AIZOTH predicts concrete compressive strength

AIZOTH has published a case study showing its Multi-Sigma® platform predicting and optimizing concrete compressive strength from unknown mix-proportion data, using a concrete production dataset and a no-code AI workflow. The company said the work covered three steps: prediction for unknown mixes, contribution analysis to identify which variables mattered most, and optimization to search for conditions that maximize strength. (aizoth.com) The case study is posted on AIZOTH’s website, which describes Multi-Sigma® as a browser-based AI analysis platform for prediction and optimization. The reported numbers are specific. AIZOTH said it split 1,030 samples into 930 training samples and 100 test samples, then built an AI model from the training data. On the test set, the company reported a relative error of 5.14%, root mean square error of 2.31 and a correlation coefficient of 0.98. ### What exactly did AIZOTH say the model was doing? AIZOTH said the model was built to predict compressive strength from explanatory variables tied to the concrete mix and production conditions. The company framed the exercise as a way to estimate strength for unknown mix-proportion data rather than only describe already-tested batches. (aizoth.com) Multi-Sigma® is described by AIZOTH as a no-code AI analysis platform that combines prediction functions with optimization tools, including deep learning, Bayesian analysis and genetic algorithms. In this case study, the company said those capabilities were used in sequence: first to predict, then to quantify variable contributions, and then to search for a stronger mix. (aizoth.com) ### Which variables did the case study identify as most important? AIZOTH said age since production had the strongest positive effect on compressive strength in its contribution analysis. The company also said higher proportions of cement and blast-furnace slag had positive effects, while higher proportions of water and fine aggregate had negative effects. (aizoth.com) The contribution analysis matters because AIZOTH presented it as more than a black-box forecast. The company said Multi-Sigma® includes a sensitivity-analysis-based feature that evaluates how each explanatory variable affects the target variable. (aizoth.com) ### What mix did the optimization step produce? AIZOTH said the optimization search estimated a highest-strength set of conditions that included an age of 172 days, cement content of 397.87 kg/m³, blast-furnace slag content of 358.93 kg/m³, water content of 133.76 kg/m³ and fine aggregate content of 737.00 kg/m³, among other factors. (aizoth.com) The company said the maximum and minimum values of compressive strength both increased as the number of generations progressed during the search. Those figures came from an optimization routine inside Multi-Sigma®, not from a newly disclosed field pour or construction project. (aizoth.com) AIZOTH’s page presents the result as an estimated optimum generated from the dataset analysis. ### Where did the data come from? AIZOTH said the data used in the analysis was processed and edited from the Kaggle Concrete Compressive Strength dataset, which it cited on the case-study page under a Creative Commons Attribution 4.0 International license. The company did not present the study as a proprietary customer deployment in the posted materials. (aizoth.com) That means the demonstration is best read as a worked example of the platform’s prediction, contribution-analysis and optimization functions using a public dataset. AIZOTH’s website separately says the company is based in Tsukuba, Japan and provides AI system development and consulting built around Multi-Sigma®. (aizoth.com) ### Where can readers check the underlying claims? AIZOTH has posted the full concrete-strength case study on its English-language use-case page, where the train-test split, error metrics and optimization outputs are listed. The company also maintains a download page for Multi-Sigma® materials and other use cases. (aizoth.com) AIZOTH’s current product page describes Multi-Sigma® as a cloud-based, no-code AI analysis platform, and the company’s website says it operates from Tsukuba, Ibaraki Prefecture, Japan. Those pages are the next stop for readers looking for the concrete case study and related product documentation. (aizoth.com 1) (aizoth.com 2)