The most important Euroclass test method for products with a non-negligible contribution to fire is the Single Burning Item (SBI) test . Correlation between the results of the SBI test and the cone calorimeter is an issue of great interest. The cone calorimeter  is a well-established and acknowledged test method, and it requires only a small amount of specimen material. Even though the official classification of products in Europe is made on the basis of the SBI test results, the cone calorimeter can be a useful tool for product development and quality control. Several modelling approaches on the prediction of heat release and classification in the SBI test have been published [50, 51, 52, 53]. Many models predict well the performance of untreated wood products in the SBI tests, but fire retardant treated wood has proven problematic in several cases. An application of a model, developed especially for fire retardant treated wood products, is presented below.
This section contains following topics:
The starting point of the calculation is the basic flame spread equation 
As a result of calculations according to Eqs. (8) and (9), predicted heat release rate (HRR) curves of SBI tests are obtained. The predicted HRR curves agree well with experimental SBI data especially for the first peak of the curve. The early phases of the test are of major importance in the determination of the FIGRA index and the classification of the product. Only a HRR curve from a single cone calorimeter test at 50 kW/m2 is needed as input data for the model.
The basic model works reasonably well for products with minor or moderate lateral flame spread in the SBI test. For these materials, the one-dimensionality of the model can be compensated with the selection of the input parameters. In the data set studied, the classification on the basis of the FIGRA index was predicted correctly for 89 % of the products .
During the development of the basic model, it was noted that the selection of input parameters is not optimal for certain groups of materials, for example fire retarded (FR) wood products. The resulting prediction inaccuracies can be reduced by optimising the input parameters separately for these products groups.
The modelling study of FR wood products included about 20 different products: impregnated or brush-applied FR wood materials, and special FR treated plywoods. In this product set, different heat release behaviour patterns in cone calorimeter tests were identified. The main patterns are introduced in Figure 7. Strongly FR impregnated materials release heat at a low level showing no sustained flaming (7a) or ignition after a long heat exposure time (7b). A third typical behaviour for this kind of products is slowly increasing heat release at a relatively low level (7c). Milder impregnations and brush-applied FR treatments typically result in even heat release at a moderate level (7d), or a heat release behaviour typical of also non-FR wood products showing two maxima with an intermediate plateau (7e). Some products exhibit a sharp peak in the very beginning, followed by even heat release level (7f) or the "non-FR wood behaviour".
Considering the one-dimensional thermal flame spread model presented above, the cases shown in Figs. 7a and 7b, and sometimes 7c, can be included to the basic model as special cases. If the heat release rate of the product does not reach 50 kW/m2 (the ignition criterion) within 570 seconds from the beginning of the heat exposure, the class prediction is B.
For products showing a heat release pattern of Figs. 7c, 7d, 7e or 7f, an optimisation of model input parameters was performed. It was found that the best prediction for HRR is obtained using w = 0.20 m as the pyrolysis width. Other input parameters were the same as in the basic model, that is, xp0 = 0.26 m, kf = 0.048, and n = 0.77. Furthermore, the ignition time, time scale and HRR values from cone calorimeter tests at the exposure level of 50 kW/m2 were scaled to lower levels describing the exposure conditions in SBI tests. The choice of exposure levels for scaling was a part of the optimization process of the model.
Assuming thermally thin behaviour, the ignition time is inversely is inversely proportional to the net heat flux . Thus, tig was scaled from heat exposure level of 50 kW/m2 to 30 kW/m2 as follows:
An analytical model for the charring rate of wood  was used as the basis of scaling the HRR curve to the exposure level of 25 kW/m2. The relationship between charring rate β (in mm/min) and an external heat flux (in kW/m2) is roughly
HRR is directly proportional to the charring rate. Thus, HRR at 25 kW/m2 is approximately 2/3 of the value measured at 50 kW/m2. Assuming that the whole specimen burns eventually and the total amount of heat produced is constant, the burning time (i.e. the time scale of the test) should be multiplied by 3/2, respectively.
Since the calculation procedure presumes that HRR used as the model input reaches 50 kW/m2, the original unscaled HRR measured at 50 kW/m2 must reach 75 kW/m2 to be applicable in the calculations. If HRR measured at 50 kW/m2 is less than 75 kW/m2 throughout the test, class B is predicted without calculations.
The procedure of using the one-dimensional thermal flame spread model for the prediction of Euroclasses of FR wood products is presented in Table 9.
Table 9. Phases of predicting SBI product classification for FR wood products.
Examples of HRR curves measured in SBI tests and predicted using the model for FR wood products are presented in Figure 8. The example products are a board with a brush-applied FR treatment and a special FR treated plywood. The predicted and measured FIGRA0,2MJ and FIGRA0,4MJ values for FR wood products are compared in Figure 9. As seen from these figures, the calculated HRR curves and FIGRA values are in good agreement with the experimental results.
It is emphasized that the predictive procedure developed is a non-physical model intended for engineering applications. Thus, it includes several approximations and simplifications. However, the model provides a practical tool for product development and quality control, since its use requires only a small amount of material and data from one small-scale test.
A progressive and illustrative tool for simulating fire development is the Fire Dynamics Simulator (FDS) program [56, 57]. It can be used for simulating systems of varying complexity, ranging from a simple small-scale test for a single material to whole buildings including different materials and structures. When predicting the fire performance of wood products in real-scale fires, the method works basically in two steps. Firstly, a cone calorimeter test is modelled using FDS and material parameters for the model are tuned so that the calculated heat release rate (HRR) curve agrees with the measured HRR. Secondly, the real fire scenario is calculated by using the material parameter values obtained in the first step.
An example of FDS simulation for 10 mm thick spruce linings in the ISO 9705 Room/Corner test is shown in Figure 10 . The calculated HRR curve agrees well with the measured HRR curve, which confirms that the extrapolation from a small-scale test to a full-scale scenario works well.
The same simulation method can be applied also to FR wood. 22-mm thick FR treated spruce timber was tested in the ISO 9705 room with a modified setup: instead of full coverage of the rear and side walls and the ceiling, only partial linings with a surface area of ca. 14 m2 were installed as shown in Figure 11. The measured and calculated HRR data are presented in Figure 12 . The comparison of the experimental and computational results clearly demonstrates that the methodology of using FDS simulation based on cone calorimeter HRR data can be used also for FR wood.