Comparative modelling of strength properties of hydrated-lime activated rice husk-ash (HARHA) modified soft soil for pavement construction purposes by artificial neural network (ANN) and fuzzy logic (FL)
Artificial neural network and fuzzy logic based model soft-computing techniques were adapted in the research study for the evaluation of the expansive clay soil-HARHA mixture’s consistency limit, compressibility and mechanical strength properties. The problematic clay soil was stabilized with vary...
Saved in:
| Main Authors: | , , , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Penerbit Universiti Kebangsaan Malaysia
2021
|
| Online Access: | http://journalarticle.ukm.my/17817/1/20.pdf http://journalarticle.ukm.my/17817/ https://www.ukm.my/jkukm/volume-332-2021/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Artificial neural network and fuzzy logic based model soft-computing techniques were adapted in the research study for
the evaluation of the expansive clay soil-HARHA mixture’s consistency limit, compressibility and mechanical strength
properties. The problematic clay soil was stabilized with varying proportions of HARHA (stabilizing agent) which is an
agricultural waste derivative from the milling of rice ranging from 0% to 12%; the utilization of the alkaline activated
wastes encourages its recycling and re-use to obtain sustainable, eco-efficient and eco-friendly engineered infrastructure
for use in the construction industry with economic benefits also. The obtained laboratory and experimental responses were
taken as the system database for the ANN and fuzzy logic model development; the soil-HARHA proportions with their
corresponding compaction and consistency limit characteristics were feed to the network as the model input variables
while the mechanical strength (California-bearing-ratio (CBR), unconfined-compressive-strength (UCS) and Resistance
value (R-values)) responses of the blended soil mixture were the model target variables. For the ANN model, feed forward
back propagation and Levernberg Marquardt training algorithm were utilized for the model development with the optimized
network architecture of 8-6-3 derived based on MSE performance criteria; while for the fuzzy logic model, the mamdani
FIS with both triangular and trapezoidal membership function with both models formulated, simulated and computed using
MATLAB toolbox. The models were compared in terms of accuracy of prediction using MAE, RMSE and coefficient of
determination and from the computed results, 0.2750, 0.4154 and 0.9983 respectively for ANN model while 0.3737, 0.6654
and 0.9894 respectively was obtained for fuzzy logic model. The two models displayed robust characteristics and performed
satisfactorily enabling the optimization of the solid waste derivatives utilization for soil mechanical properties improvement
for engineering purposes. |
|---|