Multi-Dilution Conductivity Analysis Platform for Determining Ion Concentrations in Water

Background

The analysis of ionic compositions in aqueous solutions underpins critical applications across environmental monitoring, water treatment, and industrial process control. Accurate, real‐time assessment of key ions such as Na⁺, K⁺, Ca²⁺, Mg²⁺, Cl⁻, and SO₄²⁻ is essential for regulatory compliance, early detection of contamination, and optimizing chemical dosing. Traditional laboratory assays and sensor networks have driven demand for technologies that can deliver rapid, cost‐effective, and field‐deployable measurements. With growing emphasis on decentralized water quality management and on‐site decision support, there is a clear need for methods that circumvent the logistical hurdles of sample transport, lengthy analysis times, and complex instrument operation.

Current approaches to ionic quantification often rely on single‐point conductivity readings or sophisticated laboratory instrumentation. Conductivity alone cannot distinguish individual ion species, as different ions contribute variably to the bulk signal. Conventional analytics like ion chromatography and ICP‐MS, while precise, require extensive sample preparation, high capital investment, and skilled operators, making them impractical for continuous monitoring. Portable probes suffer from limited selectivity and sensitivity, and calibration drift leads to unreliable long‐term data. Altogether, these constraints hinder timely, accurate profiling of multi‐ion systems in situ.

Technology

This technology measures the electrical conductivity of a water sample across a series of strategically chosen dilution factors, identified through an unsupervised time-series mapping process that applies PCA for dimensionality reduction, t-SNE for further embedding, and DBSCAN for clustering. The resulting multi-dilution conductivity readings feed into an adapted cross-entropy method (CEM) algorithm: the probability density function (PDF) for each ion is first established, concentration vectors are then sampled from the PDFs under the charge-balance constraint. A small subset of the best-performing vectors with the lowest loss values is selected to update the PDFs. Upon convergence, the algorithm outputs the individual concentrations of Na⁺, K⁺, Ca²⁺, Mg²⁺, Cl⁻, and SO₄²⁻. Finally, these estimates initialize a constrained numerical solver based on modified Nernst–Einstein conductivity equations and SciPy’s trust-constr minimization to reconcile modeled and measured conductivities across all dilution levels, thereby refining the ion concentration results.  

This approach is differentiated by its hybrid fusion of physical-chemical modeling, experimental design and advanced machine learning, replacing reliance on expensive chromatography or mass spectrometry. Unlike uniform sampling, unsupervised factor selection ensures the most informative dilution points, reducing redundancy. The CEM algorithm with physics-informed loss functions embeds domain constraints directly into the optimization process, while the final solver refines predictions using first-principles equations. Together, these features create a cost-effective, accessible platform capable of in-situ, continuous and real-time monitoring of aqueous ion concentrations.

Benefits

• Cost-effective and accessible alternative to traditional analytical methods like ion chromatography and ICP-MS
• Eliminates the need for complex, expensive laboratory instrumentation
• Enables in-situ, continuous and real-time monitoring of multiple ionic concentrations
• Supports portable field test kits and integrated sensor networks
• Facilitates software integration into water engineering, environmental monitoring and industrial process platforms
• Hybrid data-driven and physics-based approach improves prediction accuracy and robustness

Applications

• Portable field ion analyzers
• Continuous water quality sensors
• Industrial effluent monitoring systems
• Agricultural water testing solutions
• Drinking water safety analyzers

Additional info

Water sample conductivity is measured at optimized dilution factors selected by PCA, a modified Manhattan distance metric, t-SNE and DBSCAN. An adapted CEM with a physics-informed loss function estimates the concentrations of Na⁺, K⁺, Ca²⁺, Mg²⁺, Cl⁻, and SO₄²⁻ while enforcing physicochemical constraints. Finally, SciPy’s trust-constr solver—incorporating modified Nernst–Einstein equations and the CEM outputs—refines individual ion concentrations by minimizing conductivity prediction errors across all dilutions.