Steel Mill Production and Order Tracking Software: What Happens to the Order When a Coil Downgrades at the Temper Mill?

Why a steel mill breaks every generic production system

Coil 44821 comes off the temper mill on a Tuesday afternoon with an edge defect over the last eight tonnes. The quality inspector downgrades it from an exposed automotive grade to commercial quality. That coil was applied to an order for a stamper whose delivery window closes Thursday, and the order line called for 60 tonnes of which this coil was 22. Somebody now has to answer four questions quickly. Is there another coil in the applied pool that meets the specification. If not, does a slab exist that can be rolled in time. Can the downgraded coil serve another open order within tolerance, or does it join the secondary inventory ageing in the yard. And what do we tell the customer.

In most mills those four questions are answered by three people on the phone, a query someone wrote against the level 2 database in 2011, and a spreadsheet called ORDERBOOK that a planner updates twice a shift. The mill has an ERP, usually SAP, excellent at purchasing and finance and blind to everything between the caster and the shipping bay. It has level 2 systems from Primetals, SMS or Danieli depending on who supplied each line, each with its own data model. It may have PSI Metals if the group invested in a proper metals planning suite.

The structural reason generic systems fail here is that steel is a divergent process with continuous re identification. One heat becomes several slabs, one slab becomes a coil, one coil becomes child coils and then cut lengths, and at every step the material can change grade, owner, destination and value. Standard manufacturing software models convergence and treats material as fungible against a part number. In your mill nothing is fungible.

Problem 1: order to material matching is continuous, not a one time allocation

An order is not filled once. It is applied, re applied, partially fulfilled, and re applied again as material is produced, tested, downgraded or damaged. The right question is never what material is allocated to this order, it is what is the best assignment of all available material to all open orders right now, given tolerances, delivery windows and the cost of substitution.

SAP will allocate a batch to a sales order. It will not tell you that coil 44821, now commercial quality, satisfies four other open lines within thickness and width tolerance, one of which is late, and that applying it there beats putting it in secondary at a discount. PSI Metals genuinely does this and does it well, which is why it is the reference in this market. The cost is a long implementation, licence weight and a model you configure rather than own.

What a custom build does: hold an applied inventory pool as a live object and run re application continuously, not nightly. Every piece carries its actual measured attributes, thickness, width, weight, chemistry, mechanical results, surface class, and every open order line carries its acceptance envelope including customer specific tolerances that are wider or narrower than the standard. When a downgrade event arrives from quality, the system proposes the re application within seconds and shows the value difference between options. That single loop is usually what pays for the project, because material that sits in secondary inventory for six months is capital you already spent.

Problem 2: genealogy has to survive every split and every downgrade

The chain is heat to ladle to cast sequence to slab to coil to child coils to cut lengths, and in a mill with pickling, cold rolling, annealing and coating it is longer still. Your customers ask for a mill test certificate under EN 10204 3.1 that ties actual chemistry and mechanical test results to the specific piece they received. Automotive and pressure vessel customers audit that chain and they are not gentle about it.

The place this breaks is the splits. A coil slit into four child coils inherits chemistry from one heat, which is easy. A coil rolled from a transition slab in a mixed grade cast sequence inherits from two, which is not. Level 2 systems track the piece on their own line, and the joins between lines are often reconstructed later from timestamps.

What a custom build does: model every material piece as a node with parents and children, with the transformation event recorded between them. Transition material is flagged explicitly rather than assumed. Test results attach to the piece they were taken from and propagate to descendants with the propagation rule visible. Certificate generation then reads the graph rather than a spreadsheet, and a recall query in either direction returns in seconds. Build this as an append only event log, because a genealogy record that can be silently edited is worth nothing in an audit.

Problem 3: level 2 systems speak four different dialects

If your caster came from one supplier, your hot strip mill from another and your galvanising line from a third, you have three data models, three timestamp conventions, three piece identifier schemes and three opinions about what a production event is. Nobody joins them, so plant wide questions such as yield from slab to shipped tonne by grade require a person with database access and a free afternoon.

Each equipment supplier offers a layer above its own line, which is rational for them and unhelpful for you, since your problem is the joins between lines.

What a custom build does: an acquisition layer per line that normalises into one plant wide piece and event model, running continuously rather than in nightly batches. This is where a large share of the effort in a steel project actually goes, and it is what makes everything else possible: once every line reports into one model, scheduling, genealogy, yield and costing read the same truth. Expect real time on identifier reconciliation, since the same coil carries a different number on three systems.

Problem 4: practice rules and campaign sequencing are yours alone

Which grades can follow which on the caster without a tundish change. How wide a width jump the hot strip mill accepts in a rolling campaign. Which coating line campaigns run weekly and which monthly. Your mill has hundreds of these rules and most exist as the knowledge of two or three schedulers.

Generic advanced planning tools model capacity and setup time. They do not model a cast sequence where a grade transition creates mixed chemistry material that must be applied somewhere or scrapped.

What a custom build does: encode the practice rules as data, so schedulers change them without a release, then schedule campaigns against them with the order book as demand. The value is not a perfect optimiser. It is that the rules stop living in three heads, and that when a scheduler retires the mill does not lose its sequencing logic. In our delivery experience this is the feature operations directors underestimate at kickoff and value most a year later.

Problem 5: nobody can say what a tonne actually costs

Yield loss happens at every step: scale, crop, side trim, edge defects, downgrades, secondary sales at a discount. Standard costing assigns a planned yield and moves on, so the mill knows its overall margin monthly and cannot say which grade, which width or which customer specification is destroying it. A build that computes actual yield per step from the piece graph, and values downgraded material at what it realised rather than at standard cost, turns margin by grade, width band and customer into a report. Mills routinely discover that a demanding automotive specification they chase for prestige is worse than the construction order they treat as filler.

What this costs and how long it takes

Across the 2,000 plus projects Digital Heroes has delivered, steel is at the heavier end. A first release runs $120,000 to $250,000 over 16 to 24 weeks, covering the plant wide piece and event model, level 2 acquisition for the lines that matter most, order to material matching with re application, and genealogy with certificate generation. The full platform, adding caster and rolling campaign scheduling, all remaining lines, yield costing, secondary inventory management and customer portal access, runs $350,000 to $900,000 phased over 12 to 24 months.

Cost drivers specific to a steel mill:

• The number of production lines and how many different level 2 suppliers are represented, since each acquisition integration is a separate piece of work.
• Whether piece identifiers are consistent across lines, because reconciliation of the same coil under three numbers is slow, manual discovery work.
• Certificate requirements, especially if you serve automotive or pressure equipment customers with their own formats and portals.
• Depth of scheduling, since a viewing and sequencing tool is a fraction of the cost of a genuine campaign optimiser and most mills should start with the former.
• Whether SAP remains the system of record for orders, which it usually should, making the integration bidirectional and therefore careful work.

What keeps cost down: start downstream. Order to material matching and genealogy from the hot mill onward deliver value first, and the caster scheduling work is easier once the piece model already exists.

Build versus buy, and when PSI Metals is the right answer

Buy PSI Metals if you are a large integrated producer with the budget and the appetite for a multi year programme, and if your practices are close enough to the industry standard model that configuration will reach them. It is a serious product built by people who know steel. Buy from your equipment supplier if you run a single line and want the tightest integration with it, accepting that you are then inside their ecosystem.

Build when two or more of these are true. You run lines from different equipment suppliers and nothing joins them. Your re application decisions are made by phone and your secondary inventory is growing. Your genealogy is reconstructed from timestamps rather than recorded as a graph. Your scheduling practice rules live in the heads of two people approaching retirement. You have looked at a full metals suite and the licence and implementation cost is out of proportion to a mill of your size.

The threshold is complexity rather than tonnage. A single line re roller working to stock can run on ERP with lot tracking. The moment you have divergent material, order applied production and more than a few hundred live orders, the coordination logic becomes the business and it deserves a system you control.

How to choose a developer for steel mill production software

Ask them to draw the material model on a whiteboard before you sign anything. Heat, cast sequence, slab, coil, child coil, cut length, each with parents, children and transformation events. If they draw a batch and a quantity, they have built warehouse software and are about to learn metallurgy at your expense.

Ask how they will handle transition material in a mixed grade cast sequence. It is a small question that reveals immediately whether anyone on the team has worked in a mill.

Ask what level 2 systems they have read from, by supplier and by line type. Reading a Primetals database is different from consuming an SMS X-Pact feed, and both are different from screen scraping an ageing HMI, which some mills genuinely need.

Ask how the mill test certificate is produced and whether the genealogy is append only. If a piece history can be edited without an audit trail, your certificate is not defensible, and one automotive audit will make that expensive.

Ask who owns the code and put it in the contract before kickoff. You should own the repository, the infrastructure and the right to hire another firm to continue. At Digital Heroes the code is yours from the first commit, and for a system that will run your mill for a decade, that is the most important commercial term in the agreement.