In different industrial sectors, alarms are used to ensure satisfactory performance and prevent damages to the system. In modern industrial complexes, many variables are measured by sensors. During normal operation, each individual measurement should fall in an acceptable range. But this is not always the case, and often various measurements exceed their range due to faults. To inform the operator, who is responsible for safe and optimal operation, of these situations, alarm systems are used. But then, industrial alarm systems have various problems, which often confuse the operator and reduce reliability of the system. Some of these problems are caused by ambient conditions. If the ambient is noisy, the sensor readings will be different from the actual values of the variables, which can result in an alarm being triggered. These are called false alarms. On the other hand, a fault might occur without triggering an alarm, which is known as a missed alarm. Moreover, repetitive triggering of an alarm in a short period, and the latency between fault occurrence and alarm triggering are other concerns in alarm systems. To improve the performance and reduce errors in alarm systems, performance indices have been introduced, including false alarm rate, missed alarm rate, chattering index and alarm delay. Also, to improve these indices, different methods have been introduced, such as deadband, filtering of process data, and delay timers. In this thesis, we first investigate alarm latching method. Then we present reduced delay timer as a new method for improving alarm quality. Then, using theory of Markov chains, we study the effect of these two methods on alarm performance indices and determine their relationships. Keywords: Alarm Systems, Alarm Management, Delay Timer, Reduced Delay Timer, Alarm Latch