Demonstration of compliance with the parameters of portable fire extinguishers is provided in the form of certification testing for reliability and fire extinguishing capacity. Reliability testing is performed on 32 randomly selected from the batch extinguishers, and the acceptance number of failures should be zero. Fire extinguishing capacity is determined by fire test, and type of tested fire extinguishers considered successfully passed test if 2 of 3 extinguishers have extinguished model fire site (MFS) of specified class and rank, i. e. the acceptance number of failure is one. Those sample arrays aren’t aligned to each other, to quality control normative documents and to reliability estimation procedure. To determine the size of sample array of fire tests sequential analysis method is used. Allowable risk in this case is determined from the requirement that the probability of rejection of batch of extinguishers p does not exceed predetermined value a ( p £ p0), in which p0 is the upper limit of the probability of failure, and the probability of passing the test of the batch doesn’t exceed b, in case p1 £ p . At every m-numbered test number of defective products dm is counted. Compliance with MFS rank of tested extinguishers batch is determined by results of every test extinguishing MFS ( d = 1 or d = 0). Option of passing the test of all of the 32 extinguishers is economically inexpedient, so the size of sample array should be reduced without reducing product quality. The size of sample array depends on value of lower limit of failure probability firstly, so physico-chemical processes of MFS extinguishing were analyzed to substantiate the size of sample array. According to combustion theory fire stops if heat removal from burning area exceeds evolution of heat from combustion focus. Heat absorption of extinguishing powder, operator actions and constructive features of fire extinguisher have influence on extinguishing effect. Numeric value of heat removal from combustion focus will vary from test to test of extinguishers from even one batch respectively. In this case amount of heat absorbed by extinguishing powder will be found in between values corresponding to MFS of lower and higher rank. It is known in theory that the probability of random value to be found in the interval between standard deviations is rated 0.67. So, the probability of excess or lack of heat taking place is rated 0.165. That’s why value 0.165 is taken as lower limit of probability of failure for determining the size of sample array for fire tests. With this in mind limit amount of samples is determined by sequential analysis method (with b = a = 0.2 and p0 = 0.004) and equals 8 extinguishers in case of zero amount of defective units. Advantage of this model is based on possibility to continue test with increased amount of extinguishers in case of failure of one. For example in case of failure at first iteration of test whole batch will be considered valid if 22 following extinguishers won’t be rejected, i. e. amount of samples should be 30. In case of failure at subsequent iteration sample array should be extended to 52 extinguishers. The novelty of this method lies in the differentiated approach to the testing of fire extinguishers.