自动制图综合人工神经元网络方法的研究

现代神经生理学研究表明 ,人脑的大量神经元构成有所分工而又紧密联系的神经元网络 ,它的结构和功能可以采用物理可实现的系统人工神经元网络来模拟。制图综合是人脑神经元网络获取、处理、输出地理信息的复杂视觉思维过程 ,可以用人工神经元网络来模拟。文中探讨了制图综合的人工神经元网络的设计 ,并对用于实现自动制图综合的结果进行分析 ,指出其应用前景。     

自由网平差方法及统计特性新探

本文结合参数平差和主成分估计理论,导出了误差方程中含多重共线性时未知参数的求解公式,并以定理的形式,证明了主成分估计的解是极小范数解。由此,将主成分估计推广到秩亏网平差中,同时导出了未知参数估值之协因数阵的计算公式,同时,证明了自由网平差的传统特性。     

Development of Seismic Risk Scenarios Based on a Hybrid Method of Vulnerability Assessment

A hybrid methodology of vulnerability analysis is presented, involving elements from both empirical and theoretical methods. A model for correlating analytically calculated structural damage indices to loss (in monetary terms) is also proposed and calibrated against available statistical data. Probability damage matrices derived using this methodology are incorporated into a cost-benefit model tailored to the problem of estimating the feasibility of seismically rehabilitating the existing stock of reinforced concrete buildings in Thessaloniki, Greece. Losses calculated using the suggested procedure are found to be in good agreement with losses incurred during the 1978 Thessaloniki earthquake. The results of the present study also indicate that benefit/cost ratios for reinforced concrete buildings are quite low. Hence, it appears that a pre-earthquake strengthening programme is not economically justifiable.     

Seismic Zoning of Croatia

The territory of Croatia and neighboring regions is divided into 17 seismic source zones, considering available seismological and geological data. On this basis, seismic hazard elements (seismicity rate, maximum magnitude, b-value, probabilities of exceedance and return periods for a predefined set of magnitudes) are computed using the maximum likelihood method appropriate for treating data-sets with variable completeness thresholds. The values of long term expected peak horizontal acceleration obtained by using a combination of the deterministic and the probabilistic procedure are the highest in the Dubrovnik zone, while the Zagreb zone has the highest earthquake hazard in the continental part of the country. This revised version was published online in June 2006 with corrections to the Cover Date.     

Seismic quiescence before the M 7, 1988, Spitak earthquake, Armenia

A detailed analysis of the 35  yr of seismicity between 1962 and 1997 using a gridding technique shows that the M 7, Spitak earthquake of 1988 December 7 was preceded by a quiescence anomaly that started at approximately 1984±0.5, and lasted about 5±0.5  yr, up to the main shock. This quiescence anomaly had a radius of about 20±3  km, estimated from circular areas with 75 per cent rate decrease, centred at the point of maximum significance of the anomaly. The quiescence was clearly present in the aftershock volume during the 5  yr before the 1988 main shock, but its statistically strongest expression was located 30  km NW of the epicentre. This anomaly fulfills the association rules between precursory quiescence anomalies and main shocks, even for a tight definition, and is therefore proposed as a case of precursory quiescence. The largest value of the standard deviate Z , found by random selection of samples by gridding, was Z =14 for a time window of T _w=3  yr, using a sample size of N =300 events. This makes this anomaly the strongest observed so far, and it is the first documented in an environment of continental collision. There are no false alarms exceeding in significance the precursor. The Armenian earthquake catalogue used for this study had 4600 earthquakes with M ≥ M _min=2.2 in the area bounded by 39.5° to 42°N/42.5° to 47°E. From the point of view of homogeneous reporting this is the best catalogue we have analysed so far. The limits of the data used and the density of the grid are dictated by the data, and have no influence on the results. The choice of free parameters does not influence the results significantly within the following limits: 100≤ N ≤500, 2≤ T _w≤7, 2.2≤ M _min≤2.8.