Comparison of statistical methods of radiation monitoring

Shumakov, A.V.; Kasatkin, V.P.

doi:10.1007/s10512-010-9218-y

Cited by 3 publications

(1 citation statement)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…based on the Neyman-Pearson criterion [2]. According to this criterion, a threshold count or detection level L is found from the chosen false-alarm probability α (significance of the criterion) with known background K b for the detection block of the monitor:…”

Section: Threshold Of a Radiation Monitormentioning

confidence: 99%

Determination of the detection threshold of a radiation monitor

Shumakov

2013

At Energy

Self Cite

View full text Add to dashboard Cite

A method is proposed for determining the detection threshold of radiation monitors, using a radiation source with arbitrary flux, activity or mass of nuclear material. Good agreement is shown between the detection threshold of several neutron radiation monitors, determined using 244 Cm and 252 Cf sources with different fission-neutron fluxes. An example of converting the detection threshold measured by the proposed method or the number of detections with different background to other conditions is presented.A radiation monitor recoding radiation from nuclear materials and radioactive substances serves as a means for detecting unauthorized movement of such materials. The main characteristic of a radiation monitor is the detection threshold. The experimental determination of the detection threshold of a radiation monitor recording photon and (or) neutron radiation is based on the use of a source of corresponding radiation with parameters as a close as possible to it [1].In practice, problems arise in determining the detection threshold. One is that it is impossible to guarantee in advance that standard samples -fission neutron sources or photon radiation -with the required characteristics exist. For this reason, in many cases the numerical values of the detection threshold are not presented, especially for foreign-made monitors. Ordinarily, it is simply indicated that the monitor satisfies the requirements of the corresponding standards adopted in this country.The detection threshold determines the sensitivity of the radiation monitor -the lower the threshold, the higher the sensitivity. Usually, it is expressed as the mass of the nuclear material M th . An alternative unit of measurement is the activity A th of the radionuclide for a γ-ray monitor or the fission neutron flux Φ th for the neutron monitor. The relation between the γ-ray flux and M th or A th is not unique and depends on the density of the material, chemical composition, geometric dimensions and configuration of the source. Depending on the detection threshold, the categories of radiation monitors are determined by a standard [1].The concept of a detection threshold has a statistical basis. The aim of statistical processing in radiation monitoring practice is to detect a source of radiation. It is necessary to determine whether the detected radiation is background radiation or the alternative hypothesis is true -the detected radiation is significantly above background.Usually, the parameters of the background statistical distribution are relatively well known. Before the monitoring starts, the background is measured for 10-20 working time intervals of the monitor. Next, the background is refreshed, if the monitor has not been actuated, i.e., a significant excursion above the detection level adopted has not occurred, or remains unchanged.The alternative hypothesis that a source is present is not definite. The available quantitative information on the hypothesis is limited to several measurements. A decision must be made on the basis of this informatio...

show abstract

Section: Threshold Of a Radiation Monitormentioning

confidence: 99%

Determination of the detection threshold of a radiation monitor

Shumakov

2013

At Energy

Self Cite

View full text Add to dashboard Cite

show abstract

Experimental comparison of algorithms used in radiation monitoring

Shumakov¹,

Okishev²

2012

At Energy

View full text Add to dashboard Cite

The results of an experimental comparison of algorithms used in radiation monitoring are presented. Serially produced TSRM-type radiation monitors, which are manufactured by the All-Russia Research Institute of Automatics, and radioactive sources are used. The data coming from the monitor's detector are processed using all algorithms simultaneously in real time. Cases of low and high counts, which conform to Poisson and Gaussian statistics, with average values of about 10 and 100, respectively, are examined. Good agreement is obtained between the experimental and computed data. The probability ratio and a priori probability algorithms give the best results. The use of these algorithms in radiation monitors will make it possible to increase the probability of detecting radiation sources by a factor of 2-3 without increasing the frequency of false alarms.The algorithms used for statistical processing of information flowing from the detectors in radiation monitors as well as the means for monitoring a medium are based on the detection of weak, short-time, random neutron or photon radiation. Algorithms based on the Neyman-Pearson and probability ratio criteria as well as moving average, digital recursive filter, a priori probability, half-sum, and variance ratio methods have been compared by the Monte Carlo method in previous work [1].The results of an elementary check of model calculations and comparison of these algorithms performed using radioactive sources and TSPM serial radiation monitors manufactured by the All-Russia Research Institute of Automatics (VNIIA) are presented below. The data entering from the detecting block of a monitor were processed simultaneously in realtime. Cases of low and high counts, conforming to Poisson and Gaussian statistics with average value about 10 and 100, respectively, are examined.The moving average principle with five sub-interval measurements and duration 0.2 sec was used for all algorithms. The algorithm based on the Neyman-Pearson criterion is identical to the moving average method. The number of tests or the size of the experimental count array did not exceed 5·10 3 events. It is best to perform a large number of tests but this requires a long measurement time. The average count with a source exceeded the background level by 10 to 50% for the Poisson distribution and by 2 to 10% for the Gaussian distribution.Algorithms were compared for two false alarm probabilities α = 10 -2 and α = 10 -3 , characteristic for portable and stationary monitors, respectively. The fact that the probability of a false alarm does not correspond to the same or close real values for all processing algorithms was taken into account. For different algorithms and the same false-alarm probability, the real probability can differ by more than a factor of 100. For this reason, modeling was used to set the probability α mod so that the number of false alarms is the same for all algorithms.In Table 1, the experimental data are compared with the model calculations. The average of three events is given as ...

show abstract