Original Research Paper
A systematic framework to monitor mulling processes using Near Infrared spectroscopy

https://doi.org/10.1016/j.apt.2016.03.022Get rights and content

Highlights

  • Agglomerate size does not affect NIR prediction accuracy.

  • Smoothness and roughness of sample surface has little effect on NIR results.

  • Sample–detector distance should ideally fall between 9 mm and 24 mm.

Abstract

The optimal design of sensor location and setup is essential to ensure the accuracy and precision of in-line process monitoring of water/moisture content. This manuscript presents a systematic framework of using Near Infrared (NIR) spectroscopy to monitor moisture content in an alumina mulling process that is commonly used in the upstream operation of catalyst supports production. For this, the optimal conditions of NIR sensor setup and critical quality attributes (CQAs) of a mulling process have been first identified and then calibration models for monitoring moisture at various conditions were developed and validated. The results suggest that there is a strong relationship between sensor setup and prediction accuracy. Therefore, optimal conditions such as operating distance of the NIR sensor, sample thickness and acquisition number need to be identified prior to installment of the sensor into the manufacturing plant. In mulling processes, the particle size distribution (PSD) and surface roughness/smoothness can also vary during operation, making the monitoring of moisture content a difficult task. In this study, the effects of PSD and powder surface characteristics on moisture content measurement has been investigated and it has been found that if suitable raw data preprocessing has been applied, the effect of agglomerate size and sample surface characteristics on the accuracy of the in-line measurements can be significantly minimized. This will allow the use of a single calibration model for a range of PSDs and powder bed smoothness/roughness and that will save a significant amount of time and resources. Here, a mulling process where a microNIR sensor has been used for monitoring of moisture content has been considered as a demonstrative example. However, the approach is generic and can be applied for any combination of process and sensor.

Introduction

Inorganic metal-oxide supports offer unique properties for the immobilization of metal complexes, such as mechanical strength, surface area and porosity [1]. Alumina is one such inorganic support which has been widely used for several decades in the catalyst industry with applications in the petroleum [2], petrochemical [3], pharmaceutical [4] industries and biofuels [5]. The presence of the alumina support [5] and even the structure of alumina can significantly affect the performance of catalysts, as indicated by Luo et al. [3].

Catalysts supported on alumina, which are synthesized by ammonia precipitation, exhibit higher activity and selectivity to isobutene than the corresponding catalyst synthesized from hydrochloric acid reflux. Due to its importance, there is a large demand for alumina supports in the catalyst industry. However, the large-scale manufacturing of extrudable alumina with desired particle and pore size distribution, and surface area is still a challenging task in the current catalyst production scenario due to difficulties in handling powders.

Usually, the production of alumina is processed in three steps: mulling, extrusion and calcination. Mulling is similar to granulation but is applied in catalyst manufacturing and incorporates wetting, mixing, primary particle size reduction and particle shaping, all within a single unit operation. It improves flowability, reduces dustiness and increases bulk density. Moreover, several highly favored properties including high porosity and large surface area, which brings about faster reaction rates, are introduced by mulling. This effect is achieved by decreasing primary particle size with the help of peptizing agents and high shearing action. A well-mulled paste after undergoing extrusion, to gain the desired shape, is subsequently calcined. The process of calcination removes any residual moisture and mainly volatile fractions leaving behind pores in these alumina supports that allow for the impregnation of catalyst. A set of process variables, namely temperature, moisture content, pH, impeller speed and chopper speed, are known to affect the quality of the final product [6], [7], [8]. Therefore, it is a challenging process to optimize and scale up based on the process variables alone. Extrusion is a commonly used shaping process in many industries, as well as in the catalyst industry, since extruded catalyst supports are much easier to handle and recycle. The main condition that the extruded materials have to fulfill is having sufficient plasticity [9]. Alumina is a non-plastic material and, therefore, extrusion of alumina pastes requires processing additives such as binder and lubricant agents to impart plasticity and flow characteristics, which is another reason for the need of mulling in catalyst support production. In the mulling process, water, with the possibility of a peptizing agent such as nitric acid, can be used as the binder, wherein the binder to solids ratio has significant effect on the plasticity. This makes the monitoring of water/moisture content extremely important. Much less attention however has been paid to the monitoring of this moisture content.

Generally, batch processing is a commonly used method since the machinery for batch production is already installed and is more flexible, in terms of both machine application and productivity adjustment. However, continuous processing is more popular since it helps with stabilization of product quality as well as reduction of manufacturing time and cost of goods. It has been well established in the chemical, cosmetics and food industry [10], [11], [12]. For continuous process monitoring, non-destructive monitoring techniques are generally used, especially vibrational spectroscopic methods such as Near Infrared (NIR) and Raman, which are combined with multivariate calibration routines for use as in-line process analyzers [13], [14], [15]. These monitoring techniques offer several advantages over conventional wet chemistry techniques including non-invasiveness, little or no sample preparation and rapid measurements [16], [17]. Recent examples where process monitoring has been utilized include raw material dispensing, chemical reactions, granulation, drying and powder blending [18], [19], [20], [21], [22], [23], [24]. Similarly, a PAT (process analytical technology) based monitoring is illustrated from an industrial perspective by Chandra et al. [25]. Moreover, several works also incorporate NIR for feedback process control [26], [27]. All of these applications show unique advantages and feasibilities of using vibrational spectroscopy as real time PAT tools.

The –OH stretching vibration is a very strong absorption band in the NIR region and is thereby used for water content measurement, resulting in the successful use of NIR in food and chemical manufacturing [15], [28], [29], [30]. Analysis of samples in the form of solids, liquids and pastes are simpler using NIR as sample preparation is minimal. Since moisture content plays a key role in the mulling process, this paper discusses the use of NIR as a suitable PAT tool in process monitoring. During the mulling process, material characteristics such as particle size distribution and powder surface properties such as smoothness and roughness could slightly vary. The effect of these variations on the moisture content measurement is still an open area of research. Accuracy and precision of measurement also significantly depends on the location and installment of the sensor into the manufacturing plant. Therefore, prior to sensor integration, optimal conditions such as sensing location distance between sensor and sample, characteristics of sensing interface, and data acquisition specifications need to be identified. However, because of the unavailability of a systematic method, the identification of the optimal sensor conditions is still based on a heuristic approach.

Critical operation parameters of the NIR spectrometer such as scan number, acquisition time, distance between sample and detector, sample thickness and sensor locations in order to obtain accurate measurements have been identified in this study. The effect of particle size and water content of the particles on NIR measurements have also been investigated since they change the physical properties of material surface which would impact the NIR absorption. The methodology discussed is generic and can be used for any combination of process and PAT tools.

Section snippets

Materials and equipment

A specific grade of alumina, CATAPAL boehmite alumina, was used in this study, which was manufactured by Sasol, and the distilled water was generated by a Millipore Milli-Q Water Purifier.

A Design of Experiment (DOE) analysis was performed using the Minitab 16 statistical software. For the calibration model, the alumina and water were mixed in a vortex mixer from VWR International. The batch mulling experiments were carried out in a KG-5 high shear granulator from Key International, Inc. and

Pre-treatment of raw spectra data

The plots of the raw spectral data collected by varying the working distance between sample and spectrometer, are given in Fig. 5. In the NIR region, the –OH band is usually observed clearly at 1895 nm (–OH stretch free water) and 1428 nm (OH 1st overtone) [31]. Based on this, a wavelength range from 1399 to 1493 nm was selected for water content analysis. As shown in this figure, the absorbance value decreases as the distance between sample and detector increases. Also, it can be seen in the

Conclusions

The possibility of application of near infrared spectroscopy, a technique used for substance discrimination and quantification based on perpetual molecular motion, on batch mulling is discussed. To achieve the above goals, adequate calibration models are necessary. Many models for calculating liquid content at different conditions were created. All the liquid content models passed the validation test and have higher R-square and low RMSE values, which indicate adequate and reliable models.

The

Acknowledgements

This work is supported by Rutgers Catalyst Manufacturing Science and Engineering Consortium. The authors would like to acknowledge SASOL for the donation of their CATAPAL B alumina. The authors would also like to acknowledge Krizia M. Karry for useful discussions and Yinmingzi Liu, Aditya Donepudi and Meera Patel for their generous help in performing experiments.

References (31)

Cited by (14)

  • A comparison of benchtop and micro NIR spectrometers for infant milk formula powder storage time discrimination and particle size prediction using chemometrics and denoising methods

    2022, Journal of Food Engineering
    Citation Excerpt :

    However, micro NIR spectral noise if not removed can mask spectral features of the measured samples. Spectral noise in the micro NIR spectra originated from the limitations of the hardware configuration and measurement conditions (Manley, 2014; Wu et al., 2016). In order to employ micro NIR spectrometers as PAT tools, there is a need to embed denoising algorithms in spectral data analysis software.

  • Flow of granular materials in a bladed mixer: Effect of particle properties and process parameters on impeller torque and power consumption

    2018, Advanced Powder Technology
    Citation Excerpt :

    The torque required to turn an impeller in a particle bed depends on a number of factors, including blade design, agitator rotational speed, material fill level, and mixer size [15–17]. In the case where agglomeration is being carried out in a high shear mixer, torque measurement helps in control and monitoring of the granulation process [18–20]. Since torque is sensitive to changes in operating conditions, with rapid measurement and good precision, it can be used as a parameter to probe the efficiency of mixer designs [18].

  • Critical review of real-time methods for solid waste characterisation: Informing material recovery and fuel production

    2017, Waste Management
    Citation Excerpt :

    Ever since the introduction of PAT, the use of advanced technology in the pharmaceutical industry has increased (Dickens, 2010; Simon et al., 2015) with multiple examples of successful implementations (Chen et al., 2011; Mantanus et al., 2011; Reich, 2005; Rosas et al., 2012; Wu et al., 2011). Process analysis concepts and tools are transferable (Simon et al., 2015), and activities other than pharmaceutical have adopted the PAT approach, including the production of biofuels (Chadwick et al., 2014), biotechnology (Gomes et al., 2015), food production (Tajammal Munir et al., 2015; Troy et al., 2016; van den Berg et al., 2013), and manufacturing of catalysts (Wu et al., 2016). In all these examples, the production plants are managed using PAT tools to collect key process information to feed-back or feed-forward a central control system, improving the quality of the product in real-time.

View all citing articles on Scopus
View full text