Abstract: To reduce the maintenance cost of the integrated manufacturing software system,the aperiodic inspection strategy is an important method to trigger software rejuvenation.In the existing optimization research on this strategy,the reliability objectives or constraints adopted are fixed in the entire software operating cycle,and cannot be dynamically adjusted according to the software operating status.Therefore,a multi-state aperiodic inspection strategy and its optimization method were proposed.In this strategy,the inspection reliability in different state intervals was taken as the decision variable and the minimum maintenance cost of the system as the object to establish an optimization model to dynamically determine the optimal inspection period.Based on the Markov semi-renewal process characteristics of the aperiodic inspection strategy,a discrete maintenance cost evaluation model under multi-state reliability constraints was established.The optimal strategy configuration parameters were searched by the genetic algorithm to obtain the optimal inspection period.Experimental results showed that compared with the existing single-state aperiodic inspection method,the proposed method could reduce the system maintenance cost by about 5.32% on average,and the maximum could be reduced by about 11%.

Key words: aperiodic inspection, software rejuvenation, Markov semi-renewal cycle, reliability optimization

摘要： 为减少集成制造软件系统维护成本,非固定周期检测主要用于触发软件再生。现有针对该策略优化的研究中,采用的可靠性目标或约束在整个软件运行周期内固定不变,无法根据软件运行状态动态调整。故提出一种多状态非固定周期检测策略及其优化方法。该策略将不同状态区间内的检测目标可靠性作为决策变量,以最小化系统维护成本为目标,建立优化模型动态确定最优检测周期。基于非固定检测策略的Markov半更新过程特性,建立多状态可靠性约束下离散化的维护成本评估模型,通过遗传算法求解最优策略配置参数,进一步得到最优检测周期。实验结果表明,与现有单状态非固定周期检测方法相比,平均可降低约5.32%的系统维护成本,最高可降低约11%维护成本。

关键词: 非固定周期检测, 软件再生, Markov半更新过程, 可靠性优化