基于特征的全船焊接工时聚类分析

Clustering analysis of welding man-hours based on characteristics of shipbuilding

以某型船全船焊接信息为样本,首先验证了数据样本服从对数正态分布;再以距离作为度量标准,对焊接信息进行基于不同距离测量方法的聚类分析,并重点对比平方欧氏距离和夹角余弦距离的聚类树形图,得出全船焊接信息合理的聚类数目;最后对每个类进行物理信息的解释,得到完整实用的基于特征的全船焊接工时估算标准,为新型船经济性论证及方案选型提供快捷、可靠的依据。

Taken a certain type of ship welding information as the sample. Firstly,to verify the sample data obeys the lognormal distribution; secondly,with distance as a metric,to cluster the welding information as different distance measurement method,and compare the results of clustering tree between the squared euclidean distance and the cosine distance,then to get the reasonable number of welding clustering; finally,to explain the physical information of each cluster results,the whole ship welding man-hour estimate standard based on the characteristics had been measured,and they would provide fast,reliable basis for the selection of new ship economic demonstration and scheme. Specific ideas are that,firstly descriptive statistical analysis of welding information by the data mining technology, including exploring analysis,Q-Q distribution test,to get the distribution of the sample data; secondly based on the welding man-hours quota standard,according to the weld work information of operation area and dispatch list description,to do cluster analysis by different distance measurement method; thirdly to compare the different clustering tree,get the very clustering scheme that really reflect the physical information of ship welding characteristics; finally,date process each characteristics clusters,obtain the man-hour weight of this cluster, and form a complete and practical ship welding man-hour estimate standard based on manufacture characteristics.