Methods and systems for predicting a remaining shelf-life of a millet flour using data-driven approach

Patent metadata

Field	Value
Publication number	US-2025182860-A1
Application number	US-202418931321-A
Country	US
Kind code	A1
Filing date	Oct 30, 2024
Priority date	Dec 2, 2023
Publication date	Jun 5, 2025
Grant date	—

Abstract

The disclosure relates generally to methods and systems for predicting a remaining shelf-life of a millet flour using data-driven approach. Conventional techniques for estimating the remaining shelf life of the millet flour are very limited and not effective. The methods and systems of the present disclosure employs a Long short-term memory (LSTM) network architecture-based model to predict the remaining shelf-life of the millet flour based on both the nutritional quality and the rancidity using the information of millet variant and days after milling of the grain. The LSTM network based model is trained on the training data obtained using the chemical reaction kinetic model from the input data comprising a peroxide content, and a fat acidity. The present disclosure predicts the remaining shelf-life of the millet flour in terms of both the remaining nutritional shelf-life and the remaining rancid shelf-life.

First claim

What is claimed is:

1 . A processor implemented method, comprising:
receiving, via one or more hardware processors, a plurality of training millet flour samples, wherein each training millet flour sample of the plurality of training millet flour samples comprises one or more millet flour input parameters, one or more nutritional parameters, and one or more rancidity parameters; generating, via the one or more hardware processors, a training dataset of each training millet flour sample, to obtain a plurality of training datasets from the plurality of training millet flour samples, using a chemical reaction kinetic model, wherein the training dataset of each training millet flour sample comprises a value of each of the one or more nutritional parameters, a value of each of the one or more rancidity parameters, a value of a remaining nutritional shelf-life, and a value of a remaining rancid shelf-life; and training, via the one or more hardware processors, a LSTM network-based model, with the plurality of training datasets associated to the plurality of training millet flour samples, to obtain a trained millet flour shelf-life estimation model.

2 . The processor implemented method of claim 1 , further comprising:
receiving, via the one or more hardware processors, one or more millet flour input parameters of a test millet flour sample; and passing, via the one or more hardware processors, the one or more millet flour input parameters of the test millet flour sample, to the trained millet flour shelf-life estimation model, to estimate the value of the remaining nutritional shelf-life, and the value of the remaining rancid shelf-life of the test millet flour sample.

3 . The processor implemented method of claim 1 , wherein:
the one or more millet flour input parameters of each of the plurality of training millet flour samples comprises (i) a millet flour variant of a plurality of millet flour variants, and (ii) a number of hours after milling of the millet flour variant of the plurality of millet flour variants, the one or more nutritional parameters comprises an unsaturated fatty acid concentration, and the one or more rancidity parameters comprises (i) a fat acidity, and (ii) a peroxide content.

4 . The processor implemented method of claim 1 , wherein generating the training dataset of each training millet flour sample, to obtain the plurality of training datasets from the plurality of training millet flour samples, using the chemical reaction kinetic model, comprising:
receiving an input data related to each training millet flour sample associated with each training dataset, from a literature, wherein the input data comprises a value of a peroxide content, and a value of a fat acidity; optimizing a rate constant of each of the one or more nutritional parameters and each of the one or more rancidity parameters, by curve fitting of the input data, using the chemical reaction kinetic model, to obtain an optimized rate constant of each of the one or more nutritional parameters and each of the one or more rancidity parameters; determining the value of each of the one or more nutritional parameters and the value of each of the one or more rancidity parameters, using the optimized rate constant associated to each of the one or more nutritional parameters and each of the one or more rancidity parameters; determining the value of the remaining nutritional shelf-life, and the value of the remaining rancid shelf-life, by limiting the value of each of the one or more nutritional parameters and the value of each of the one or more rancidity parameters; and adding the value of each of the one or more nutritional parameters and the value of each of the one or more rancidity parameters, the value of the remaining nutritional shelf-life, and the value of the remaining rancid shelf-life, to generate the training dataset associated to the training millet flour sample.

5 . The processor implemented method of claim 1 , wherein training the LSTM network-based model, with the plurality of training datasets associated to the plurality of training millet flour samples, to obtain the trained millet flour shelf-life estimation model, comprising:
passing the one or more millet flour input parameters of each of the plurality of training millet flour samples, to the LSTM network-based model, to obtain a predicted value of each of the one or more nutritional parameters and the predicted value of each of the one or more rancidity parameters, the predicted value of the remaining nutritional shelf-life, and the predicted value of the remaining rancid shelf-life; determining value of a loss function of the LSTM network-based model, based on a difference between (i) the predicted value of each of the one or more nutritional parameters and the predicted value of each of the one or more rancidity parameters, the predicted value of the remaining nutritional shelf-life, and the predicted value of the remaining rancid shelf-life, and associated values of (ii) each of the one or more nutritional parameters and each of the one or more rancidity parameters, the remaining nutritional shelf-life, and the remaining rancid shelf-life; and updating one or more network weights of the LSTM network-based model, based on value of the loss function of the LSTM network-based model, for training with a next training millet flour sample, until the plurality of training millet flour samples is completed, to obtain the trained millet flour shelf-life estimation model.

6 . A system, comprising:
a memory storing instructions; one or more input/output (I/O) interfaces; and one or more hardware processors coupled to the memory via the one or more I/O interfaces, wherein the one or more hardware processors are configured by the instructions to: receive a plurality of training millet flour samples, wherein each training millet flour sample of the plurality of training millet flour samples comprises one or more millet flour input parameters, one or more nutritional parameters, and one or more rancidity parameters; generate a training dataset of each training millet flour sample, to obtain a plurality of training datasets from the plurality of training millet flour samples, using a chemical reaction kinetic model, wherein the training dataset of each training millet flour sample comprises a value of each of the one or more nutritional parameters, a value of each of the one or more rancidity parameters, a value of a remaining nutritional shelf-life, and a value of a remaining rancid shelf-life; and train a LSTM network-based model, with the plurality of training datasets associated to the plurality of training millet flour samples, to obtain a trained millet flour shelf-life estimation model.

7 . The system of claim 6 , wherein the one or more hardware processors are further configured to:
receive the one or more millet flour input parameters of a test millet flour sample; and pass the one or more millet flour input parameters of the test millet flour sample, to the trained millet flour shelf-life estimation model, to estimate the value of the remaining nutritional shelf-life, and the value of the remaining rancid shelf-life of the test millet flour sample.

8 . The system of claim 6 , wherein:
the one or more millet flour input parameters of each of the plurality of training millet flour samples comprises (i) a millet flour variant of a plurality of millet flour variants, and (ii) a number of hours after milling of the millet flour variant of the plurality of millet flour variants, the one or more nutritional parameters comprises an unsaturated fatty acid concentration, and the one or more rancidity parameters comprises (i) a fat acidity, and (ii) a peroxide content.

Assignees

• Tata Consultancy Services Ltd

Inventors

• Pande Harshvardhan
• Kapse Shrikant Arjunrao
• Kausley Shankar Balajirao
• Rai Beena

Classifications

• G06Q10/087

  Inventory or stock management, e.g. order filling, procurement or balancing against orders · CPC title

• G06N3/0442

  characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU] · CPC title

• G16C20/70

  Machine learning, data mining or chemometrics · CPC title

• G16C20/30 Primary

  Prediction of properties of chemical compounds, compositions or mixtures · CPC title

• G16C60/00 Primary

  Computational materials science, i.e. ICT specially adapted for investigating the physical or chemical properties of materials or phenomena associated with their design, synthesis, processing, characterisation or utilisation · CPC title