Abstract: With the popularity of standardized breeding and artificial insemination, fetal time and insemination number have become the main factors influencing the reproduction of swine herd. In the foregoing methods, a simple average of the conception rate has been used for the monitoring of breeding, which is not precise because it is highly dependent on the age structure and other influence factors of the herd. An appropriate monitoring system for breeding must be adjusted for these systematic effects, being in a position to capture correlations between fetal time and insemination number, and developing over time. In order to improve the accuracy of the fetal time and insemination number monitoring in pig breeding, this paper developed and implemented a new breeding surveillance system in pig farm based on the dynamic generalized linear model. The dynamic generalized linear model is suitable for statistical data in accordance with the binomial distribution. It includes an observation equation and a system equation. The observation equation associates observation variables and parameters, and the system equation indicates the change process of impact factor over time. The key observation used throughout the system is "farrowing rate". Since "conception rate" has to be measured indirectly through the percentage of sows that return to oestrus 21 days after service, or based on pregnancy diagnosis at about 30 days post-service. The farrowing rate is a more reliable numeric indicator of the successful conception, and it is defined as the total number of sows farrowing divided by the total number of sows mating, and expressed as a percentage. Through the analysis of historical data of the target pig farm, we found that there are no significant differences in farrowing rate during the first 5 parities of sows. From the sixth parity, farrowing rate shows a significant downward trend. We made a negative slope on behalf of this downward tendency in farrowing rate. Besides, there is a kind of data representing the destructive effect of insemination number on reproductive performance. Based on the statistics of historical data of target pig farm, we achieved automatically updating of the impact factors' values using the system equation. The results of the dynamic generalized linear model are monitored using control charts inspired by Shewhart. A control chart is composed of 3 elements: a central line (CL), corresponding to a target value; an upper control limit (UCL) and a lower control limit (LCL). With the updating equation developed by Bono based on Bayes rule and Taylor expansion, we got the values of CL, UCL and LCL. The control limits were drawn using a 95% confidence interval built on the forecast variance. The monitoring method is a weekly control of the number of observed events (observed farrowing sows) compared to the LCL. An alarm is triggered when observed events are below the LCL. Workers in pig farm can choose any time range and the monitoring information to create the chart. The result of practical application shows that the system runs stably and its rate of false positives is lower than 2%. The study not only increases the accuracy of the pig breeding information monitoring, but also provides reference for the further analysis of potential information in pig production data. Suggestions for future improvements are adding the steps of forward filtering and backwards smoothing and the inclusion of a "sow effect" in the farrowing model.