(中南大學(xué) 資源環(huán)境與建筑工程學(xué)院
采礦與巖土工程研究所,長沙 410083)
摘 要: 巖土流變是巖土工程失穩(wěn)破壞的重要原因之一。從系統(tǒng)辨識(shí)的角度，首先將巖土材料流變本構(gòu)的一般微分方程通式按實(shí)際的采樣周期轉(zhuǎn)化為線性時(shí)不變SISO系統(tǒng)的離散差分方程格式，構(gòu)建了用于巖土流變本構(gòu)模型辨識(shí)的BP神經(jīng)網(wǎng)絡(luò)模型；然后探討了該神經(jīng)網(wǎng)絡(luò)模型用于巖土流變本構(gòu)模型辨識(shí)的基本步驟以及其網(wǎng)絡(luò)結(jié)構(gòu)參數(shù)（輸入層神經(jīng)元數(shù)和網(wǎng)絡(luò)連接權(quán)值）與SISO流變系統(tǒng)差分方程模型參數(shù)間相互轉(zhuǎn)化的算法原理，并據(jù)此在Matlab軟件平臺(tái)中編制了BP網(wǎng)絡(luò)辨識(shí)算法的相應(yīng)程序CYJ1.M；最后，采用有關(guān)的考題驗(yàn)證證明該辨識(shí)算法是成功可信的。
關(guān)鍵字: 巖土材料; 流變; 本構(gòu)模型; 系統(tǒng)辨識(shí); 人工神經(jīng)網(wǎng)絡(luò)
CAO Ping, WANG Wen-xing
（Institute of Mining and Rock Soil Mechanics Engineering， College of Resources, Environment and Civil Engineering , Central South University, Changsha 410083, China）
Abstract:The rheology of rock and soil is one of the important reasons why geotechnical engineering is often apt to lose its stability and be damaged. From the view of system identification, the author first changes the general constitutive differential equation of the rheology of rock and soil into the discrete difference equation of the linear time—invariant single—input single—output system using the actual sampling period and builds the BP neural network model to identify the rheological constitutive model of rock and soil, and then probes into the basic steps of the identification of the rheological constitutive model of rock and soil using the BP model and studies the algorithmic principle to translate the BP network structure parameters (the number of the network’s neurons and the values of the network’s connections weights) into the model parameters of the difference equation of the SISO rheological system. Based on the principle the author also compiles the application program CYJ1.M for the identification arithmetic using the BP neural network on the base of MATLAB software .At last, by using an examination question, the identification arithmetic is testified to be successful and reliable.
Key words: rock and soil material; rheology; constitutive model; system identification; artificial neural network
