Wear Parts & Mining Equipment
Wear parts: making drawings, process and failure history answerable
A crusher wear part — jaw plate, concave, mantle, roll shell, tooth plate — compresses four kinds of knowledge into a single casting: the drawing, the material grade, the heat treatment and the inspection record. In most plants that knowledge lives in 2D drawings, process cards, QA sheets and a few veterans’ heads, spread across PDFs, scans, spreadsheets and chat threads.
RatioMax ingests all of it into a knowledge base that stays inside the plant. Drawings and scans are parsed automatically; parts, material grades, heat-treatment steps and inspection criteria are linked into a knowledge graph. Engineers ask in plain language and get answers cited back to the exact drawing sheet, process clause or inspection record.

Where it breaks
Scanned drawings you cannot search
Many older drawings exist only as JPGs or scanned PDFs, named with a part number and a date. There is nothing to full-text search. Finding a drawing depends on memory and folder archaeology, and even then someone has to check by eye whether it is current.
Several versions of one part in circulation
After a revision, engineering, the shop floor, subcontractors and field service may each hold a different copy. Nobody can confirm who is on the old one until an assembly clash or a customer complaint surfaces — by which point the rework cost is already spent.
Process know-how lives with a few veterans
Soaking time for water toughening, pouring temperature, whether a defect is repairable — most of it is passed on verbally. It never becomes a searchable, comparable record, so when the veteran retires the knowledge goes with them.
Slow answers to technical enquiries
When a customer asks whether a grade suits their rock, which hardness standard applies, or how inspection is done, engineering has to cross-check drawings, process cards and past work orders. Answers drift: the same question asked two months later may get a different reply.
How RatioMax responds
Scanned drawings become searchable
Complex-layout PDFs go through OpenDataLoader; images and scans through PaddleOCR-VL. Title blocks, dimension callouts and technical notes on 2D drawings all land in the search index without manual data entry.
Platform capability: Dual-engine parsing
Version snapshots show what changed
Chunk-level snapshots and diff views highlight exactly what changed, with one-click rollback. Metadata filters on part number, material grade or customer keep answers scoped to the part actually being asked about.
Platform capability: Chunk-level versioning
Drawings, materials and process linked
Wiki mode has the agent build interlinked pages and a knowledge graph from the source documents: a jaw plate links to Hadfield steel, the steel links to water toughening, toughening links to the hardness and NDT acceptance criteria — not just a paragraph of summary.
Platform capability: Auto Wiki and knowledge graph
Every answer cites its source
Hybrid retrieval plus reranking produces the answer, and every conclusion carries a citation. Open it and you land on that drawing sheet, that process clause, that inspection row — so engineering can verify before replying.
Platform capability: Traceable citations
Deployment scenarios
Customer technical enquiries
A customer asks whether a grade suits granite crushing or which hardness standard applies. The engineer asks directly and gets an answer backed by past work orders and drawings.
Process preparation before a changeover
Before a changeover, pull that part’s heat-treatment requirements, inspection items and past nonconformities in one pass, cutting trial runs and first-article rework.
Failure and nonconformity investigation
When a part cracks or wears out early, trace the material batch, heat-treatment record and inspection data by part number — grounding the failure analysis in records instead of impressions.
Onboarding engineers and service staff
New process engineers and field service staff ask the questions they actually have and get cited answers, shortening the path from onboarding to answering customer questions independently.
Terms and knowledge entities
- Hadfield manganese steel
- An austenitic wear-resistant steel with manganese as the main alloying element, typically around 11–14% Mn. It work-hardens under impact while the core stays tough, which is why it is used for jaw plates and concaves.
- Water toughening
- The heat treatment for Hadfield steel: soak at roughly 1050–1100°C and quench in water to obtain a single-phase austenitic structure, suppressing grain-boundary carbides that would otherwise cause brittle failure in service.
- Jaw plate
- The liner on the fixed and moving jaws of a jaw crusher. It takes the compression, splitting and sliding wear directly and is one of the highest-consumption wear parts.
- Concave and mantle
- The fixed bowl liner (concave) and the moving cone liner (mantle) of a cone crusher. Together they form the crushing chamber and are normally assessed as a pair for replacement.
- Casting tolerance grade
- Under GB/T 6414, casting dimensional tolerances are expressed as CT grades — a higher CT number means a wider tolerance. It is a separate system from machining tolerance, and drawings normally specify both independently.
- Surface roughness Ra
- Arithmetic mean deviation of the assessed profile, in micrometres. Drawings use it to state the required finish on a machined surface, which in turn constrains the machining process.
- Work hardening
- The rise in strength and hardness a metal undergoes during plastic deformation. Hadfield steel’s wear resistance depends on work hardening under impact rather than on its initial hardness.
- Brinell hardness HBW
- Hardness measured with a tungsten-carbide ball indenter, commonly used on castings and after heat treatment. Test location, force and conversion method must match the drawing or technical agreement or the values are not comparable.
- NDT: UT and MT
- Ultrasonic testing (UT) finds internal defects; magnetic particle testing (MT) finds surface and near-surface ones. Both are routine acceptance items for steel castings and the results must be archived with the part for traceability.
These definitions are also exposed structurally so search engines and AI engines can read and cite the domain concepts accurately.
Related deployment
Chongqing Boyue Machinery Manufacturing Co., Ltd.
Wear parts for mining crushers
Makes wear-resistant castings — jaw plates, concaves, mantles, roll shells and tooth plates — for crusher OEMs and mining customers.
Solution: An engineering-drawing knowledge base on RatioMax with multi-format parsing and smart chunking; Wiki mode auto-generates structured pages linking drawings, parts, materials and processes.
- 20 Drawings & tech docs ingested
- 107 Knowledge-graph nodes
- Seconds Natural-language answers vs manual lookup
- Standard Quality evidence & issue handling
Results shown are project targets
Industry questions
- What hardness does a water-toughened Hadfield jaw plate have, and why is low hardness still wear resistant?
- After water toughening the steel is austenitic and its initial hardness is modest — commonly in the HBW 180–220 range. Wear resistance comes from work hardening under impact in service, where the surface hardens while the core stays tough. Acceptance therefore cannot rest on initial hardness alone: microstructure, impact toughness and the actual duty matter too. RatioMax captures these acceptance conventions with their source documents so new staff get a cited explanation rather than a guess.
- What does the CT casting tolerance grade on a drawing mean, and how does it differ from machining tolerance?
- CT grades follow GB/T 6414 and describe the permissible deviation of the as-cast surface; a higher CT number means a wider tolerance. Machining tolerances apply to the finished, mated dimension. They are independent systems and drawings normally state both. Confusing them leads straight to wrong stock allowance or wrong assembly clearance. RatioMax files these conventions per part and returns them alongside the original drawing so the interpretation does not drift between people.
- A customer asks whether a given grade can crush granite. What do you answer from?
- The defensible route is to pull comparable duty history: rock hardness and abrasiveness, feed size, chamber profile, achieved service life and failure mode, then combine that with the material grade and heat-treatment state. RatioMax keeps past work orders, drawing specifications and failure records in one knowledge base, so the engineer can surface similar duty cases with citations and defend the recommendation later.
- How do you trace chemistry and heat-treatment records across batches of the same grade?
- Traceability depends on binding heat number, part number, heat-treatment batch and inspection result into one queryable chain instead of leaving them across paper furnace logs and spreadsheets. RatioMax supports custom metadata and metadata-filtered retrieval by part, heat or batch, and both answers and wiki pages cite back to the original record — so a nonconformity can be traced to a specific batch rather than guessed at.
- Most wear-part drawings are scans. Can an AI actually read them?
- What can be read is the legible content: title blocks, part numbers, material grades, dimension callouts, technical notes and tables. RatioMax uses PaddleOCR-VL for scans and images and OpenDataLoader for complex-layout PDFs, then chunks the result semantically. To be clear, this is document understanding, not 3D geometry reconstruction — spatial relationships in complex views should still be confirmed against the original sheet.
- Can formulation and heat-treatment parameters go on a public cloud?
- For most wear-part manufacturers the answer is no: process parameters and customer drawings are core IP. RatioMax supports full on-premise deployment on your own servers or intranet, so data never leaves your boundary. It ships as a Docker Compose single-node deployment or a Kubernetes cluster, the four core services deploy and scale independently, and each knowledge base can bind to a private OSS or S3 bucket.
- How do you retain expertise when a veteran retires?
- Archiving chat logs does not work. Experience has to become retrievable, citable and checkable. A practical path is to let Wiki mode generate interlinked knowledge pages from drawings, process cards, inspection records and nonconformance reports, then have process engineers revise them. Pages carry version snapshots and diffs, so who changed what stays visible — the knowledge grows on the document system instead of living in one person’s memory.
- How long before a new process engineer can answer customer questions independently?
- It depends on two things: whether the knowledge is searchable and whether the answer can be verified. RatioMax does not promise a specific number of days, but it compresses the two slowest steps — locating the material and confirming the accepted interpretation — because a question returns a cited answer instead of requiring weeks of folder archaeology. Whether someone is ready to answer alone still depends on judgement and mentoring; the tool only makes verification faster.
Other industries
- Equipment Manufacturing & Drawings — Drawing versions, process and inspection knowledge
- Modified Engineering Plastics — Engineering Plastics R&D and Quality Knowledge Base
- Government & Public Sector — AI Knowledge Base for Public Services
- Financial Services — AI Knowledge Base for Financial Services Teams