Solid recovered fuels in the cement industry – semi-automated sample preparation unit as a means for facilitated practical application

Abstract

One of the challenges for the cement industry is the quality assurance of alternative fuel (e.g. solid recovered fuel, SRF) in co-incineration plants – especially for inhomogeneous alternative fuels with large particle sizes (d₉₅⩾100 mm), which will gain even more importance in the substitution of conventional fuels due to low production costs. Existing standards for sampling and sample preparation do not cover the challenges resulting from these kinds of materials. A possible approach to ensure quality monitoring is shown in the present contribution. For this, a specially manufactured, automated comminution and sample divider device was installed at a cement plant in Rohožnik. In order to prove its practical suitability with methods according to current standards, the sampling and sample preparation process were validated for alternative fuel with a grain size >30 mm (i.e. d₉₅=approximately 100 mm), so-called ‘Hotdisc SRF’. Therefore, series of samples were taken and analysed. A comparison of the analysis results with the yearly average values obtained through a reference investigation route showed good accordance. Further investigations during the validation process also showed that segregation or enrichment of material throughout the comminution plant does not occur. The results also demonstrate that compliance with legal standards regarding the minimum sample amount is not sufficient for inhomogeneous and coarse particle size alternative fuels. Instead, higher sample amounts after the first particle size reduction step are strongly recommended in order to gain a representative laboratory sample.

Keywords

Alternative fuel solid recovered fuel (SRF)‘Hotdisc SRF’representative sampling sample preparation comminution quality assurance

Introduction

A continued increase in prices of conventional fossil fuels goes hand in hand with an increase in use of alternative fuels (i.e. solid recovered fuel, SRF) in the cement industry. The utilization of alternative fuel in co-incineration processes, however, requires an adequate quality assurance of SRF at the cement plant, which covers sampling, sample preparation and analyses.

Relevant influencing factors on the quality of alternative fuels

The quality of alternative fuel produced and delivered to the cement industry is mainly depending on two influencing factors. The first factor is the type and origin of input materials used in a SRF production plant (e.g. non-hazardous mixed household and similar industrial waste) and the second is the volume and intensity of waste processing technologies applied in the plant (cf. Lorber et al., 2011, 2012; Pomberger and Schmidt, 2005). SRFs of high quality with strict specifications (i.e. narrow particle size (d₉₅) range between 10 and 30 mm, and a net calorific value ⩾20 MJ kg⁻¹_DM) such as main burner fuel (MBF) require a great extent of processing technologies. Coarse alternative fuels with lower quality demands (i.e. net calorific value ⩽20 MJ kg⁻¹_DM, wider particle size (d₉₅) distribution between 80 and 300 mm) usually result from less equipped (i.e. without fine shredding steps) SRF production plants (Sarc et al., 2014). As the latter is less expensive and easier to produce, coarse alternative fuel will gain more importance in current and future substitution of conventional fuels. The quality of SRF is a crucial factor, because it highly affects the complexity of quality assurance procedure.

Standards and guidelines for quality assurance

A common basis for sampling, sample preparation and characterization of SRF is given in a series of European Standards (EN). However, the approaches described in these standards were designed for rather homogeneous materials with small grain sizes. Challenges resulting from inhomogeneous materials with large grain sizes are not covered by these norms.

Waste analysis consists of a chain of operations (sampling, sample preparation, analysis, evaluation of data) whereas each step contributes to the overall uncertainty (expressed as variance s²) of an analysis result (Flamme and Geiping, 2012). Although the analysis itself may be very precise, errors that occur during sampling or the sample preparation process will in any case falsify the results. This fact highlights the importance of a carefully considered sampling and sample preparation procedure.

One of the main factors in such procedures is the minimum sample amount. According to ÖNORM EN 15442 (ASI, 2011a), the minimum amount of a sample for an increment (m_m [kg]) is calculated as follows, with d₉₅ being the 95th percentile of particle size [mm] and λ_b being the bulk density of sample material [kg m⁻³]):

m = 2.7 \cdot 10^{- 8} \cdot d_{95}^{3} \cdot λ_{b}

(1)

A field sample consists generally of 6–10 increments. Because the minimum sample amount for the field sample is drastically reduced after particle size reduction and the minimum sample amount is directly related to the particle size, coarse grinding is always highly recommended. Equation 2 should be taken under consideration repeatedly in the course of sample preparation, in order to determine the minimum sample amount after each size reduction step, the formula terms m₁ and m₂ being the sample amount [kg] before and after the size reduction step, and d₁ and d₂ being the respective maximum particle size [mm]. The determination of the shape factors (s₁, s₂) for inhomogeneous materials are assumed to be 1 as a common approach (ASI, 2011b).

m_{1} \geq \frac{s_{1}}{s_{2}} \cdot {(\frac{d_{1}}{d_{2}})}^{3} \cdot m_{2}

(2)

Challenges in sampling and sample preparation for coarse alternative fuel

The present paper focuses on the challenges in terms of chemical quality monitoring of a coarse particle size alternative fuel, which are described by using the so-called Hotdisc fuel (HDF) as an example. The HDF is used at the cement production plant of Holcim in Rohožnik, Slovakia, additionally to MBF, which is the input material for the ‘main burner’. At the plant in Rohožnik, the cement kiln features a special precombustion chamber called HOTDISC®. The HOTDISC® precombustion system is an extensively discussed subject in FLSmidth (2013), Karrahe (2014), Lorber et al. (2011) and Pomberger (2008).

Besides compliance with national legal requirements in Slovakia, the owner of the cement plant in Rohožnik has also voluntarily imposed the stricter requirements of the Austrian Waste Incineration Ordinance (BMLFUW, 2010). The quality assurance according to the Austrian Waste Incineration Ordinance (BMLFUW, 2010) requires the monitoring of heavy metals and the net calorific value for both input streams (HDF and MBF) disclosed separately for all the suppliers (i.e. type and origin of SRF).

For a coarse alternative fuel such as HDF, the sampling and sample preparation process is especially challenging due to the inherent material inhomogeneity but even more due to the large particle size. Taking into account that the material is compressed in fully loaded trucks, the bulk density is estimated to be 250 kg m⁻³ at delivery (Sarc and Lorber, 2013). With this bulk density and d₉₅ being 100 mm, the minimum mass for an increment equals to 6.8 kg when applying equation 1. As a resulting amount for a field sample (consisting of at least six increments), about 41 kg can be assumed. According to ÖNORM EN 15443 (ASI, 2011b), the entire sample amount has to be comminuted before being reduced. In practice, a sample amount of 41 kg (or c. 165 litres) is very difficult to handle for the plant personnel, particularly when there is no shredder for coarse grinding available at the site.

This challenging initial situation was subject matter of a project in cooperation of Holcim Slovensko and the Chair of Waste Processing Technology and Waste Management at the Montanuniversitaet Leoben. The project’s main objective was the development of a sample preparation procedure that could easily be handled by plant personnel, could perfectly be integrated into the company’s daily routine and guaranteed adequate analysis results for monitoring purposes. The sampling was excluded from the process, as a fully automated sampling device is almost impossible to realize due to the physical properties of the material (e.g. hardness, humidity). The sample preparation system should be largely automated to minimize possible human errors, facilitate the handling and allow a fast preparation process to cope with the high number of daily truck deliveries for SRF. Good reproducibility and stable representativeness were also essential requirements. In this contribution, the newly installed system at the cement plant in Rohožnik for the preparation of coarse alternative fuel is described and the results of the conducted validation of the sampling and sample preparation process are presented.

Materials and methods

Previous sampling procedure of HDF

Prior to installation of the small shredding plant and implementation of the adapted sampling procedure, the sampling for HDF (cf. Figure 1) at the Holcim plant at Rohožnik was done manually according to the instructions laid out in the legal requirements of the Austrian Waste Incineration Directive (BMLFUW, 2010) and standards such as ÖNORM EN 15442 (ASI 2011a), ÖNORM EN 15443 (ASI, 2011b) and ÖNORM EN 15413 (ASI, 2011c).

Sampling took place in front of the storage bunkers, where the HDF was delivered by walking floor trucks. From standpoint of logistics, this was the easiest way to establish the direct retraceability to every supplier, as each truck could be allocated accordingly. The samples were taken directly from the truck by plant personnel during the unloading process. To cover the whole truck content, samples were taken from different locations (spots) across the total truck length. The individual samples were then merged into a collective sample in a rubbish bin and afterwards transported to the plant’s own analytical laboratory. There, the complete sample was dried at 40°C and afterwards comminuted using a slow-running cutting mill (SM 2000; Retsch) for coarse shredding. Subsequently, the sample amount was reduced applying the coning and quartering method and thereafter a cutting mill for high-speed comminution (Pulverisette 18, Fritsch) was applied in order to gain a representative laboratory sample. Results from the continuous monitoring throughout the past year, presented in this study as reference values, were obtained using X-ray fluorescence analysis (XRF) at the plant’s own analytical laboratory.

Description of the newly implemented sampling procedure

As the previous approach was rather time consuming mainly because of the comminution and drying process to be applied for very large sample amount, a new route with fully automated comminution techniques was aspired. In June 2014, an automatic device for comminution and sample amount reduction was installed close to the storage bunkers in the SRF delivery area at the Holcim plant in Rohožnik. The device is a small shredding plant, consisting basically of two conveyer belts (b, e), a single-shaft-shredder with a punched sieve segment (d) and a rotary sample divider (f) (cf. Figure 2). Sampling itself is still carried out manually by plant personnel and the samples are taken directly from the truck during the unloading process. The individual samples taken from different spots of the truck load are then integrated into a collective sample in a rubbish bin. The rubbish bin has a volume of 240 litres, which corresponds to approximately 34 kg of SRF (bulk density of loosely piled material: 140 kg m⁻³ according to Sarc and Lorber, 2013), a little bit less than the necessary theoretical amount for a material with this particle size according to the Eqs 1–2 presented in the introduction. Immediately after sampling, the bin is then transferred to the rubbish bin tipping device (a) by plant personnel (cf. Figure 2). After that, the comminution and sample amount reduction are carried out automatically. In the next step, the rubbish bin is lifted overhead and emptied on the feeding conveyer belt (Erdwich, Feeding-conveying belt H400). A magnetic separator (c) on top of the belt removes metallic pieces, which are collected in a separate bin (cf. Figure 2). Subsequently, the Hotdisc material falls into the funnel of a single-shaft-shredder (Erdwich, M600/1-600, motor power 5.5 kW, cutter diameter 220 mm, cutter width 25 mm, number of cutters 25). A perforated sieve segment integrated in the shredder unit guarantees a maximum grain size (d₉₅) of 20 mm. The comminuted sample falls onto the second conveyer belt (Erdwich, Discharge-conveying belt H300) and is transported to the rotary sample divider (Erdwich, Rotary sample divider DRV 75). An essential part of the rotary sample divider is a time-controlled clapper that allows for the sample amount to be reduced to approximately one-tenth of its original amount, which equals approximately 6.8 kg (‘laboratory sample’) and can easily be transported in a bucket to the analytical laboratory, where the sample is dried at 40°C and further comminuted in a high-speed cutting mill (Pulverisette 18, Fritsch). The rest of the material is discarded. Applying equation 2 shows that the theoretical minimum sample amount of 0.3 kg is easily met for the conditions at hand (d₁ being 20 mm, d₂ being 100 mm and m₂ being 41 kg).

Validation of the newly implemented sampling procedure

Once the new sample preparation and dividing system was implemented, the whole device had to be validated in order to establish its functional capability. An overview of the validation steps is given in Figure 3.

One truck of a regular supplier was randomly picked and two independent collective samples (collective sample 1 and collective sample 2) were gathered by sampling the different areas (spots) of the whole truck. These samples were consecutively subjected to the automated comminution and size-reduction system. After completing the process, collective sample 1 (CS 1) resulted in a laboratory sample (LS1) and a parallel sample (PS1), which consisted of the rest of the material and would normally get discarded. The laboratory sample was homogenized and divided into five sub-samples, namely LS1.1, LS1.2, LS1.3, LS1.4 and LS1.5. The same procedure was used on the parallel sample resulting in five sub-samples (PS1.1, PS1.2, PS1.3, PS1.4 and PS1.5, respectively). All sub-samples were separately handled during the subsequent preparation and analytical procedure in the accredited laboratory of the Chair of Waste Processing Technology and Waste Management at the Montanuniversitaet Leoben. The samples were dried at 40°C, in parallel, and the dry mass and water content according to ÖNORM CEN/TS 15414-1 (ASI, 2010) were determined. Subsequently, the samples were comminuted to analytical fineness (<0.5 mm) using a cutting mill (Pulverisette 18, Fritsch) after sorting out of extraneous materials. A final grain size of <0.5 mm for the analysis sample was considered particularly important, based on the recommendations of several studies (Denner 2009; Denner and Kügler, 2006). The net calorific value as well as the gross calorific value, respectively, were determined according to ÖNORM EN 15400 (ASI, 2011d), the elements chlorine and sulphur were analysed using ion chromatography after calorimetric digestion in compliance with ÖNORM EN 15408 (ASI, 2011e). The selected metals Sb, As, Pb, Cd, Cr, Co, Ni, Hg, Zn, Sn, V, Tl and Cu were also determined, using mass spectrometry with inductively coupled plasma (ICP-MS) after digestion in a mixture of HNO₃, HCl and HF according to ÖNORM EN 15411 (ASI, 2011f). All analysis procedures were done in duplicate for every sample.

The same sample preparation and analytical procedure was applied to collective sample 2 (CS 2), resulting in five laboratory sub-samples LS2.1, LS2.2, LS2.3, LS2.4 and LS2.5, and five parallel sub-samples PS2.1, PS2.2, PS2.3, PS2.4 and PS2.5, respectively.

In addition to the 20 sub-samples, two composite samples were analysed. The first composite sample (COMS 1) was prepared from the analytically fine (i.e. d₉₅ <0.5 mm) sub-samples LS1.1, LS1.2, LS1.3, LS1.4 and LS1.5, whereas the second composite sample (COMS 2) was prepared from the sub-samples of the second laboratory sample (LS2.1, LS2.2, LS2.3, LS2.4 and LS2.5). All analytical parameters mentioned above with the exception of dry matter were also analysed separately for the two composite samples.

Results and discussion

Determination of accuracy for the newly implemented procedure

The main objective of sampling and the sample preparation process including the primary sample through to analysis sample is to obtain representative samples. A basic statistical evaluation of sampling operations always comprises the aspects of precision and accuracy. The latter is difficult to verify as establishment and definition of the ‘true value’ is rather challenging for an inhomogeneous material such as HDF with large grain sizes, as can be seen from the high yearly standard deviation (cf. Table 1). For this reason, the authors chose to use the yearly average values as well as the yearly minimum and maximum values as a reference for the ‘true value’. Table 1 shows the results from monthly monitoring, which were obtained by the manual method according to ÖNORM EN 15442 (ASI, 2011a), ÖNORM EN 15443 (ASI, 2011b) and ÖNORM EN 15413 (ASI, 2011c) without usage of the newly implemented shredding plant. In addition, the table contains the mean value of the entire dataset as well as the yearly maximum and minimum values. The results for composite samples 1 and 2 (COMS 1 and COMS 2) obtained by the new procedure described in this paper were compared with this data, as the composite samples would be the actual laboratory sample to be analysed in the laboratory. A composite sample is representative for the total amount of the laboratory sample because it was blended after the comminution to fineness d₉₅⩽0.5 mm of all individual laboratory sub-samples. The comparison of yearly average, yearly maximum and minimum values with the results for the composite samples are shown graphically in Figure 4. It can distinctly be seen, that the analysis results for the composite samples lie within the yearly maximum and minimum values for all parameters and are in good accordance with the respective yearly average value too. The reported sampling in combination with the newly developed automated sample preparation procedure can be considered a suitable approach to deliver representative results for coarse grain size SRF. The monthly monitoring comprises only the parameters laid down in the Austrian Waste Incineration Ordinance. Hence, only eight heavy metals and the net calorific value were available for comparative purposes.

Table 1.

Comparison of the results for consecutive sample 1 and 2 to yearly reference values for the chosen supplier’s material.

Sample	Sb	As	Pb	Cd	Co	Cr	Ni	Hg	Hu
Sample	mg kg⁻¹_DM								MJ kg⁻¹_DM
RS 1	18	<0.25	29	2.2	16	79	23	0.27	14.5
RS 2	48	<0.25	221	1.3	6.4	58	19	0.45	16.5
RS 3	97	<0.25	105	1.5	8.9	92	8.9	0.32	21.7
RS 4	64	1.7	153	1.0	4.7	53	8.4	0.51	16.4
RS 5	52	3.1	152	1.1	8.8	92	18	0.24	16.0
RS 6	29	<0.25	222	5.2	3.8	25	7.0	0.15	20.2
RS 7	7.4	0.3	166	4.8	15	137	25	0.23	20.2
RS 8	71	1.4	113	3.7	18	54	34	0.54	16.7
RS 9	140	1.6	296	2.1	12	94	22	0.43	14.8
RS 10	12	<0.25	88	0.61	4.7	31	10	0.20	18.7
RS 11	25	4.5	226	4.5	8.1	64	18	0.45	14.9
RS 12	104	0.5	106	1.3	6.6	37	14	0.47	22.0
Yearly average value	56	1.6	157	2.4	9.4	68	17	0.36	17.7
Yearly standard deviation	41	1.5	74	1.6	4.9	32	8.1	0.13	2.7
Yearly maximum value	140	4.5	296	5.2	18	137	34	0.54	22.0
Yearly minimum value	7.4	<0.25	29	0.61	3.8	25	7.0	0.15	14.5
COMS 1	80	3.1	230	4.3	5.5	96	23	0.18	17.3
COMS 2	41	4.3	200	3.4	9.0	50	20	0.11	17.0

RF, reference sample; COMS, composite sample; DM, dry matter.

Verification of reproducibility

For the validation process, the examinations of two independent samples (samples 1 and 2) were carried out. These two selected samples underwent the same sampling process and were both comminuted with the newly implemented shredding device. Therefore, each of those samples should be a representative sample for the sampled truck and should deliver the same analysis results. The selection of parameters for analysis was mainly based on minimum parameters for classification, according to BS EN 15359 (BSI, 2011). These include limit values and a classification system for a technical (combustion process) parameter (chlorine), an economic relevant parameter (net calorific value) and an environmental parameter (mercury). The water content is also considered an economic attribute, whereas other metals were classified as environmentally relevant WRAP, 2014). As can be seen from Figure 4, as well as Table 1, the results for the composite samples 1 and 2 are matching well.

Table 2 lists the calculated mean values for all analysed sub-samples. For this purpose, the results from the laboratory sub-samples and parallel sub-samples were combined to include potential outliers. As can be seen from the results, sample 1 and sample 2, again, only differ slightly from each other. Copper, however, represents the only noticeable exception which is not surprising as copper is one of those elements that show a very heterogeneous distribution in waste due to its predominate presence as metallic species (EC, 2002).

Table 2.

Mean values of laboratory and parallel samples for sample 1 and sample 2 for all analysed parameters.

Sample name	Sb	As	Pb	Cd	Cr	Co	Ni	Hg	Zn
Sample name	mg kg⁻¹_DM
Sample 1: Mean value	76	3.5	189	5.8	80	5.4	23	<0.25	364
Sample 2: Mean value	64	5.5	188	2.6	64	4.2	22	<0.25	360
Sample name	Sn	V	Tl	Cu	Cl	S	Hu	Ho	WC
Sample name	mg kg⁻¹_DM						MJ kg⁻¹_DM		w%_OM
Sample 1: Mean value	17	8.7	<0.25	677	9040	5358	16.8	18.2	16.2
Sample 2: Mean value	27	6.1	<0.25	168	7475	5218	17.3	18.7	16.0

DM, dry matter; OM, original matter; WC, water content.

A conclusion that can be drawn from the data presented so far in this article is that the amount of a rubbish bin (i.e. 240 l, corresponding to 41 kg) to be taken from the truck and treated in the shredding plant is enough to be representative for coarse grain size SRF because average results are quite similar.

Check for potential malfunction of the comminution unit

One of the major drawbacks of a shredding plant having an automatic sample divider system would be any kind of physical separation or segregation processes as well as non-representative enrichment of certain sample portions or components of the material flow throughout the comminution plant. That this fact does not apply for the implemented shredding plant in Rohožnik is shown in Tables 3 and 4, where the mean value for all laboratory sub-samples and the mean value for all parallel sub-samples as well as the respective standard deviation of sample 1 are presented. The same was calculated for sample 2 (cf. Tables 3 and 4). As shown, the results for the mean values of the laboratory sample and parallel sample are very similar for all parameters except copper, which confirms that the generated laboratory sample is representative for the whole material after coarse grinding and sample amount reduction. The matching of laboratory sample and parallel sample also indicate that no significant enrichment or segregation processes occur in the shredding plant. This fact is supported by a visual check for any residue material left in the comminution device. The visual check was done after finishing the preparation process for each of the samples.

Table 3.

Results for all laboratory and parallel sub-samples as well as their respective mean values for Sb, As, Pb, Cd, Cr, Co, Ni, Hg and Zn.

Sample	Sb	As	Pb	Cd	Cr	Co	Ni	Hg	Zn
Sample	mg kg⁻¹_DM
LS1.1	39	3.1	210	9.7	140	7.1	35	<0.25	440
LS1.2	110	3.4	110	2.4	75	5.2	21	<0.25	530
LS1.3	25	3.1	250	3.8	72	5.0	23	<0.25	600
LS1.4	116	3.1	180	10	94	5.0	20	<0.25	290
LS1.5	93	4.1	100	3.9	85	5.4	25	<0.25	340
LS 1 Mean value	77	3.4	170	6.0	93.1	5.5	25	<0.25	440
LS 1 Standard deviation	42	0.4	64	3.6	28	0.9	6.0	–	129
LS 1 Relative standard deviation [%]	54.5	11.8	37.6	60.0	30.1	16.4	24.0	–	29.3
PS1.1	75	3.6	120	8.9	58	4.6	21	<0.25	310
PS1.2	13	3.5	250	2.0	79	5.0	21	<0.25	290
PS1.3	110	3.5	230	4.4	78	8.8	33	<0.25	330
PS1.4	130	3.4	190	1.8	58	4.0	19	<0.25	260
PS1.5	44	4.4	250	11	61	4.1	16	<0.25	250
PS 1 Mean value	74	3.7	208	5.6	67	5.3	22	<0.25	288
PS 1 Standard deviation	48	0.4	55	4.1	11	2.0	6.5	–	33
PS 1 Relative standard deviation [%]	64.9	10.8	26.4	73.2	16.4	37.7	29.5	–	11.5
LS2.1	39	3.4	190	1.7	54	3.9	21	<0.25	470
LS2.2	22	6.6	440	6.4	59	7.0	20	<0.25	520
LS2.3	110	15	240	1.5	55	4.0	21	<0.25	340
LS2.4	99	3.9	130	7.6	62	4.3	31	<0.25	370
LS2.5	60	3.3	90	1.0	48	3.9	21	<0.25	300
LS 2 Mean value	66	6.4	218	3.6	56	4.6	23	<0.25	400
LS 2 Standard deviation	38	5.0	137	3.1	5.3	1.3	4.6	–	92
LS 2 Relative standard deviation [%]	57.6	78.1	62.8	86.1	9.5	28.3	20.0	–	23.0
PS2.1	130	3.9	210	2.1	96	4.2	42	<0.25	330
PS2.2	17	6.3	120	0.9	50	3.5	19	<0.25	380
PS2.3	120	3.8	120	0.65	140	3.6	21	<0.25	300
PS2.4	18	4.8	120	2.7	40	3.8	15	<0.25	270
PS2.5	20	4.4	220	1.7	40	3.6	14	<0.25	320
PS 2 Mean value	61	4.6	158	1.6	73	3.7	22	<0.25	320
PS 2 Standard deviation	59	1.0	52	0.8	44	0.3	11	–	41
PS 2 Relative standard deviation [%]	96.7	21.7	32.9	50.0	60.3	8.1	50.0	–	12.8

LS, laboratory sample; PS, parallel samples; DM, dry matter.

Table 4.

Results for all laboratory and parallel sub-samples as well as their respective mean values for Sn, V, Tl, Cu, Cl, S, Hu, Ho and water content (WC).

Sample	Sn	V	Tl	Cu	Cl	S	Hu	Ho	WC
Sample	mg kg⁻¹_DM						MJ kg⁻¹_DM		w%_OM
LS1.1	10	9.9	<0.25	2040	11600	4300	17.3	18.8	16.4
LS1.2	16	10	<0.25	3510	7980	5070	17.0	18.4	16.6
LS1.3	11	11	<0.25	660	7300	5500	17.0	18.0	14.2
LS1.4	20	8.6	<0.25	97	9590	5070	16.0	17.0	15.5
LS1.5	35	10	<0.25	180	7040	4870	15.4	16.8	16.8
LS 1 Mean value	18	9.9	<0.25	1297	8702	4962	16.5	17.8	15.9
LS 1 Standard deviation	10	0.9	–	1461	1900	436	0.8	0.9	1.1
LS 1 Relative standard deviation [%]	55.5	9.1	–	112.6	21.8	8.8	4.8	5.1	6.9
PS1.1	19	7.5	<0.25	52	8870	5480	17.1	18.6	16.8
PS1.2	15	6.8	<0.25	54	9640	6020	17.2	18.7	16.4
PS1.3	11	7.5	<0.25	46	9090	5900	17.7	19.2	16.9
PS1.4	13	6.2	<0.25	63	9520	5660	16.3	17.7	16.5
PS1.5	17	9.2	<0.25	63	9770	5710	17.3	18.8	15.8
PS 1 Mean value	15	7.4	<0.25	56	9378	5754	17.1	18.6	16.5
PS 1 Standard deviation	3.2	1.1	–	7.4	382	211	0.5	0.6	0.4
PS 1 Relative standard deviation [%]	21.3	14.9	–	13.2	4.1	3.7	2.9	3.2	2.4
LS2.1	17	5.8	<0.25	650	6600	5680	17.9	19.0	14.9
LS2.2	14	8.2	<0.25	260	9890	6040	17.9	19.4	15.0
LS2.3	16	6.4	<0.25	95	8400	5210	16.2	17.6	16.2
LS2.4	72	6.9	<0.25	140	9640	5960	18.0	19.6	16.5
LS2.5	13	6.0	<0.25	140	7270	4670	17.0	18.5	16.5
LS 2 Mean value	26	6.7	<0.25	257	8360	5512	17.4	18.8	15.8
LS 2 Standard deviation	26	1.0	–	228	1438	572	0.8	0.8	0.8
LS 2 Relative standard deviation [%]	100.0	14.9	–	88.7	17.2	10.4	4.6	4.3	5.1
PS2.1	86	5.7	<0.25	71	6300	5060	17.6	19.2	16.9
PS2.2	16	6.3	<0.25	67	6180	4800	16.1	17.5	16.0
PS2.3	10	4.9	<0.25	70	7040	4760	18.0	19.5	16.1
PS2.4	12	5.8	<0.25	95	7590	5150	16.6	18.1	16.2
PS2.5	15	5.2	<0.25	88	5840	4850	17.4	18.9	16.1
PS 2 Mean value	28	5.6	<0.25	78	6590	4924	17.1	18.6	16.3
PS 2 Standard deviation	33	0.5	–	12	710	171	0.8	0.8	0.4
PS 2 Relative standard deviation [%]	117.9	8.9	–	15.4	10.8	3.5	4.7	4.3	2.5

LS, laboratory sample; PS, parallel samples; DM, dry matter; OM, original matter.

In addition to the good agreement of mean values for laboratory samples and parallel samples, the standard deviations for these samples are quite similar for most parameters such as Sb, Cd, Co, Ni, Sn, V, S, Hu, Ho and the water content. This indicates that each sample represents and covers the total truckload composition very well. These findings show that the chosen technique is highly reproducible and the extracted amounts for the laboratory sample are sufficient.

Determination of minimum amount for laboratory sample

In addition to the representative mean values and the respective standard deviation, the results of all analysed parameters for every laboratory and parallel sub-samples are shown in Tables 3 and 4. While the individual results for the net calorific value as well as the gross calorific value show little relative deviation from the mean value with a maximum of 8% as depicted in Figure 5, the results for copper display a very high degree of fluctuation of up to 400%. Figure 5 also illustrates the results for cadmium: similar to those for antimony, arsenic and tin, which all show a maximum deviation between 100% and 200%. The fluctuation range for lead, nickel and cobalt is a little bit less with 80–100%, whereas it is only 15–50% for zinc, vanadium, sulphur and chlorine as well as the parameter water content. The graphical presentation for nickel and chlorine is shown in Figure 5 as an example for the other parameters mentioned. These high deviation ranges clearly show that required representativeness would not be reached, if the designated laboratory sample is divided into sub-samples directly after the coarse shredding plant and before further comminution, even when it complies with the theoretical minimum sample amount of 0.3 kg, calculated with equation 2. The representative amount is, in fact, even larger with an actual minimum weight of approximately 1.3 kg. This shows very clearly, that for a coarse alternative fuel such as HDF, larger amounts of samples are required for representative results. Therefore, it is essential toprocess the whole amount of laboratory sample, i.e. the entire laboratory sample is to be taken to the laboratory for drying and further fine grinding before coning and quartering again.

Statistical deviation attributable to material inhomogeneity

The selection of a certain sample amount and subsequent comminution is one of the crucial influencing factors in sample representativeness, which is also supported by the results in Figure 6. It illustrates the relative deviation of experimentally determined results for the composite samples 1 and 2 from the calculated mean of all investigated laboratory and parallel sub-samples. This comparison shows, that the homogeneity and representativeness of a sample improves with additional comminution of the material, as was done with the composite samples, and correlates well with statistical data of a higher amount (i.e. in [kg]) of samples. For most parameters, the deviation adds up to only 20–30%; for others, it is a little bit higher with up to 50–60% but these figures are still considered good results for a heterogeneous material such as HDF. Once more, the only exception is copper, with a relative deviation of up to almost 600%, but this result is, again, attributable to the inhomogeneous distribution of copper in waste, which is notorious. The results for mercury and thallium were below the limit of detection (i.e. 0.25 mg kg⁻¹) and are therefore not shown in Figure 6.

Conclusions

The use of inhomogeneous alternative fuels with large particle sizes (d₉₅⩾100 mm) (HDF) will gain more importance in the cement industry due to their low production cost. The quality assurance of these materials is, however, rather challenging and not sufficiently covered by existing standards. In the present article, a sound approach for the quality assurance of HDF is presented which was also extensively validated. The following aspects can be concluded from this research:

The sample amount equivalent to the volume of a rubbish bin (240 litres) for initial sampling was proven to be sufficient to obtain representative results for a truckload (up to c. 80 m³).

The newly implemented sample comminution and division device has been proven to be an adequate and suitable installation for the sample preparation process, as material handling is considerably facilitated for the plant personnel. The reported device, on one hand enabled the processing of the entire minimum sample amount that constitutes 41 kg per truckload. On the other hand, an easily manageable laboratory sample of approximately 6–7 kg is provided by further comminution followed by automatic sample dividing.

The results for two independent laboratory samples (composite samples 1 and 2) correlate well with the yearly average results and no significant outliers were obtained. This suggests that this new approach allows the generation of reasonable and representative results.

The validation of this innovative procedure also showed that segregation and enrichment of certain sample portions or components of the treated HDF through the comminution device do not occur. This was verified by visual checks as well as a comparison of results (mean value and standard deviation) obtained for the laboratory sample and those of a parallel sample made out of remaining excess material that would be discarded in practice.

The minimum sample amount after coarse grinding, calculated according to ÖNORM EN 15443 (ASI, 2011b), is not sufficient for an inhomogeneous material such as HDF. The findings obtained in the validation process described show that a 10-fold amount is necessary to ensure representative results of the material under investigation.

Footnotes

Acknowledgements

The authors are very grateful to Holcim (Slovensko), especially to Dr. Ernst-Michael Sipple, for enabling the research project on HDF, and to Ms Maria Reznakova for on-site assistance in all matters.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

References

Austrian Standards Institute (ASI) (2010) Solid recovered fuels – Determination of moisture content using the oven dry method. Part 1: Determination of total moisture by a reference method, ÖNORM CEN/TS 15414–1. Vienna.

Austrian Standards Institute (ASI) (2011a) Solid recovered fuels – Methods for sampling, ÖNORM EN 15442. Vienna.

Austrian Standards Institute (ASI) (2011b) Solid recovered fuels – Methods for the preparation of the laboratory sample, ÖNORM EN 15443. Vienna.

Austrian Standards Institute (ASI) (2011c) Solid recovered fuels – Methods for the preparation of the test sample from the laboratory sample, ÖNORM EN 15413. Vienna.

Austrian Standards Institute (ASI) (2011d) Solid recovered fuels – Determination of calorific value, ONORM EN 15400. Vienna.

Austrian Standards Institute (ASI) (2011e) Solid recovered fuels – Methods for the determination of sulphur (S), chlorine (Cl), fluorine (F) and bromine (Br) content, ÖNORM EN 15408. Vienna.

Austrian Standards Institute (ASI) (2011f) Solid recovered fuels – Methods for the determination of the content of trace elements (As, Ba, Be, Cd, Co, Cr, Cu, Hg, Mo, Mn, Ni, Pb, Sb, Se, Tl, V and Zn), ÖNORM EN 15411. Vienna.

British Standards Institute (BSI) (2011) Solid recovered fuels – Specifications and classes, BS EN 15359. London.

Bundesministerium für Land-und Forstwirtschaft, Umwelt und Wasserwirtschaft (BMLFUW) (2010) Verordnung über die Verbrennung von Abfällen. Abfallverbrennungsverordnung (AVV) (‘Waste Incineration Directive’). Vienna, Austria.

10.

Denner

(2009) Ermittlung der Einflüsse von alternativen chemisch-analytischen Aufarbeitungsverfahren auf die Bewertung von festen Ersatzbrennstoffen „EBS-Analytik” (Identification of influencing effects of alternative chemical-analytical preparation techniques on evaluation of solid recovered fuels‚ ‘SRF analytics’) – Final report REP-0125. Austria: Environmental Protection Agency.

11.

Denner

Kügler

(2006) Erarbeitung eines Beprobungskonzeptes für Ersatzbrennstoffe (Development of a sampling concept for solid recovered fuels) Project ERSA – Final report REP-0059). Austria: Environmental Protection Agency.

12.

European Commission (EC) DG ENV. E3, Project ENV.E.3/ETU/2000/0058 (2002) Heavy metals in waste, Final Report. COWI A/S, Denmark.

13.

Flamme

Geiping

(2012) Quality standards and requirements for solid recovered fuels – a review. Waste Management & Research 30: 335–353.

14.

FLSmidth (2013) HOTDISC^TM combustion device, C 03–11 400–13-ENG. Available at: www.flsmidth.com (last accessed: 2 January 2015).

15.

Karrahe

(2014) FLSmidth HOTDISC^TM – a perfect combustion device for coarser alternative fuels. In: Zement, Kalk, Gips international (ZKG international), Issue 6, ISSN: 0949–0205.

16.

Lorber

Sarc

Aldrian

(2012) Design and quality assurance for solid recovered fuel. Waste Management & Research 30: 370–380.

17.

Lorber

Sarc

Pomberger

(2011) Herstellung und Einsatz von Ersatzbrennstoffen (EBS) in Österreich (Production and application of refuse derived fuels in Austria). In: Kühle-Weidemeier (ed.) Waste-to-Resources 2011, 4. Internationale Tagung MBA und Sortieranlagen. Göttingen: Cuvillier Verlag.

18.

Pomberger

(2008) Entwicklung von Ersatzbrennstoff für das HOTDISC-Verfahren und Analyse der abfallwirtschaftlichen Relevanz (Development of SRF for the HOTDISC-process and evaluation of its importance for waste management). PhD thesis at IAE-Institute, Montanuniversitaet Leoben.

19.

Pomberger

Schmidt

(2005) Betriebserfahrungen in der Ersatzbrennstoffproduktion und verwertung am Beispiel ThermoTeam (Operating experience in production and recovery of solid recovered fuels by the example ThermoTeam). In: Thomé-Kozmiensky

(ed.): Ersatzbrennstoffe 5 – Herstellung und Verwertung. Neuruppin: TK Verlag.

20.

Sarc

Lorber

(2013) Production, quality and quality assurance of refuse derived fuels (RDFs). Waste Management 33: 1825–1834.

21.

Sarc

Lober

Pomberger

Rogetzer

Sipple

(2014) Design, quality and quality assurance of solid recovered fuels (SRF) for the substitution of fossil feedstock in the cement industry. Waste Management & Research 32(7): 565–585.

22.

Waste & Resources Action Programme (WRAP) (2014) A classification scheme to define the quality of waste derived fuels. Available at: http://www.wrap.org.uk/sites/files/wrap/WDF_Classification_6P%20pdf.pdf (Last accessed: 2 January 2015).