Patent Protection For Desalination Membranes Optimised By Machine-Learning Models.
1. Conceptual Foundation
(A) What is being patented?
Modern desalination innovation combines:
- Physical product → membrane (e.g., nanocomposite, RO membrane)
- Process → method of desalination or fabrication
- Algorithmic layer → ML model optimizing:
- pore size
- permeability
- fouling resistance
- energy efficiency
ML is increasingly used for:
- predicting membrane performance
- optimizing structure–property relationships
- guiding material discovery
(B) Patentable Subject Matter
A typical ML-optimized desalination patent may claim:
- A membrane composition
- A method of designing membrane using ML
- A system combining ML + desalination unit
- A training method for predictive models
Example patents:
- Recent desalination membrane patents show structural innovations and chemical compositions
- Earlier patents include membrane systems and purification technologies
(C) Legal Requirements
To be patentable, the invention must satisfy:
- Novelty
- Inventive Step (Non-obviousness)
- Industrial Applicability
- Patentable Subject Matter
⚠️ Special issue:
ML models may be treated as:
- mathematical methods (non-patentable) OR
- technical applications (patentable)
2. Key Legal Challenges in ML-Optimized Membranes
(1) Algorithm vs Technical Effect
- Pure ML model → not patentable
- ML applied to improve membrane performance → patentable
(2) Data Dependency
- Training data may not be disclosed → sufficiency issues
(3) Inventive Step Problem
- If ML merely automates known optimization → may be obvious
(4) Enablement (Disclosure)
- Must explain:
- input parameters
- training method
- technical output
3. Detailed Case Laws (More than 5)
1. Diamond v. Diehr (1981, US Supreme Court)
Facts:
- Patent for rubber curing using mathematical formula (Arrhenius equation)
Issue:
Is a process using a mathematical algorithm patentable?
Judgment:
✔ YES — because it produced a technical industrial result
Principle:
“Application of algorithm in a physical process is patentable.”
Relevance:
- ML model optimizing desalination membrane =
algorithm + physical transformation (water purification)
👉 Therefore:
✔ Patentable if tied to technical improvement in membrane performance
2. Gottschalk v. Benson (1972, US Supreme Court)
Facts:
- Algorithm for binary conversion
Judgment:
❌ Not patentable
Principle:
- Pure mathematical algorithms are abstract ideas
Relevance:
If ML model is claimed as:
- “A neural network predicting permeability”
❌ Not patentable
BUT if:
- “A desalination membrane designed using ML to reduce fouling”
✔ Patentable
3. Alice Corp. v. CLS Bank (2014, US Supreme Court)
Facts:
- Computer-implemented financial method
Test Established:
Two-step test:
- Is it an abstract idea?
- Does it add “something more” (inventive concept)?
Relevance to ML membranes:
- ML optimization alone → abstract
- ML + improved desalination efficiency → inventive concept
4. KSR v. Teleflex (2007, US Supreme Court)
Facts:
- Mechanical invention combining known elements
Judgment:
❌ Not patentable (obvious)
Principle:
Combination of known elements with predictable results = obvious
Relevance:
If:
- ML is used to optimize known membrane parameters
- and results are predictable
❌ Patent may fail for obviousness
BUT if:
- ML discovers non-intuitive membrane structures
✔ Inventive step satisfied
5. EPO Case: T 641/00 (COMVIK Approach)
Principle:
- Only technical features contribute to inventive step
- Non-technical (algorithms) ignored unless producing technical effect
Application:
In ML desalination patents:
✔ Considered:
- membrane structure
- filtration efficiency
- energy reduction
❌ Ignored:
- mathematical model itself
6. EPO Case: T 1227/05 (Circuit Simulation Case)
Facts:
- Simulation method
Judgment:
✔ Patentable because it had technical purpose
Relevance:
ML model for:
- simulating membrane filtration
✔ Patentable if:
- linked to real-world engineering outcome
7. Indian Case: Ferid Allani v. Union of India (2019)
Facts:
- Computer-related invention rejected under Section 3(k)
Delhi High Court:
✔ Allowed patentability
Principle:
Computer programs with “technical effect” are patentable
Examples of technical effect:
- higher efficiency
- improved performance
- reduced resource consumption
Relevance:
ML-designed desalination membranes:
✔ Clearly produce technical effect (better filtration, less energy)
8. Indian Case: Telefonaktiebolaget LM Ericsson v. Intex (2015)
Principle:
- Technical contribution must be real and measurable
Relevance:
For desalination:
✔ Must show:
- improved salt rejection %
- reduced fouling
- energy savings
9. US Case: McRO v. Bandai Namco (2016)
Facts:
- Automated animation using rules
Judgment:
✔ Patentable
Reason:
- Not just automation; improved technical process
Relevance:
ML in membranes:
✔ Patentable if:
- it changes how membranes are designed, not just automates
4. Application to ML-Optimized Desalination Membranes
Patent Strategy
(A) Strong Claims
- Product Claim
- Membrane with specific ML-optimized nanostructure
- Process Claim
- Method of fabricating membrane using ML outputs
- System Claim
- Integrated ML + desalination plant
(B) Weak Claims (Likely Rejected)
- “An AI model for predicting membrane efficiency”
- “Algorithm for optimizing desalination”
(C) Drafting Tips
To survive legal scrutiny:
- Emphasize:
- technical improvement
- measurable parameters
- Include:
- training data characteristics
- feature selection
- physical outcomes
- Avoid:
- claiming algorithm alone
5. Emerging Legal Trends
- Patent offices increasingly accept:
- AI-assisted material discovery
- ML-guided engineering systems
- But require:
- clear technical linkage
- industrial application
- Desalination is a strong field due to:
- global water scarcity
- high industrial applicability
6. Conclusion
Patent protection for ML-optimized desalination membranes sits at the intersection of:
- Material science (membranes)
- AI/ML algorithms
- Patent law on software & technical effect
Core Legal Rule:
ML becomes patentable only when it produces a technical contribution to a physical system.
RELATED Blog
comments