel or CSV, with a printable week record for the quality meeting.
- Dashboards and alarms
- Boards of up to 12 charts for a control room or a meeting, and alarms on live values from machines and instruments.
- Pareto and scrap trends
- Defects ranked by count or cost, split by part, line and shift, with control limits that keep a small batch from looking like a large change.
- Correlations and mechanism tests
- Which conditions move most with each defect, and known mechanisms such as wet sand with gas porosity, each reported as supported, suggestive, not supported or not yet testable.
- Traceability
- From a heat and its chemistry to the ladles, molds and castings. A recall finds every casting from a suspect heat. A commonality search finds what a group of bad castings shared.
- Statistical process control
- Control charts, process windows and capability, the last calculated only where the plant has given a real specification.
- Power and material balance
- Every ton of metal, sand and binder, and every kilowatt-hour, from purchase to good castings, returns or losses, so you can see true yield and the true cost of a casting.
- The Knowledge page
- What the platform has learned about your plant, listed as confirmed findings or candidates, each with its evidence. Engineers acknowledge or dispute each one with a reason, so the plant's own reasoning is written down and kept.
- Actions and improvement projects
- Each change has an owner and a date, and its effect is checked from the day it's made. A3 and Six Sigma DMAIC projects start from any defect in the brief.
- The Consultant
- An assistant that answers questions about your data in plain language and prepares reports, alarm rules and projects as cards. Nothing is created until a person approves it.
The rules it keeps.
Every number traces back. A figure that can't be traced to its records isn't shown.
Associations, not claimed causes. A pattern in the data stays a lead until the plant confirms it.
A person approves what enters. New kinds of files are read only after someone has checked how they'll be read.
It sits beside your controls. Up to Twin Stage 3 it reads data and never writes a setpoint.
Your data stays yours. It can be exported in full or deleted on request.
See what it takes to start