DEVELOPMENT OF THE FR 13 RISK FRAMEWORK – DEMONSTRATED WITH ONE- AND TWO-STEP STEADY-STATE MEMBRANE PROCESSING OF JUICES by Mr Wei ZOU School of Chemical Engineering The University of Adelaide A thesis submitted for examination for the degree of Master of Philosophy (Chemical Engineering) December – 2015 ii STATEMENT OF DECLARATION I, Wei Zou, certify that this work contains no material which has been accepted for the award of any other degree or diploma in any university or other tertiary institution and, to the best of my knowledge and belief, contains no material previously published or written by another person, except where due reference has been made in the text. I give consent to this copy of my thesis when deposited in the University Library, being made available for loan and photocopying, subject to the provisions of the Copyright Act 1968. The author acknowledges that copyright of published works contained within this thesis1 resides with the copyright holders of those works. I also give permission for the digital version of my thesis to be made available on the web, via University’s digital research repository, the Library catalogue and also through web search engines, unless permission has been granted by the University to restrict access for a period of time. Signature: Date: 1 Davey, K.R., Zou, W., 2015. Fruit juice processing and membrane technology application (sic) – A response. Food Eng. Rev. – submitted Oct. 2015. Zou, W., Davey, K.R., 2014. A Friday 13th risk model for failure in cross - flow membrane filtration of passion fruit juice. In: Proc. 26th European Modelling and Simulation Symposium - EMSS 2014, Sept. 10 – 12, Bordeaux, France, paper 106. ISBN: 9788897999386 Zou, W., Davey, K.R., 2014. A novel Friday 13th stochastic assessment of failure of membrane filtration in juice clarification. In: Proc. 44th Australasian Chemical Engineering Conference (Processing Excellence, Powering our Future). Sept. 27 – Oct. 1, Perth, WA, Australia, paper 1259. ISBN/ISSN: 1922107387 Zou, W., Davey, K.R., 2015. An integrated two-step Fr 13 synthesis – demonstrated with membrane fouling in combined ultrafiltration-osmotic distillation (UF-OD) for concentrated juice. Chem. Eng. Sci. – submitted on Nov. 2015. iii EXECUTIVE SUMMARY Steady-state processing is used extensively in foods and chemical, engineering, and more widely. Importantly, however there will be naturally occurring, random fluctuations in parameter values about an apparent steady-state ‘set’ mean. These are not sufficient, on their own, to be considered transient i.e. unsteady-state. Generally, random small change in one parameter is ‘off-set’ by a corresponding change in another - with the output seemingly remaining steady. Traditional chemical engineering does not address these random fluctuations explicitly. However, Davey and co-workers (e.g. Abdul-Halim and Davey, 2015 a, b; Davey, 2015 a; Davey et al., 2016) have reasoned that process failures can result from the accumulation of these fluctuations within an apparent steady-state process itself. Their hypothesis is that naturally occurring chance fluctuations can unexpectedly combine and accumulate in one direction and leverage significant change across a binary ‘not failure - failure’ boundary. That is to say, even with good design and operation of plant, there can be unexpected (surprise and sudden) occasional failures. This they titled Fr 13 (Friday 13th) to underscore the nature of the event. A current limitation of the Fr 13 framework however that is it has been largely limited to one-step (single) unit-operations. It was not known therefore if there was any benefit in developing the framework as a useful tool for integrated multi-step foods and chemicals engineering unit-operations. A research program was therefore undertaken with the aim to advance the Fr 13 framework to gain unique insight into how naturally occurring fluctuations in apparent steady-state plant parameters can be transmitted and impact in progressively complex (in the context of ‘integrated’ not ‘complicated’) multi-step processes, and to assess the Fr 13 framework as a new design tool. A logical and stepwise approach was implemented as a research strategy. Because steady-state membrane clarification and concentration of fruit juices is becoming a widespread alternative to traditional thermal treatments, a two-step membrane concentration was selected as a timely and stringent test of development of the Fr 13 risk thesis. iv Two, preliminary single-step Fr 13 membrane models, ‘dead-end’ and ‘cross-flow’, were initially synthesized and tested with independent experimental data for clarification of orange (n = 25) and blood orange (n = 34) juice. Fr 13 simulations of the key input parameters, transmembrane pressure (∆P), filtration time (t) and volumetric flow rate (Q), revealed that some 16.8 % of dead-end and 4.0 % of cross-flow membrane filtrations, over an extended time, will fail to meet the required operational flux plus a practical tolerance (2 %) as a design margin of safety. If each filtration is thought of as a daily batch-continuous operation, then an unexpected fouling failure could result every six (6) and 26 days, respectively, in batch–continuous processing. A more commercially representative integrated two-step Fr 13 membrane global model was then synthesized for combined ultrafiltration (UF) - osmotic distillation (OD) (UF-OD), and validated with extensive independent data (n = 27) for pomegranate (Punica granatum) juice. Overall global failure of the integrated two-step UF-OD was defined as a fouled (unwanted) OD flux. Fr 13 simulations showed that the integrated UF-OD is expected to be vulnerable to surprise fouling failure in 10.5 % of all operations over an extended time. This translates to 39 surprise failures per year with a 3 % design tolerance. In completing this work an important error in the membrane literature was discovered, corrected and addressed2. Fr 13 simulations of these newly synthesized models underscored that the three (3) apparent steady-state membrane processes should be more correctly thought of as a combination of successful and failed operations. This new insight is not available from traditional risk and hazard analyses, with or without sensitivity analyses. Findings were applied in ‘second-tier’ studies to assess re-design of membranes processing of juices. The aim was to improve process reliability and reduce vulnerability to unexpected failure. For example, repeat Fr 13 simulations revealed that for the integrated 2 The permeate flux (J ) used for measuring membrane permeability was found to be widely incorrectly 0 defined e.g. by Schafer et al. (2005), Boerlage et al. (2002) and Echavarria et al. (2011) in which it was not possible to reconcile the form or units in engineering science. The effect was to significantly overestimate the predicted vulnerabilities to Fr 13 fouling failure, in the preliminary work. This is being addressed by Davey and Zou (2015) (see Appendix C). v UF-OD, reducing variance about the mean value of transmembrane pressure (∆P ) and filtration time (t ) of UF significantly reduces overall UF-OD failures UF 1-1 UF 1-1 (p ). Practically, this suggests costs for increased precision control to limit fluctuations in 2 the UF parameters could be readily justified. Further, second-tier simulations of the integrated UF-OD showed that the addition of an enzymatic treatment step prior to UF could significantly reduce the overall UF-OD failures through a reduction in the required (design) operational UF flux (J ). UF 1-1, required These findings will aid an enhanced understanding of factors that contribute to unwanted fouling as membrane failure, and to increased confidence in steady-state membranes operations. It is concluded that the Fr 13 framework is generalizable to an integrated two-step, steady-state processes. Additionally, there appears no methodological barriers to advancement. Therefore results auger well for further advancement of the Fr 13 framework to a range of steady-state processes of increasing complexity and inter-connectedness. If properly developed, it is thought that Fr 13 could become a new decision tool, in both design analysis and synthesis, for improved understanding of process behaviour outcomes. This research is original and not incremental work. Outcomes are of immediate interest to researchers in risk analyses and processors and manufacturers of membrane equipment. vi ACKNOWLEDGMENTS I would like to express my gratitude to Dr K R (Ken) Davey (FIChemE), my principal supervisor from the School of Chemical Engineering, The University of Adelaide, for his patience, instruction, encouragement and time in guiding and helping throughout my candidature. I especially acknowledge his immense help and support in addressing various issues that arose. Also I wish to thank Dr Brian O’Neill, my co-supervisor from the School of Chemical Engineering, The University of Adelaide, for providing guidance and advice. I would also like to thank Professor Peter Ashman, Head of the School, and to all staff members in the School of Chemical Engineering, for giving me the opportunity to complete my research degree. I am greatly indebted to my parents who gave me the opportunity to come to Australia and I would also like to thank them for providing me with the financial assistance for doing so. I would also like to thank my colleagues, especially Saravanan, Priyantha, Nadiya, James, and all my friends here in Adelaide for being there when needed, and for providing moral support. I hope that the results of my efforts justify the expectations and confidence of the people concerned, and the interest, help, and encouragement of all my family, friends and colleagues. vii TABLE OF CONTENTS PAGE EXECUTIVE SUMMARY iii ACKNOWLEDGEMENTS vi TABLE OF CONTENTS vii LIST OF FIGURES x LIST OF TABLES xii CHAPTER 1 INTRODUCTION 1 CHAPTER 2 LITERATURE REVIEW 5 2.1 Introduction 6 2.2 Traditional single value assessment (SVA) solution 6 2.3 Development of the Fr 13 framework 7 2.3.1 A Fr 13 risk assessment 8 2.3.2 The 5-step Fr 13 risk algorithm 10 2.3.3 Fr 13 applications 10 2.3.4 Fr 13 and other risk approaches 14 2.3.5 Fr 13 as terminology 15 2.3.6 Benefits and limitations of Fr 13 16 2.4 Membranes 17 2.4.1 Structure 17 2.4.2 Materials 18 2.4.3 Module 19 2.5 Membrane unit-operations 20 2.5.1 Classification of membrane unit-operations 21 2.5.2 Advantages and disadvantages of membrane processing 22 2.5.3 Dead-end and cross-flow filtration 23 2.5.4 Osmotic distillation 26 2.6 Membrane failure as fouling 28 2.7 Membrane unit-operations models 30 2.8 Fr 13 framework for membranes 33 2.9 Chapter summary and conclusions 34 CHAPTER 3 A PRELIMINARY ONE-STEP FR 13 MEMBRANE MODEL FOR STEADY-STATE DEAD-END FILTRATION OF 36 ORANGE JUICE 3.1 Introduction 37 3.2 Dead-end membrane model with independent data for orange juice 37 3.2.1 Synthesis of a unit-operations model 37 3.2.2 Required (design) operational flux for orange juice 39 3.3 Traditional single value assessment (SVA) 40 viii 3.4 Fr 13 risk assessment 40 3.4.1 Defining dead-end membrane failure (risk factor) 40 3.4.2 Fr 13 simulation 41 3.5 Results 42 3.6 Discussion 43 3.6.1 Dead-end membrane failures 43 3.6.2 Visualizing Fr 13 risk failures 45 3.7 Chapter summary and conclusions 47 CHAPTER 4 A PRELIMINARY ONE-STEP FR 13 MEMBRANE MODEL FOR STEADY-STATE CROSS-FLOW FILTRATION OF 49 BLOOD ORANGE JUICE 4.1 Introduction 50 4.2 Cross-flow membrane model with independent data of blood orange juice 50 4.2.1 Synthesis of a unit-operations model 50 4.2.2 Required (design) operational flux for blood orange juice 52 4.3 Traditional single value assessment (SVA) 53 4.4 Fr 13 risk assessment 53 4.5 Results 54 4.6 Discussion 57 4.6.1 Cross-flow membrane failures 57 4.6.2 Fr 13 second-tier simulation 57 4.6.3 Comparison of dead-end and cross-flow membrane models 59 4.7 Chapter summary and conclusions 60 CHAPTER 5 A NOVEL TWO-STEP Fr 13 MEMBRANE GLOBAL MODEL FOR INTEGRATED ULTRAFILTRATION-OSMOTIC 62 DISTILLATION (UF-OD) OF CONCENTRATED JUICES 5.1 Introduction 63 5.1.1 Two-step membrane processing of juices 63 5.2 A two-step membrane global model of integrated UF-OD concentration 65 5.2.1 Cross-flow UF clarification 65 5.2.2 OD concentration 68 5.2.3 UF-OD membrane global model 70 5.3 Deterministic single value solution (SVA) 71 5.4 Fr 13 model and simulations 71 5.5 Results 75 5.6 Discussion 79 5.6.1 UF-OD global simulations 79 5.6.2 UF-OD membrane failures 79 5.6.3 Input probability distributions 83 5.6.4 Process tolerance 84 5.6.5 Impact of filtration time 86 ix 5.6.6 Depicting membrane failure in the integrated two-step Fr 13 86 synthesis 5.6.7 Reducing process vulnerability through Fr 13 second-tier simulations 89 5.7 Advancing the Fr 13 framework 93 5.7.1 Integrated multi-step processes 93 5.7.2 Time, cost, effort and benefit 94 5.8 Summary and conclusions 95 CHAPTER 6 CONCLUSIONS AND FUTURE DEVELOPMENT 96 6.1 Conclusions 97 6.2 Recommendations for future research 99 APPENDICES A - F 100 A A definition of some important terms used in this research 101 B Fish bone diagram for Fr 13 simulation of integrated UF-OD membrane global model for pomegranate juice 103 C Correcting the literature – Letter to Editor 104 D Refereed publications from this research 107 E A detailed solution to the model synthesis for integrated UF-OD 108 F A detailed simulation data of a Fr 13 risk assessment for integrated UF-OD 111 NOMENCLATURE 122 REFERENCES 127 x LIST OF FIGURES PAGE Fig. 2-1 A simplified schematic of a typical membrane unit-operation for 21 juices (adapted from Domingues et al., 2014) Fig. 2-2 Schematic diagrams of steady-state dead-end filtration mode (a) and 24 cross-flow filtration mode (b) (adapted from Echavarria et al., 2011) Fig. 2-3 Mechanism of OD through a porous hydrophobic membrane with 27 different water vapour of feed and stripping solutions (adapted from Hogan et al., 1998) Fig. 2-4 Schematic representation of fouling mechanisms for porous 29 membranes: (a) Complete pore blocking; (b) Internal pore blocking; (c) Partial pore blocking; (d) Cake layer formation (reproduced from Cui et al., 2010) Fig. 3-1 Experimental data for permeate flux (n = 25) in steady-state 39 dead-end filtration of orange juice at ∆P = 344.74 kPa (adapted from Nandi et al., 2012) and T = 25 oC showing determination of J = 38.2 x 10-6 m3 m-2 s-1 required Fig. 3-2 Summary of Fr 13 simulation of 100,000 scenarios of risk factor (p) 43 for steady-state, dead-end filtration of orange juice with a tolerance = 2 %. The 16,843 failed scenarios are seen on the right side of the figure where p > 0 Fig. 3-3 3D plot of 40 selected Fr 13 failures (p > 0) in steady-state 46 dead-end filtration of orange juice with tolerance = 2 %: 3D scatter plot (a) and 3D surface plot (b). The highlighted failure of Table 3-1 is indicated by the data marker ӿ Fig. 4-1 Independent experimental data (n = 34) for permeate flux in 52 steady-state cross-flow filtration of blood orange juice at ∆P = 85 kPa, showing determination of J = 3.292 x 10-6 m3 m-2 s-1 required Fig. 4-2 An overall summary of Fr 13 simulation of 100,000 scenarios of 55 risk p for the steady-state cross-flow filtration of blood orange juice with a tolerance = 2 %. The 4,000 failures are seen in the right side of the figure, evidenced by p > 0 Fig. 4-3 Impact of process control as %stdev in distribution for both ∆P and 58 t on the failures (p) numbers in steady-state cross-flow filtration of blood orange juice Fig. 5-1 Schematic for integrated UF-OD membrane processing of juice 64 (adapted from Cassano et al., 2011)