Artificial neural networks are increasingly being used by mathematicians and scientists for complex calculations but apparently not by civil engineers. Such networks have many advantages over rigid equations and design codes, not least that they can be continually ‘retrained’ when new field or research data become available. On the downside, like people, they are not very transparent and not always reliable. This paper shows how a relatively simple neural network can outperform Burland and Burbidge's well-established 1985 ground settlement equation. The advantages and disadvantages of the technique are also discussed.
📖 افتح في inklap 🔗 DOI 📮 اطلب بحثاً