Abstract: The optimization of engineering control network,especially that of the deformation monitoring network and precise control network for large construction is a very important work in network design.But in practice it is still difficult for the users to realize the network optimization with analytical or analogical method.In this paper,after interpretation on the optimization design methods of engineering control network and the quality standards such as precision,reliability,sensitivity and costs,a new algorithm of optimization design with simulative method for engineering control network based on the inner reliability of observations is presented.The relationship between inner reliability and other quality standards is discussed in detail.The great advantages and characteristics of this new method are that the result of optimization is determined uniquely and not on the knowledge and experience of the designer.Otherwise some new concepts,such as average observation redundant,the design number of observation and matching criterion for the precision of distance and direction observation,are proposed and defined. A general software packet for data processing and adjustment of surveying network (cosa) used for the network optimization design with this method is introduced.With this software,a primary network observation plan is simulated by *.OB2 file.Then the corespondent observation file *.IN2 will be created automatically according to the above *.OB2 file.After adjustment,the redundant of every observation will be calculated.Based on the average redundant and observation design number,surveyors may make a decision that which observations could be deleted.Finally,two examples are given.One is a simulated triangu-trilateration network for bridge project.The work load of the final optimal measuring plan is reduced to 36 from the initial 75 observations and the precision and reliability are still satisfied for the design.The other is a real large construction network.For this network,there are 186 observations at the beginning.After the optimal processing the observations are reduced to 119 and the precision of the point with the lowest precision is not degraded significantly.The optimization benefits and necessity can be seen clearly through the above examples.Furthermore,this method can not only make the optimization of network,but also be used to interpret the rationality of an existing network.