How to Find Better Solutions to Engineering Problems with AI
Find evidence-backed candidate solutions to engineering problems with AI — using Patsnap Solution Engine to turn a problem statement into TRIZ-backed candidate directions.
In one example run, the workflow returned nine candidate directions for a battery-electrode drying problem. The top-ranked one proposed staged temperature-gradient control across oven zones, linked to analogous cases in food drying and polymer-film curing — with a directional benefit estimate, not a guarantee.
Example run, August 2026 — one observed instance, not a performance benchmark.When you face a technical challenge — reducing energy consumption, improving performance under conflicting constraints, cutting cost without losing function — the range of possible solutions can feel overwhelming. Brainstorming tends to circle back to familiar approaches, and manual patent search can return a large document set that needs substantial expert review before useful analogies emerge. Claude can help you organize the problem and interpret candidate solutions, but a general Claude conversation doesn’t by itself give you controlled access to a structured patent-case library or a systematic method for applying it. Patsnap Solution Engine is a concept-generation service that uses TRIZ patterns and analogous patent cases to expand your range of solution directions — you don’t need any prior TRIZ knowledge; your job is just to describe the problem and constraints clearly, and the method structures the process in the background.
Which engineering problems are a good fit
- Reduce energy use without cutting throughput
- Reduce material use without losing structural strength
- Improve heat dissipation without increasing volume
- Lower defect rate without replacing equipment
- “I just want more creative ideas,” with no technical constraint
- No clearly stated technical function or trade-off
- You need directly validated engineering parameters
- You need an FTO, patentability, or legal conclusion
Use this template to check whether your problem is specific enough before you run it:
Example: battery-electrode drying
The problem: reduce energy consumption and drying defects in a battery-electrode drying process, while surface temperature stays controlled to prevent cracking, moisture removal stays complete, and throughput doesn’t drop.
Generate engineering concepts to reduce energy consumption and drying defects in a battery-electrode drying process. Surface temperature must stay controlled to prevent cracking, moisture removal must stay complete, and throughput must not drop.Describe the problem in plain language like this. Claude handles the workflow and returns the completed candidate set when it’s ready — in the recorded run, that took about seven minutes and returned nine candidates.
| Field | Top-ranked candidate |
|---|---|
| Concept | Staged temperature-gradient control across oven zones |
| Contradiction addressed | Drying speed vs. surface defect (thermal stress) |
| Directional benefit | Lower peak power demand, reduced thermal shock, improved multi-layer adhesion (from analogous cases — not a measured result for your process) |
| Main risk / assumption | Requires oven zone control upgrade; plant-specific energy and defect baseline not supplied or verified |
| Evidence source | Analogous patent cases in food drying and polymer-film curing |
| Suggested validation | Small-batch DOE with process instrumentation |
The recorded run returned nine candidates in total; full field-level detail was captured for this top-ranked one. The other eight followed the same structure — ask Claude for detail on any candidate from your own run using its solution ID.
How to run it
https://connect.patsnap.com/c18800/mcp?apikey=yourapikey Replace yourapikey with your own Patsnap Open Platform API key before connecting. Don’t paste the completed link into a public chat, a screenshot, or a code repository — anyone with it can use your credits. Keep the key and the URL private, and regenerate it if you think it’s been exposed.
- Get a Patsnap Open Platform API key.
- Copy the connection URL or JSON config from the Connect panel on the Solution Engine page.
- Add the server to Claude Desktop or another MCP-compatible client, then restart it.
Once connected, describe your problem to Claude as shown above — a full run takes about 5–10 minutes.
How to read and compare results
For each candidate, look at the same fields shown in the example above — concept, contradiction addressed, directional benefit, main risk, evidence source, suggested validation — and lay them against your fixed constraints (capital budget, equipment footprint, material compatibility, regulatory requirements). Eliminate anything that violates a hard constraint, then rank what’s left by benefit, implementation complexity, and technical risk. It helps to be clear about what’s actually doing the work at each stage:
- Expanding the concept search space
- Surfacing cross-domain analogies
- Linking candidates to TRIZ principles and cases
- Providing initial comparison dimensions
- Organizing your problem statement
- Explaining the returned results
- Building a comparison table
- Drafting a validation plan
- Confirming material and equipment constraints
- Judging physical feasibility
- Running simulation, DOE, pilot tests, safety review
- The final engineering decision
Moving to engineering validation
This is a concept-exploration step, not a replacement for patent search or engineering validation — here’s where it sits in a typical process:
For each candidate you take forward, define a test plan: which parameter you’ll vary, which metric you’ll measure, and what result confirms or rejects the hypothesis. Document the TRIZ reasoning and supporting cases alongside that plan for manufacturing, quality, and finance stakeholders. If validation confirms a concept, return to Solution Engine with refined constraints for optimization directions; if it doesn’t, restart from the problem statement with a different trade-off emphasis.
TRIZ and technical details (optional)
What TRIZ actually is. TRIZ (Theory of Inventive Problem Solving) comes from analyzing patterns across hundreds of thousands of patents, organized into a set of recurring inventive principles — instead of brainstorming from scratch, you match your contradiction to a principle that’s already solved something structurally similar in another field. Solution Engine automates that matching step against a patent-case library.
The server exposes six tools across two workflows. This tutorial uses the innovation set (run_triz_innovation_task, fetch_triz_innovation_task_stream, fetch_triz_innovation_solution_detail) — submit a problem statement, stream progress and final candidates, then retrieve full detail for one selected candidate. A separate cost-reduction set (run_triz_reduction_task, fetch_triz_reduction_task_stream, fetch_triz_reduction_solution_detail) follows the same shape but takes product structure, function, and cost information instead of a technical problem. The input is always a single user_input string, under 500 characters; the output is a list of candidates, each with a technical description, the TRIZ principle applied, a supporting patent-case reference, and a directional benefit indicator.
Explore engineering solution directions in Claude
Connect Patsnap Solution Engine and turn a problem statement into TRIZ-backed candidates.
Frequently asked questions
What kind of engineering problems work best with Solution Engine?
Problems involving a technical contradiction — where improving one parameter typically worsens another. Clearly stated constraints, materials, and process parameters help the engine match relevant cases and apply the right inventive principles.
Are the expected benefits guaranteed for my specific process?
No. Directional benefit indicators come from analogous patent cases in related domains, not from your process — treat them as reasons to prototype, not as performance guarantees.
Do I need a Patsnap subscription to use Solution Engine?
You need a Patsnap Open Platform account and API key. The Starter plan includes free credits for initial testing — check the pricing page for how far those go before committing to a paid plan.
Disclosure & disclaimer
Commercial disclosure and sources
Published by Patsnap, which develops and sells the Patsnap Solution Engine MCP server described here — an editorial tutorial, not an independent review. Server capabilities and connection details come from the product page and Patsnap Open Platform documentation as accessed August 2026; the example is one recorded run, not a repeated benchmark, and product capabilities change frequently.
Engineering limitations
Candidate concepts and benefit indicators are AI-assisted research leads drawn from analogous patent cases, not validated engineering results, a novelty search, or a legal opinion. Confirm feasibility with your own material samples, process instrumentation, and qualified engineering and IP review before implementation.
Trademarks, product names, and company names mentioned — including Claude and Cursor — belong to their respective owners and are used here for identification only.
