Daniel Ingo Hefft and Dr Federico Alberini of the University of Birmingham describe a new technology for the in-line measurement of complex rheological properties of liquid foods during processing operations.
Reducing production losses
Advanced Measurements for Industrial Application (AMFIA) is a new research group within the School of Chemical Engineering at the University of Birmingham. This group has been specifically designed to address industry needs to reduce preventable production losses and to improve process performance and monitoring. The group’s speciality area lies within mixing/blending of fluids, however, the activities also extend to projects within the aerospace and pharmaceutical sector.
New/existing product development is an ongoing activity in any fast moving consumer goods (FMCG) environment. Key for any development is scaling up from a pilot to full-scale production. Traditional approaches often only focus on pilot scale development without paying sufficient attention to a formulation’s manufacturability. This can often make it very challenging for operators and process engineers to deliver an optimal product quality and consistency and often involves post launch changes and adjustments. Such changes can mean additional on-costs and a reduction of line rates.
These critical challenges also limit the ability to achieve more efficient and flexible processes. A way to overcome these hurdles is to use an Industry 4.0 approach and in situ measurement techniques. When it comes to fluid handling, complex rheological characterisation is one of the most desired measurements in industry as the micro and macro structure of formulated products is strongly related to their rheological properties. Currently there are only a limited number of products that can deliver such measurements live and on a non-invasive basis.
When it comes to fluid handling, complex rheological characterisation is one of the most desired measurements in industry as the micro and macro structure of formulated products is strongly related to their rheological properties.
Detection and quantification of flow
Currently, industry mainly uses electromagnetic flow meters and Coriolis flow meters for rheological measurements. The disadvantages are obvious with electromagnetic flow meters, which in general are only calibrated on water. Coriolis flow meters often struggle with multiphase flow. Technologies, such as Ultrasound (Doppler) Velocimetry or other systems based on active acoustic emission, are high in energy consumption, costly and limited in their penetration depth or their ability to measure multiphase systems.
For complex rheological measurements, there are even fewer options available. In fact, most of the in-line or in situ rheological measurement techniques that are available in the marketplace can only describe the viscosity of a fluid at a fixed shear rate. The non-linear relationship between shear rate and shear stress, means that viscometer measurements cannot provide reliable information for the characterisation of complex fluid rheology. Much more reliable off-line measurement, such as the cone and plate rheometer, are used to acquire a detailed flow ramp (shear rate vs shear stress curve). However, this approach has the drawback of the length of time required for measurement that can easily exceed 30 minutes, which is well beyond the desired live measurement timeframe. Also, such rheometer measurements require line sampling, meaning the potential introduction of product contaminants. An off-line measurement can only deliver an indication of the product/process performance, but will not deliver a true picture of the state of the fluids in the pipe. Other modern technologies, such as Electrical Resistance Rheometry (ERR) work well, however, they are only applicable for conductive fluids, limiting their ability to offer an easy off the shelf solution.
In-line measurement of rheology
One of AMFIA’s most recent innovations has been in the in-line measurement of rheology. The basic idea is to move away from a traditional rheology measurement and use an Industry 4.0 approach. The group has found a way to measure rheology live amongst other factors within a processing pipe. The patent pending technology (GB1909291.5), named Rheality, is designed as a plug and play system and works for any single and multiphase fluid passing through a pipe.
The technology works on a combination of transient energy release measurements and machine learning. The fluid passes through a specifically designed pipe segment, which causes it to release transient energy. This is detected by a piezoelectric sensor on the outer wall of the pipe segment. With the help of self-developed algorithms and statistical analysis, the incoming data is reduced by over 99% and selected features are given to machine learning. The machine learning algorithms are trained to predict the following features on a live basis:
Relative rheology,
Blockages and leakages in the pipe system,
Solid/gas content for multiphase systems,
Flow and flow rate.
Rheality consists of two main components, which are a specifically designed pipe segment and the piezoelectric sensor kit. The signals detected by the sensor are complex, have very high information content and many data points. The data is converted in a first step from a time signal into a frequency spectrum (Fourier transform) (Figure 1).
Image may be NSFW. Clik here to view.
Figure 1 Transformation of a time signal into its frequency domain (FFT)
In order to reduce the number of data points, in a first step, self-developed filters are applied. Further data point reductions are based on statistical analysis algorithms, the creation of signal rankings and point selection and finally a principal component analysis. This initially simplifies between 1,000,000 and 500,000 raw points to 5,000 to 1,000 data points.
The last step is the use of supervised machine learning. As a learning algorithm, the quadratic SVM is used, which can solve classification problems. This type of machine learning is based on four key principles: the hyperplane, maximum margin hyperplane, the soft margin and the kernel function. Principle schemes for each of these are given in Figure 2.
Image may be NSFW. Clik here to view.
Figure 2 Principle scheme of the SVM machine learning algorithm.
Rheality is a non-invasive and non-destructive measurement method that can predict complex rheological problems and measure flow features live. Machine learning prediction accuracies are 95% or better (measured on previously unseen data sets). The system works both for single and multi-phase systems, as well as for laminar and turbulent flow.
Call to industry
AMFIA has been awarded an Innovate UK (ICURe Midlands) grant to further develop applications for this technology. The purpose of the grant is to visit companies in the FMCG sector in the UK as well as worldwide to explore opportunities to apply the technology. This will enable us to build industry collaboration and to give AMFIA the unique opportunity to get in touch with industrial partners. We wish to gain an understanding of how rheology and flow are measured currently, how important this information is for the production process and its impact on the process and product.
Daniel Ingo Hefft, Research Fellow and Dr Federico Alberini, Lecturer, University of Birmingham, School of Chemical Engineering, Advanced Measurements for Industrial Application
