Hello All,

Dangerous Goods classification is one of the critical activities for any organization to maintain compliance with regulatory requirements and also ensuring not to harm persons and environment. The classification of Dangerous goods is a skill which is getting scarce in organizations due to various reasons like availability of SME’s and knowledge gaps etc. Use of Machine Learning and Deep Learning for Dangerous Goods management offers a very good opportunity to compliment the resources who are already working on the DG classification. Based on DG based preliminary classifications a model can be trained over a period of time to provide you the information as accurately as possible.

white freight truck close up photography
Photo by Craig Adderley on Pexels.com

There are various ways by which a model can be built. For example applications of Machine Learning / Deep Learning for Dangerous goods managements spans across i) Identifying whether a material is dangerous goods ii) The UN number into which it falls and iii) Identification of characteristics for Dangerous Goods classification.

Machine Learning approach for DG Classifier:

  • A material can be designated as a Dangerous Goods based on certain primary data like Physical Chemical properties, Hazards Identification, Fire Fighting Measures, Handling and Storage, Stability & Reactivity, Toxicological and Eco-Toxicological information
  • The Components information like Standard Composition
  • Using Machine Learning Classifier Algorithms, a prediction model can be developed which can help in identification of whether a Product / Mixture falls into the category of Dangerous or Non Dangerous Good
  • Since the composition of products vary from company to company and technique of DG varies as well, the model needs to be calibrated with target company Data.

Machine Learning approach for identification of UN Classifier:

  • There is Regulatory Content packages available which provides a rule-based system to support experts in classifying dangerous goods
  • This model uses Classifier Rules Logic for specifically predicting the Dangerous Goods UN Number, Risk Classifier information
  • Based on Materials, components and existing UN number information, new material is determined for the UN Number information
  • Since the composition of products vary from company to company and technique of DG varies as well, the model needs to be calibrated with target company Data.

Deep Learning approach for prediction of Dangerous Goods characteristics:

  • Deep Learning model creates a complete prediction of DG Characteristics for any material.
  • It uses Composition, Component and Product Physio-chemical data along with Toxicological & GHS information
  • This model will be developed based on a specific company data only as it needs careful analysis of existing data.
  • The output from the model would be Prediction of Dangerous Goods Class, Labels, Hazard Inducers and Transport classification data
  • It needs huge dataset and massive computing power

I have made a demo video on the Machine Learning / Deep Learning based DG Classification with examples using Wikipedia pages. I have used Ananconda / Tensorflow & Knime to demo the scenarios. I will be posting them in next couple of days.

Besides above machine learning / deep learning can also help organizations in management of I) Determination of fire / explosion hazard ii) Toxicity prediction iii) Chemical Control Laws determination iv) Chatbots for Product Safety data dissemination.



Leave a Reply

Close Menu
%d bloggers like this: