Optimisation and Modelling of Anaerobic Digestion of Whiskey Distillery / Brewery Wastes after Combined Chemical and Mechanical Pre-Treatment

: Whiskey distillery waste streams consisting of pot ale (liquid residue) and spent grain (solid residue) are high strength organic wastes and suitable feedstock for anaerobic digestion (AD) from both economic and environmental stand points. Anaerobic digestion of pot ale and pot ale / spent grain mixtures (with mixing ratios of 1:1, 1:3, and 1:5 by wet weight) was performed after implementation of a novel hybrid pre-treatment (combined chemical and mechanical) in order to modify lignocellulosic structure and ultimately enhance digestion yield. Lignin, hemicellulose, and cellulose fractions were determined before and after chemical pre-treatment. E ﬀ ects of di ﬀ erent inoculum rates (10–30–50% on wet basis) and beating times (0–7.5–15 min) on anaerobic digestion of pot ale alone and of pot ale / spent grain mixtures were investigated in lab scale batch mode with a major focus of optimising biogas yield by using response surface methodology (RSM) in Design Expert Software. The highest biogas yields of 629 ± 8.5 mL / g vs. (51.3% CH 4 ) and 360 ± 10 mL / g vs. (55.0 ± 0.4) with anaerobic digestion of pot ale alone and spent grain mix after 1M NaOH and 7.5 min beating pre-treatments with 50% inoculum ratio respectively. The optimum digestion conditions to maximise the biogas quality and quantity were predicted as 10 and 13 min beating times and 32 and 38 ◦ C digestion temperatures for anaerobic digestion of pot ale alone and spent grain mix respectively. M.C., C.C., and J.L.; Methodology, B.G., K.B., J.S., P.D., C.C., and J.L.; Project administration, J.L.; Resources, C.C.; Supervision, J.S., P.D., C.C., and J.L.; Visualization, K.B.; Writing—original draft, B.G.; Writing—review and editing, J.S., P.D., C.C., and J.L. All authors


Introduction
Widespread usage of fossil fuels as non-renewable energy source has been targeted for reduction because of many ecological and environmental concerns for impact on climate conditions and human migration due to continuous greenhouse gas emissions such as CO 2 [1,2]. Therefore, the European Union aims to reduce the total greenhouse emissions to 80-95% of 1990 levels by 2050 in developed countries [3]. In addition, European Council Directive 999/31/EC aims to reduce landfilling of organic waste [4]. As a result of increasing energy demand as well as environmental concerns, exploration

Sample Characterisation
Pot ale samples were initially characterised in terms of solid content; total solids (TS), volatile solids (VS), and moisture content (MC) in accordance with [46]. The concentrations of phosphorous (Molybdovanate Hach Method 8114) and nitrogen in forms of nitrate (Cadmium Reduction Hach Method 8039), nitrite (Ferrous Sulphate Hach Method 8153) were determined by using Hach DR 2000 Spectrometer and ammonia (Salicylate Test N Tube Hach Method 10031) was measured by using Hach DR 900 Spectrometer as part of sample characterisation. The concentrations of copper, iron, magnesium, manganese, and zinc of the pot ale were determined using ICP-OES (Agilent 5100).
Hemicellulose, cellulose, and lignin contents of pot ale and spent grain were measured according to the detergent method developed by [47,48] prior to and after the implementation of each pre-treatment for evaluation of amendment of lignocellulosic structure with regard to the applied pre-treatment to enhance AD yield.
Chemical oxygen demand (COD), biological oxygen demand (BOD), and sulphate concentration were measured prior to and after AD in order to evaluate the efficiency of organic matter removals and bacterial competition between methanogenic and sulphate reducing bacteria as well as for characterising pot ale as received. COD and sulphate concentration were measured according to Hach Method 8000 by using standard Hach high range plus vials (0-15,000 ppm) and USEPA Sulfa-Ver 4 Hach Method 8051 provided by Hach Lange Company respectively, and DR 2000 spectrometer was used for the measurement. BOD was measured using the Lovibon BD600 unit provided by Lennox. To prevent the risk of sludge masking on organic content removals, the supernatant of all samples was analysed in the form of a mix of sludge after centrifuging at 10,000 rpm for 30 min using a Sorvall RC 5B Plus centrifuge [49,50]. Similarly, concentrations of acetic acid, propionic acid, isobutyric acid, butyric acid, isovaleric acid and valeric acid were determined using gas chromatography (Agilent 7890) with an Agilent CP 7686 column, in order to calculate total short chain volatile fatty acids in substrate (pot ale) for sample characterisation as well as substrate and inoculum mixture prior to and after AD to assess potential VFA inhibition [51].

Hollander Beater Specifications
The Hollander Beater specifications were explained in detail by [52], having been shown to be an excellent tool for pre-treatment of solid wastes prior to anaerobic digestion. The beater can perform a cutting action as a result of rotation of the bladed drum and a high-pressure homogenisation action caused by an inclined plate located at the exit of the drum concurrently. Therefore, the Hollander beater combines the effects of particle size reduction and substrate homogenization which is essential for pre-treatment of high solid content biomass.

Bioreactor Set-Up
A lab scale batch system was used as anaerobic digester. The bioreactors consist of 2 main parts which are 500 mL conical flasks with an active working volume of 400 mL, and the 5 L capacity biogas compatible sampling bag ( Figure 1). The connection in between those parts is provided by plastic tubing, quick release tubing connectors, and three-way valves. The system was flushed with nitrogen twice for 2 min to ensure anaerobic conditions [53,54]. Water baths were employed to keep the reactors at a constant mesophilic reaction temperature. A biogas analyser, BIOGASS 5000, was used for the qualitative and quantitative analysis of the generated biogas. The biogas quality and quantity was measured based upon DIN 2006 standard 38 414-8 explained by [54] which is commonly used for biomethane potential tests (BMP) at lab and pilot scales. The residence time was 21 days due to significant decrease in biogas generation (less than 1% of the previous measurement) after that time period.
Processes 2020, 8, x FOR PEER REVIEW 4 of 30 reactors at a constant mesophilic reaction temperature. A biogas analyser, BIOGASS 5000, was used for the qualitative and quantitative analysis of the generated biogas. The biogas quality and quantity was measured based upon DIN 2006 standard 38 414-8 explained by [54] which is commonly used for biomethane potential tests (BMP) at lab and pilot scales. The residence time was 21 days due to significant decrease in biogas generation (less than 1% of the previous measurement) after that time period.

Experimental Methodologies
Pot ale was chosen as the main substrate of AD. Biodegradability of pot ale was assessed and the effects of increased solids content within the reactors was tested by anaerobic co-digestion (AcoD) of pot ale and spent grain.

Assessment of the Pre-Treatments
Initial screening experiments were performed to assess the impact of alkaline and beating pretreatment on AD of pot ale. Alkaline pre-treatment was performed in accordance with the literature [31,32] in 400 mL batches by dropwise addition of 1 M NaOH into the pot ale sample until pH 10 was obtained. Once pH 10 [29] was achieved the sample was kept stirred for 6 min at 600 rpm. For alkaline pre-treatment of spent grain, the total volume of 400 mL was kept constant and spent grain was treated with required amount of pH 10 adjusted DI water by dropwise addition of 1 M NaOH to obtain a spent grain:DI water ratio of 1:5 by weight. For beating pre-treatment, the beater was fed with 25 L of alkali pot ale sample to have a full circulation in the beating tub. Beating pre-treatment was performed on pot ale prior to, after and without pH 10 1M NaOH treatment for 0, 5, and 7.5 min. All samples were seeded with 50% of sludge (by volume, on wet basis).
The modifications in the lignocellulosic character of pot ale and spent grains were measured prior to and after the application of the alkali pre-treatment. Samples were subjected to sequential steps of hydrolysis and extraction using different reactants (neutral detergent, acid detergent, and 72% H2SO4 solution), in order to assess the lignocellulosic fractions with respect to their sensitivity to acid hydrolysis [48]. However, influences of beating pre-treatment on the lignocellulosic structure could not be assessed because of grinding requirement (particle size between 0.5 mm and 1 mm) of the Van Soest method [48].

Experimental Methodologies
Pot ale was chosen as the main substrate of AD. Biodegradability of pot ale was assessed and the effects of increased solids content within the reactors was tested by anaerobic co-digestion (AcoD) of pot ale and spent grain.

Assessment of the Pre-Treatments
Initial screening experiments were performed to assess the impact of alkaline and beating pre-treatment on AD of pot ale. Alkaline pre-treatment was performed in accordance with the literature [31,32] in 400 mL batches by dropwise addition of 1 M NaOH into the pot ale sample until pH 10 was obtained. Once pH 10 [29] was achieved the sample was kept stirred for 6 min at 600 rpm. For alkaline pre-treatment of spent grain, the total volume of 400 mL was kept constant and spent grain was treated with required amount of pH 10 adjusted DI water by dropwise addition of 1 M NaOH to obtain a spent grain:DI water ratio of 1:5 by weight. For beating pre-treatment, the beater was fed with 25 L of alkali pot ale sample to have a full circulation in the beating tub. Beating pre-treatment was performed on pot ale prior to, after and without pH 10 1M NaOH treatment for 0, 5, and 7.5 min. All samples were seeded with 50% of sludge (by volume, on wet basis).
The modifications in the lignocellulosic character of pot ale and spent grains were measured prior to and after the application of the alkali pre-treatment. Samples were subjected to sequential steps of hydrolysis and extraction using different reactants (neutral detergent, acid detergent, and 72% H 2 SO 4 solution), in order to assess the lignocellulosic fractions with respect to their sensitivity to acid hydrolysis [48]. However, influences of beating pre-treatment on the lignocellulosic structure could not be assessed because of grinding requirement (particle size between 0.5 mm and 1 mm) of the Van Soest method [48].

Evaluating the Effects of Increased Solid Content on Anaerobic Digestion Performance
In order to evaluate the impacts of increased solid content on anaerobic digestion, spent grain and pot ale were seeded with 50% sludge (on wet basis) of the total working volume of 400 mL with different mixing ratios (1:1, 1:3, and 1:5 by wet weight); the experiments were performed at 35 • C. Impacts of alkaline pre-treatment stands alone and in combination with 7.5 and 15 min beating pre-treatment on biogas generation and quality were investigated while a non-treated sample was kept as the control for each mixing ratio. The pre-treatments were performed according to Section 2.4.1.

DOE for AD of Pot Ale and Pot Ale and Spent Grain Mix (5:1 by Wet Weight)
A Box Behnken design of experiments (for process optimisation using Response Surface Modelling (RSM) involving three numerical factors, A: beating time, B: temperature, C: sludge%, was adopted as the RSM.
Pot ale samples were treated with 1M NaOH first, then beaten for 0, 7.5, and 15 min. The required ratio of sample and sludge (10, 30, and 50% sludge) for 400 mL of total volume were place into 500 mL conical flasks and AD experiments were performed at three different temperatures under mesophilic conditions (32,35, and 38 • C), as specified in Table 1. All experiments were performed in triplicate as indicated for each parameter (beating time, sludge percentages, and temperature). The same design parameters were followed for AcoD of pot ale and spent grain mix. Experimental data was examined by stepwise regression and the best fit was achieved by a second order polynomial Equation (1) in order to identify the relevant mathematical model terms. Sequential F-test, lack of fit test, as well as adequacy measures were then implemented in obtaining the best models. The same statistical software was used to generate response plots [52,[55][56][57] Both probability > F (also called as p-value) of the model and the value of each term in the model was computed by analysis of variance (ANOVA). Only if a Prob. > F of the model and of each term in the model was smaller than the level of significance (in this study α = 0.05), in the confidence interval (1-α), the model is considered to be adequate. Developed models were then challenged and validated by independent experiments. The reduced model passed all the required statistical tests to be used to predict system responses and optimisation [56]. Paired t-tests were conducted on the predicted and the independent experiment results for statistical analysis. Both designs were then optimised by using a combination of numerical and graphical optimisation by employing the desirability function explained in detail by [58] offered by DOE software.

Sample Characterisation and Assessment of Pre-Treatments
Characteristics of pot ale sample as received in terms of in terms of TS, VS, moisture content, COD, BOD, sulphate, phosphorous, nitrogen-nitrate, nitrogen-nitrite, nitrogen-ammonium, hemicellulose, cellulose, lignin, VFAs, and dissolved copper are given in Table 2. The effects of alkaline pre-treatment on lignocellulosic structure was assessed for pot ale and spent grain. After implementation of alkaline pre-treatment on both pot ale and spent grain, significant increases and decreases were seen in the hemicellulose and lignin contents respectively while cellulose remained static (Figure 2a,b). One-way ANOVA tests indicated that 1 M NaOH pre-treatment led to a significant increase (5%) in hemicellulose fraction and a significant decrease (9%) in lignin fraction of pot ale with the p-values of 0.032 and 0.0254 respectively. In case of spent grain, alkaline pre-treatment achieved a significant increase in hemicellulose fraction (16%) and a significant decrease (10%) in lignin fraction with the p-values of 0.00032 and 0.000032, respectively.
The biogas results of the initial screening experiments for combined 1M NaOH and beating pre-treatment for 5, 7.5, and 15 min are given in Table 3. The biogas results of the initial screening experiments for combined 1M NaOH and beating pretreatment for 5, 7.5, and 15 min are given in Table 3. The highest cumulative biogas production was achieved with alkaline pre-treatment following by beating for 7.5 min (AB7.5) as 629 ± 8.5 mL/g vs. where the control could only reach 205 ± 21.4 mL/g vs. indicating that hybrid pre-treatment resulted in 3-fold enhancement (p: 0.0029) in biogas yield. Similarly, the highest CH4 content was seen as 51.3% in AB7.5. Moreover, a p-value of 0.0032 (<0.05) was found by application of a one-way ANOVA, indicating that the order in which the pretreatments were performed had a significant effect on total biogas generation. As such, alkaline pre-  The highest cumulative biogas production was achieved with alkaline pre-treatment following by beating for 7.5 min (AB7.5) as 629 ± 8.5 mL/g vs. where the control could only reach 205 ± 21.4 mL/g vs. indicating that hybrid pre-treatment resulted in 3-fold enhancement (p: 0.0029) in biogas yield. Similarly, the highest CH 4 content was seen as 51.3% in AB7.5. Moreover, a p-value of 0.0032 (<0.05) was found by application of a one-way ANOVA, indicating that the order in which the pre-treatments were performed had a significant effect on total biogas generation. As such, alkaline pre-treatment followed by beating pre-treatment was selected as the baseline hybrid pre-treatment for further experiments.

AcoD of Pot Ale and Spent Grain Mixture
The total solids, volatile solids, and moisture content of spent grain and pot ale mixed in ratios of 1:1, 1:3, and 1:5 by wet weight are given in Table 4. The cumulative biogas quantity and quality is given in Figure 3 and Table 5, respectively. The unknown content within the produced biogas was referred as BAL. Table 4. TS, VS, and Moisture% of the non-and pre-treated pot ale and spent grain mix. further experiments.

AcoD of Pot Ale and Spent Grain Mixture
The total solids, volatile solids, and moisture content of spent grain and pot ale mixed in ratios of 1:1, 1:3, and 1:5 by wet weight are given in Table 4. The cumulative biogas quantity and quality is given in Figure 3 and Table 5, respectively. The unknown content within the produced biogas was referred as BAL. The highest biogas yields were achieved with the mixing ratio of 1:5 regardless of the condition of pre-treatments. Among the sample group with 1:5 spent grain pot ale mixing 1 M NaOH pretreatment followed by 7.5 min beating had the highest yield of 469 ± 7 mL/g vs. with a CH4 content of 53.3 ± 1%. It also found to be significantly higher than both 1 M NaOH pre-treatment alone and followed by 15 min beating with the p-values of 0.0033 and 0.00016 respectively. On the other hand, no significant difference was found in the biogas quality regarding the different pre-treatment conditions or mixing ratios. The pH drops from 7 to around 5.5 was observed with all samples. For modelling of AcoD of spent grain and pot ale, a mixing ratio of 1:5 was chosen based on preliminary experiment results.   The highest biogas yields were achieved with the mixing ratio of 1:5 regardless of the condition of pre-treatments. Among the sample group with 1:5 spent grain pot ale mixing 1 M NaOH pre-treatment followed by 7.5 min beating had the highest yield of 469 ± 7 mL/g vs. with a CH 4 content of 53.3 ± 1%. It also found to be significantly higher than both 1 M NaOH pre-treatment alone and followed by 15 min beating with the p-values of 0.0033 and 0.00016 respectively.
On the other hand, no significant difference was found in the biogas quality regarding the different pre-treatment conditions or mixing ratios. The pH drops from 7 to around 5.5 was observed with all samples. For modelling of AcoD of spent grain and pot ale, a mixing ratio of 1:5 was chosen based on preliminary experiment results.

Design of Experiments and Organic Removals
The Box-Behnken design matrix for AD of Pot ale is given in Table 6 with experiment results. A variety of methane content was seen in the generated biogas. The maximum biogas quality and quantity were achieved with the std. no. 12 as 550 ± 6 mL/g vs. and 54.3% CH 4 in respectively. Although all reactions were started at the neutral pH, mainly low final pH values were seen in all samples with a slight difference (higher) in the ones seeded 50% sludge, however it was still lower than optimum pH range for methanogens (std. no. 7, 8, 11, and 12). The same particular samples also had significantly higher CH 4 percentages than the rest. The percentage removals of the organic compounds are given in Table 7. Mainly higher COD and BOD removals were achieved with the samples seeded with 50% sludge (std. no. 7, 8, 11, and 12 in Table 7) suggesting the higher sludge amount delayed the pH drops thus methanogens could survive longer to transform the organic matter into CH 4 . The pH drops were explained with VFA (acetic acid, propionic acid, isobutyric acid, butyric acid, isovaleric acid, and valeric acid) accumulation in the reactors. The total VFAs concentrations before and after AD are given in Figure 4.
In contrast, the highest SO 4 decompositions were seen the samples which contain 10 and 30% sludge while SO 4 removal of 50% sludge contain samples was almost one-third of the values obtained with 30% sludge. Due to the bacterial competition, the sludge seeding ratio lower than 50% results in unsuitable environment for methanogenic bacteria growth, and it triggers the activity of the sulphate reducing bacteria. Digestion under these conditions results in low CH 4 generation and high (>10,000 ppm) H 2 S generation. (acetic acid, propionic acid, isobutyric acid, butyric acid, isovaleric acid, and valeric acid) accumulation in the reactors. The total VFAs concentrations before and after AD are given in Figure  4. All results are average of triplicate runs.
In contrast, the highest SO4 decompositions were seen the samples which contain 10 and 30% sludge while SO4 removal of 50% sludge contain samples was almost one-third of the values obtained with 30% sludge. Due to the bacterial competition, the sludge seeding ratio lower than 50% results in unsuitable environment for methanogenic bacteria growth, and it triggers the activity of the sulphate reducing bacteria. Digestion under these conditions results in low CH4 generation and high (> 10,000 ppm) H2S generation. The VFA concentration after AD was found to be significantly higher for all samples. The std. no. 5, 6, 9, 10, 7, 11 and 12 (which were seeded with 30 and 50% sludge) had lower total VFAs The VFA concentration after AD was found to be significantly higher for all samples. The std. no. 5, 6, 9, 10, 7, 11 and 12 (which were seeded with 30 and 50% sludge) had lower total VFAs concentrations after AD in comparison to 10% sludge seeding. Moreover, 50% sludge containing samples (std. no. 7, 8, 11, and 12) had significantly higher biogas yields with significantly higher methane contents (Table 6) indicating that they were not affected by VFA inhibition as much as the ones containing a lower sludge amount. The std. no. 12 in particular had the lowest total VFA value (16.12 mM) after AD and it had the highest biogas quality and the quantity ( Table 6).
The same design was followed in order to evaluate the impact of process parameters on increased solid content after adding spent grain into pot ale (1:5, by wet weight). Design matrix with the corresponding responses is given in Table 8. A similar range of CH 4 generation was seen with the previous design for AD of pot ale. The highest biogas quality and quantity were obtained with the std. no. 12 as 55 ± 0.4% CH 4 in 360 ± 10 mL/g vs. biogas. Moreover, the final pH value for this particular sample was maintained within the neutral areas, whereas it dropped around the level of 5.5 for the other samples. All results are average of triplicate runs.
The percentage removal of COD, BOD, and SO 4 due to bacterial activity is given in Table 9. The experiments performed with 10% and 30% sludge resulted in low COD and BOD removal percentage levels regardless of the application of the pre-treatments, correspondingly low levels of biogas yields were seen in those samples. Pre-treatments, however, had a significant impact on biogas quality for the same samples i.e., the std. no. 1,2 and 5,6 in (Table 8), CH 4% increased 2-and 5-fold over due to the effect of 15 min beating pre-treatment, respectively. It can be explained by, in early stage of anaerobic digestion, the extended active surface area on solid material in the reactors stimulated microbial activity which resulted in higher CH 4 generation. However, the availability of large amounts of fresh feedstock caused imbalanced reaction rates, sequentially sharp pH drops inhibitory level in particular for the acetoclastic methanogens. The percentage of sulphate removal was found to be significantly higher for the samples seeded with 10 and 30% sludge, which also indicating higher activity of sulphate reducing bacteria with regard to methanogenic bacteria. Moreover, excessive H 2 S production levels (reaching beyond the sensor range >10 000 ppm) in the samples contain 10 and 30% sludge supported the same statement. Significantly higher COD and BOD removals were achieved with the samples seeded with 50% sludge along with the lower SO 4 removals (Table 9) in comparison to the lower sludge seeding ratios. In case of std. no. 7 and 8, 15 min beating pre-treatment resulted in a significant rises in both biogas yield (from 207 ± 19 to 279 ± 11 mL/g VS, p: 0.0093) and quality 28.0 ± 2.1 to 48.5 ± 2.3% (p: 0.0008) CH 4 due to higher COD and BOD removals (Table 9). Moreover, the std. no. 8 had lower H 2 S production (p: 0.0030) than 7 because of 9% less SO 4 reduction (Table 9). Different digestion temperatures were seen to significantly influence biogas production and CH 4 content (std. no. 11 and 12, Table 8). The digestion at 38 • C increased the biogas generation yield from 258 ± 8 to 360 ± 10 mL/g vs. and the percentage, because of higher COD, BOD removals which are given in Table 9, the of CH 4 from 29.3 ± 2.1 to 55.0 ± 0.4 with respect to the digestion at 35 • C. Furthermore, likewise in the std. no. 8, a lower H 2 S production was achieved with experiment no. 12 along with lower SO 4 removal percentages ( Table 9).
Results of the VFA analysis of pot ale spent grain mix before and after anaerobic digestion is given in Figure 5. Significantly higher COD and BOD removals were achieved with the samples seeded with 50% sludge along with the lower SO4 removals (Table 9) in comparison to the lower sludge seeding ratios. In case of std. no. 7 and 8, 15 min beating pre-treatment resulted in a significant rises in both biogas yield (from 207 ± 19 to 279 ± 11 mL/g VS, p: 0.0093) and quality 28.0 ± 2.1 to 48.5 ± 2.3% (p: 0.0008) CH4 due to higher COD and BOD removals (Table 9). Moreover, the std. no. 8 had lower H2S production (p: 0.0030) than 7 because of 9% less SO4 reduction (Table 9). Different digestion temperatures were seen to significantly influence biogas production and CH4 content (std. no. 11 and 12, Table 8). The digestion at 38 °C increased the biogas generation yield from 258 ± 8 to 360 ± 10 mL/g vs. and the percentage, because of higher COD, BOD removals which are given in Table 9, the of CH4 from 29.3 ± 2.1 to 55.0 ± 0.4 with respect to the digestion at 35 °C . Furthermore, likewise in the std. no. 8, a lower H2S production was achieved with experiment no. 12 along with lower SO4 removal percentages ( Table 9).
Results of the VFA analysis of pot ale spent grain mix before and after anaerobic digestion is given in Figure 5. Although total VFA concentration of all samples after AD was so much higher than the initial corresponding values, samples which were seeded with 50% sludge (sample nos. 7,11 in Figure 5) had the lowest final total VFAs values indicating that these particular samples were not affected by the VFA inhibition as much as the others [59]. Therefore, their biogas yields as well as methane percentage of the produced biogas is significantly higher than the samples seeded with lower sludge (Table 8). AD of pot ale spent grain mix have much higher final VFA concentration than pot ale alone even though the same design factors are applied for both feedstock types. This is attributed to higher impact of the lack of internal agitation within the system for AD of the mixed compounds, since it has higher solid content. In other words, microbes are physically blocked to reach spent grain which lead a competition for limited amount of pot ale as the feedstock.

Mathematical Model Estimations
The mathematical models were statistically analysed by ANOVA (using the coding format given in Table 10) and the summary of analysis is given in (Tables 11 and 12), (Tables 13 and 14) and (Tables 15 and 16) for biogas production, CH 4 and CO 2 % for AD of pot ale, and spent grain and pot ale co-digestion respectively. For each experiment, developed models achieved an adequate fit with produced data within the identified significance threshold since the value of lack of fit was <0.05 for all models except for biogas generation by AD of spent grain and pot ale mix <0.1. In other words, estimated model on biogas production by AD of spent grain and pot ale had no significant lack of fit within 90% confidence interval, whereas the confidence interval is 95% for all other models. Furthermore, the values of R 2 , adjusted R 2 (Adj. R 2 ) and predicted R 2 (Pred. R 2 ) were close to 1 for all models indicating a sufficient regression for the models, plus the difference between the adjusted R 2 (Adj. R 2 ) and predicted R 2 (Pred. R 2 ) was less than 0.2. An adequate precision value greater than 4 indicates adequate model discrimination [60] achieved here in all cases. Mathematical models were built based on actual and coded factors allowing to plot different design factors with different units in the same graph to compare their individual effects on the response of interest.      For biogas yield, the first and the second order of sludge percentage (C, C 2 ) were significant parameters regardless of feedstock type according to ANOVA. In case of AD of pot ale the first and second order of the beating time (A, A 2 ) was found to be a significant parameter (Table 11) whereas in case of AD of spent grain and pot ale temperature (B) the interaction between temperature and sludge percentage (BC) were the significant factors (Table 12).
Final equations showed the biogas generation (Y) as a function of the independent variables in terms of coded and actual factors on biogas production by AD of pot ale and spent grain and pot ale mixture given in Equations (2) and (3) and Equations (4) and (5) The coefficients of the final equation in terms of coded factors (Equations (2) and (3)) indicates the power of the design parameters on biogas generation as C > C 2 > A 2 > A and C > C 2 > BC > B for AD of pot ale and spent grain pot ale mixture.
ANOVA results for CH 4 % modelling, showed the digestion temperature (B), the first and second order effects of sludge (C, C 2 ), as well as interaction between temperature and sludge percentage (BC) as significant parameters for AD of pot ale in 95% confidence interval (Table 13).The beating time (A), first and the second order effects of temperature (B, B 2 ) and sludge% (C, C 2 ), as well as the interaction effect of temperature and sludge percentage (BC) were the significant model terms for AD spent grain and pot ale mixture (Table 14). Influence strength of the significant parameters on CH 4 concentration within the biogas generated by AD of pot ale and pot ale and spent grain mixture was identified as C > C 2 > BC > B and C > C 2 > BC > B > A > B 2 respectively by considering coefficient of the final equation in terms of coded factors Equations (6) and (7). The final models in terms of actual factors are given in Equations (8) and (9) for AD of pot ale and AcoD of spent grain and pot ale mixture respectively.
Coded: For modelling CO 2 concentrations, ANOVA analysis defined the second order effect of sludge% (C 2 ) and the interaction between temperature and sludge percentage (BC) as significant model terms within a 95% confidence interval for AD of pot ale (Table 15). The first order effects of temperature (B) and sludge% (C) were added to the model by the software to support hierarchy. The first and second order effects of sludge percentages (C, C 2 ) were the significant parameters for CO 2 modelling by AD of spent grain and pot ale mixture (Table 16). However, temperature (B) and the interaction effect of temperature and sludge% (BC) were also added to the model after the stepwise regression to improve the model by the software. The strength of the model terms on CO 2 content of generated biogas by AD pot ale and spent grain pot ale mixture was defined as C 2 > BC > B >C and C 2 > C > BC > B based on the coefficient of the final equation in terms of coded factor (Equations (10) and (11)). The final mathematical models on CO 2 content of produced biogas are given in Equations (12) and (13). Coded:

Validation of the Estimated Models
The diagnosis of the estimated models was performed by the same software as part of the post statistical analysis. The diagnostic plots, the normal plot of residuals and the predicted vs. actual plot for AD of pot ale, are given in Figures 6 and 7  to the diagonal line. The developed mathematical model is therefore adequate. The similar trend in both normal plot of residuals and actual vs. predicted plot is seen for the built models on CH4 and CO2 percentages for AD pot ale as well as all models developed for AD of spent grain and pot ale mixture.  The normal distribution of residual data in Figure 6 proves that ANOVA can be applied to the dataset. Figure 7 gives the comparison between the actual data set and the predicted data set by developed model. It is seen that the residuals are minimum since all the design points tend to be close to the diagonal line. The developed mathematical model is therefore adequate. The similar trend in both normal plot of residuals and actual vs. predicted plot is seen for the built models on CH 4 and CO 2 percentages for AD pot ale as well as all models developed for AD of spent grain and pot ale mixture. In addition to post analysis, the models were challenged by independent experiments for a further validation step. Three validation points for validating the models on AD of pot ale and two validation points for validating models on AD pot ale and spent grain (which were not used for model development) were decided and the results of the experiments were given in Table 17 and 18 respectively along with the predicted values by the models in terms of actual factors for each interest of response. The Equations (4), (8), and (12) were used to predict the results, given in Table 17, of the validation for biogas yield, CH4% and CO2% by AD of pot ale respectively. Likewise, the Equations (5), (9), and (13) were used for the predictions, given in Table 18 on biogas yield, CH4% and CO2% respectively though AD of spent grain and pot ale mix. The difference between the predicted and the experimental results were then statistically analysed by t-test (2-tailed distribution with unequal variance). The modelling of the biogas quality and the quantity was validated for both designs within a 95% confidence interval in accordance according to the p-values in Table 17 and 18  In addition to post analysis, the models were challenged by independent experiments for a further validation step. Three validation points for validating the models on AD of pot ale and two validation points for validating models on AD pot ale and spent grain (which were not used for model development) were decided and the results of the experiments were given in Tables 17 and 18 respectively along with the predicted values by the models in terms of actual factors for each interest of response. The Equations (4), (8), and (12) were used to predict the results, given in Table 17, of the validation for biogas yield, CH 4 % and CO 2 % by AD of pot ale respectively. Likewise, the Equations (5), (9), and (13) were used for the predictions, given in Table 18 on biogas yield, CH 4 % and CO 2 % respectively though AD of spent grain and pot ale mix. The difference between the predicted and the experimental results were then statistically analysed by t-test (2-tailed distribution with unequal variance). The modelling of the biogas quality and the quantity was validated for both designs within a 95% confidence interval in accordance according to the p-values in Tables 17 and 18.

Model Graphs
The perturbation plots give the effect of all factors, which have different units, on the same graph since it is plotted in terms of coded factors. The lines show the individual behaviour of each factor while keeping the other at a constant ratio (their centre points by default). It can be used to identify which factor affects the response of interest the most. Figure 8a-c and d-f shows the perturbation graphs for biogas yield, CH 4 % and CO 2 % for AD of pot ale and AcoD of spent grain pot ale mix respectively. In both cases, the curvature related to the sludge percentages (C) showed that this factor had a greater impact on the response than the beating time (A) and the digestion temperature (B). Although results were different, the digestion temperature (B) and the sludge ratio (C) had similar effects on CH 4 % and CO 2 % for both DOE experiments (Figure 8b,e) and Figure 8c,f respectively. Figure 9a-c and d-f give the contour graphs for biogas yield, CH 4 % and CO 2 % for AD of pot ale and AcoD of spent grain pot ale mix respectively. The effects of temperature and sludge percentages mainly have straight lines on all responses. However, a peak was seen for the area around 20-30% sludge for biogas yield, CH 4 and CO 2% for AcoD of spent grain and pot ale (Figure 9d-f) as well as CO 2% of the biogas produced by AD of pot ale (Figure 9c). In both experiments, the greatest biogas generation and the highest CH 4 percentage are achieved under the condition of higher sludge seeding ratio and digestion temperature, while the same conditions resulted in lower CO 2 %.
There was no interaction between the factors for the model developed on biogas yield for AD of pot ale. The interactions between the sludge percentages and the digestion temperatures for the quality of biogas produced by AD of pot ale is given in Figure 10a,b along with the interactions for biogas yield CH 4 and CO 2 percentages for AcoD for spent grain pot ale. The digestion temperature had no significant influence at low sludge seeding ratios (<20%). However, once the sludge ratio was increased, the digestion at 38 • C has significantly increased the biogas production by AcoD of spent grain and pot ale and its CH 4 content (Figure 10c,d) as well as the biogas quality of AD of pot ale (Figure 10a). In contrast, a completely opposite interaction was seen on CO 2 contents of the biogas produced by AD of pot ale and AcoD of spent grain pot ale mix. Digestion at 38 • C had a significantly lower CO 2% within the biogas produced by AD of pot ale (Figure 10b). However, the effect of digestion temperature on it became insignificant once the sludge seeding ratio is increased. An opposite figure was seen on CO 2 content of for AcoD of spent grain pot ale (Figure 10e).  There was no interaction between the factors for the model developed on biogas yield for AD of pot ale. The interactions between the sludge percentages and the digestion temperatures for the quality of biogas produced by AD of pot ale is given in Figure 10 a,b along with the interactions for

Optimisation of the Developed Models
The design factors were kept in range however biogas yield and CH4% were maximised with the importance of 5 whereas CO2% was minimised with the same importance level. The optimisation

Optimisation of the Developed Models
The design factors were kept in range however biogas yield and CH 4% were maximised with the importance of 5 whereas CO 2% was minimised with the same importance level. The optimisation conditions for AD pot ale and spent grain and pot ale mix are outlined in Tables 19 and 20 respectively. Corresponding suggested solutions are given in Tables 21 and 22 respectively. The first solutions were selected based on its higher desirability to define lower and upper limits of graphical optimisation. The graphical optimisation of AD of pot ale and spent grain pot ale mix is given in Figure 11a,b respectively. The target area coloured with yellow is delimited by constraints set according to the numerical optimisation solutions (Tables 21 and 22). For example, in the case of AD of pot ale, lower and upper limits of the optimum area are 520 and 521 mL/g vs. for biogas generation with 22 and 24% for CO 2 content. In the case of CH 4 %, only the lower limit (48.3%) is visible on the chart because the lower limit does not overlap with the other criteria. In the case of AD of spent grain, pot ale mix defining lines for the target area are 317.03 mL/g vs. biogas production, 53.2 CH 4 and 36% CO 2 percentages.

Discussion
The characteristics of non-treated pot ale were mainly found to be in agreement with the literature in terms of the concentrations of COD (38867 ± 115 mg/L), BOD (30965 ± 666 mg/L), SO4 (190 ± 31 mg/L) within the reported ranges 30,000-50,000 mg/L for COD [61][62][63][64], 25,000-35,000 mg/L for BOD [49,64,65], and 223-284.8 mg/L for SO4 [66]. The pH (3.7) was also within the reported range of 3.5-4.5 [67,68]. A slight difference was found between the concentrations of PO4 3-(778 ± 7 mg/L), N-NH3 (45 ± 7 mg/L), N-NO3 (111 ± 20 mg/L), N-NO2 (33 ± 4 mg/L), and Cu (14.7 ± 1 mg/L) and the corresponding reported concentrations: 714mg/L [63,69] for PO4 3-, 0.12-7.9 mg/L for N-NH3 [70], 8.3-53 mg/L for N-NO3, 16-20 mg/L for N-NO2, and 2-6 mg/L for Cu [64]. These differences were attributed to use of different source of raw materials and process parameters of the sampling distillery. Biogas yield and quality obtained from AD of whiskey distillery/brewery waste streams (pot ale and spent grain pot ale mix) was examined using lab scale batch experiments. The effect of alkaline pre-treatment (1M NaOH) on the lignocellulosic structure of pot ale and spent grain was found to be a significant parameter on lignin degradation as well as on hemicellulose solubilisation which was then associated with overall AD performance. A hybrid alkaline and beating pretreatment was introduced. In order to assess the efficiency of the digestion, organic matter removals in terms of COD and BOD were measured before and after AD, while the potential bacterial competition between acetoclastic methanogenic archaea and sulphate reducing bacteria was traced with SO4 removal. Up to 78 ± 5% COD and 61 ± 4% BOD removals were achieved by AD of pot ale along with 25±1% SO4 degradation, while COD and BOD removals up to 73 ± 6 and 63 ± 7% respectively were achieved by AcoD of spent grain and pot ale along with a 31 ± 7 SO4 removal. The maximum biogas yields of 629 ± 8.5 mL/g vs. with a 51.3% CH4 content (Table 3) and 360 ± 10 mL/g vs. with a 55.0 ± 0.4 CH4%content (std. 12 in Table 8) over 21 days of digestion period was achieved for samples subjected to alkaline and subsequent 7.5 min beating pre-treatments; seeded with 1:1 sludge ratio on wet basis at 35 (for pot ale) and 38 °C (for pot ale and spent grain) digestion temperature. A methane yield of 554 ± 67 mL/g vs. was reported by [70] with AD of non-treated pot ale with a substrate inoculum ratio of 1:2 (on wet basis) at lab scale batch experiments, while in this study the highest methane yield was found as 322 ± 4 mL/g vs. (Table 3) by AD of alkaline and 7.5 min beating pre-treated pot ale with a substrate to inoculum ratio of 1:1. The difference between the two yields was attributed to the lower inoculum ratio in the experiments as the reactor configurations

Discussion
The characteristics of non-treated pot ale were mainly found to be in agreement with the literature in terms of the concentrations of COD (38,867 ± 115 mg/L), BOD (30,965 ± 666 mg/L), SO 4 (190 ± 31 mg/L) within the reported ranges 30,000-50,000 mg/L for COD [61][62][63][64], 25,000-35,000 mg/L for BOD [49,64,65], and 223-284.8 mg/L for SO 4 [66]. The pH (3.7) was also within the reported range of 3.5-4.5 [67,68]. A slight difference was found between the concentrations of PO 4 3 16-20 mg/L for N-NO 2 , and 2-6 mg/L for Cu [64]. These differences were attributed to use of different source of raw materials and process parameters of the sampling distillery. Biogas yield and quality obtained from AD of whiskey distillery/brewery waste streams (pot ale and spent grain pot ale mix) was examined using lab scale batch experiments. The effect of alkaline pre-treatment (1M NaOH) on the lignocellulosic structure of pot ale and spent grain was found to be a significant parameter on lignin degradation as well as on hemicellulose solubilisation which was then associated with overall AD performance. A hybrid alkaline and beating pre-treatment was introduced. In order to assess the efficiency of the digestion, organic matter removals in terms of COD and BOD were measured before and after AD, while the potential bacterial competition between acetoclastic methanogenic archaea and sulphate reducing bacteria was traced with SO 4 removal. Up to 78 ± 5% COD and 61 ± 4% BOD removals were achieved by AD of pot ale along with 25±1% SO 4 degradation, while COD and BOD removals up to 73 ± 6 and 63 ± 7% respectively were achieved by AcoD of spent grain and pot ale along with a 31 ± 7 SO 4 removal. The maximum biogas yields of 629 ± 8.5 mL/g vs. with a 51.3% CH 4 content (Table 3) and 360 ± 10 mL/g vs. with a 55.0 ± 0.4 CH 4% content (std. 12 in Table 8) over 21 days of digestion period was achieved for samples subjected to alkaline and subsequent 7.5 min beating pre-treatments; seeded with 1:1 sludge ratio on wet basis at 35 (for pot ale) and 38 • C (for pot ale and spent grain) digestion temperature. A methane yield of 554 ± 67 mL/g vs. was reported by [70] with AD of non-treated pot ale with a substrate inoculum ratio of 1:2 (on wet basis) at lab scale batch experiments, while in this study the highest methane yield was found as 322 ± 4 mL/g vs. (Table 3) by AD of alkaline and 7.5 min beating pre-treated pot ale with a substrate to inoculum ratio of 1:1.
The difference between the two yields was attributed to the lower inoculum ratio in the experiments as the reactor configurations and alkali conditions were similar in both works. The efficiency of AD was evaluated based on the removal of organic matter by the treatment and various COD and BOD degradations have been reported for different reactor configurations and pre-treatment conditions for AD of pot ale. During the first 140 days of digestion, a typical COD removal was reported as 70-90% by AD of alkaline (NaHCO 3 ) pre-treated pot ale using a UASB [61]. Enhanced levels of COD (96%) and BOD (99%) removals were obtained by [63] for AD of pot ale with addition of NaHCO 3 and nutrient supplement for accelerated growth of culture using two stage UASB. On the other hand, 13 and 50% COD reductions were obtained by AD of pot ale without and with enzymatic pre-treatment using 10% papain in a lab scale batch reactor [67]. In this work, the highest levels of organic matter removals (78 ± 5% for COD and 61 ± 4% for BOD) were seen in AD of pot ale (std. no. 8 in Table 7), it was slightly higher than co-digestion of pot ale and draff mixture. Organic degradation rates were found to be in an agreement with published works [61,67]. The sludge percentages had the biggest impact on all responses according to the perturbation plots given Figure 8 as well as the mathematical models in terms of coded factors. Sludge percentage was found to be the most powerful factor on interest of responses. Therefore, samples containing low sludge percentage (10 and 30%) had low biogas generation and methane quality regardless of the pre-treatment. The optimum digestion conditions were identified as 10-and 13-min beating times, 32 and 38 • C digestion temperature for AD of pot ale and AcoD of spent grain pot ale mixture respectively as well as 50% sludge seeding ratio for both cases.