Novel Real-Time Diagnosis of the Freezing Process Using an Ultrasonic Transducer

The freezing stage governs several critical parameters of the freeze drying process and the quality of the resulting lyophilized products. This paper presents an integrated ultrasonic transducer (UT) in a stainless steel bottle and its application to real-time diagnostics of the water freezing process. The sensor was directly deposited onto the stainless steel bottle using a sol-gel spray technique. It could operate at temperature range from −100 to 400 °C and uses an ultrasonic pulse-echo technique. The progression of the freezing process, including water-in, freezing point and final phase change of water, were all clearly observed using ultrasound. The ultrasonic signals could indicate the three stages of the freezing process and evaluate the cooling and freezing periods under various processing conditions. The temperature was also adopted for evaluating the cooling and freezing periods. These periods increased with water volume and decreased with shelf temperature (i.e., speed of freezing). This study demonstrates the effectiveness of the ultrasonic sensor and technology for diagnosing and optimizing the process of water freezing to save energy.


Introduction
The freeze drying process, also known as lyophilization, is a critical and well-established process for the long-term storage of food, drug, biopharmaceutical products, etc. [1]. Almost 50% of the marketed biopharmaceuticals are freeze dried, thus indicating that the freeze drying process is the most common formulation strategy [2]. The advantages of this process include better stability, easy handling and storage and good product quality [3].
The traditional freeze drying process is a dehydration process by sublimation of a frozen product, and consists of three steps: freezing, primary drying and secondary drying [4]. In the freezing stage, the liquid is cooled down until ice crystals start to nucleate and grow. Then the crystalline ice forms and is removed by sublimation in the primary drying stage. In this stage, the vacuum chamber pressure is reduced below the vapor pressure of ice, and the shelf temperature is raised to provide the heat for ice sublimation [5]. The remaining 15%-20% water inside the product would be desorbed at elevated temperature and low pressure in the secondary drying stage [6].
The freezing stage is the first and the shortest step of the freeze drying process, but it governs several critical parameters, such as the sublimation and desorption rates [7], and the final quality of the lyophilized product [8]. The initial/end freezing point, freezing rate and degree of supercooling in the freezing stage are important thermodynamic factors for the prediction of thermal and physical properties. Accurate freezing point data can be utilized to determine several properties, such as effective molecular weight, water activity, frozen water, enthalpy below freezing, construction of state diagrams [9], freezing and thawing of frozen foods, etc. The freezing point is also important to estimate the freezing time and other structural properties, such as glass transition, end point of freezing, and fraction of unfrozen water in foods. Due to this importance and its wide applications, the estimate of freezing point and modeling these properties are crucial in food processing (freezing and drying) and food stability during storage [10].
The complex interplay of these properties in the freezing stage requires the use of real-time process diagnosis and quality control procedures. Currently, the most widely used freezing process diagnostic tool deploys temperature sensors for online measurement [11]. The temperature sensors, however, need to be in direct contact with the frozen samples in order to provide the temperature information [12], but a sensor embedded in the product may cause removal inconveniences after the freeze drying process. For drug and biopharmaceutical products, the direct contact of a sensor with the products may cause undesired pollution. There are also several off-line measuring methods, such as theoretical thermodynamic calculations, cryo-microscopy, differential scanning calorimetry (DSC), thermo-mechanical analysis (TMA), dynamic mechanical analysis (DMA) and dynamic mechanical thermal analysis (DMTA) [13] for measuring the thermodynamic properties, glass transition, weight, heat flow, dimensional changes, as well as viscoelastic properties of the tested samples or products. All the off-line methods need very costly instruments and operator skill to measure and interpret the data.
The ultrasonic technique is a widely known non-destructive and non-intrusive method for real-time process diagnosis [14]. The basic signatures of ultrasonic signals, such as velocity, attenuation, reflection and transmission coefficients, scatter signals from materials, all have unique relationships with process dynamics [15], material characteristics [16], and product quality [17]. Ultrasonic signals can also reveal the temperature of a material [18][19][20]. Ultrasonic technology has also been adopted for decreasing the sublimation time by controlling the nucleation temperature during the primary drying stage [21], but it has not been utilized for freezing point diagnosis during the freezing process. This study attempts to apply ultrasonic techniques to the real-time diagnosis of freezing processes. In particular, an ultrasonic transducer (UT) is integrated onto the bottom of a freezing bottle. Then, during the process of freezing water at various cooling temperatures, the process is monitored and diagnosed by ultrasonic technology to evaluate the freezing point and cooling/freezing periods to optimize the process and save energy.

Development of Ultrasound-Assisted Diagnosis System
Since the operating temperature of most ultrasonic couplant is between −18 to 100 °C, and the space available for installing sensors in a freezing machine is limited, the development of UT is one of the key factors to realize the ultrasonic diagnosis of industrial freeze drying processes at freezing temperatures. The procedure for fabricating the UT by a sol-gel spray technique is described in previous publications [22][23][24]. The UT was applicable at temperature range of −100 to 400 °C without an ultrasonic coupler. It could be operated in a medium megahertz (MHz) frequency range with a sufficient frequency bandwidth and had a sufficient piezoelectric strength and signal-to-noise ratio (SNR). Figure 1a presents the utilized freezing bottle made of stainless steel. The height of the freezing bottle is 35 mm. Figure 1b shows the top view of bottle. The inner and outer radii of the bottle are 16 and 22 mm, respectively. The inner height and capacity of the bottle are 30 mm and 6.03 cm 3 , respectively. According to the size of the freezing bottle, the UT sensor was designed with a circular shape with the following dimensions: 8 mm in radius and 103 μm in thickness, as shown in Figure 1c. The top electrode was fabricated with silver paste with the radius of 4 mm. The UT sensor was well aligned on the center of the cavity. A schematic view of the freezing bottle with the UT sensor incorporated in the freeze dryer machine during the freezing process is displayed in Figure 2. As shown in Figure 2, when electric pulses were applied to the piezoelectric film through the top and bottom electrodes, where the freezing bottle itself served as the bottom electrode, ultrasonic waves were excited and transmitted into the freezing bottle. L n (n = 1, 2,…) denote the nth round trip longitudinal-wave ultrasonic echoes reflected from the interface of the freezing bottle/water or ice, and Lw is the 1st echo propagating in the water and reflected from the water/air interface. The L 1 and Lw echoes will be used to monitor the freezing process and water state. The height of water level and thickness of bottle bottom are denoted as h and d, respectively. A temperature sensor (Type T thermocouple, Omega, Stamford, CT, USA) was set in the middle of the freezing bottle for measuring the water/ice temperature. The temperature would be measured for a comparison with the ultrasonic signals during the freezing process.   Figure 3a shows the typical ultrasonic signals acquired with the UT in Figure 2. As one can see, the L n echoes (n = 1, 2, …), reflected at the bottle/water or ice interface, appeared at 0.81 and 1.87 μs, respectively, and remain during the entire process. When the water was not frozen, the Lw echo, propagating in the water and reflected at the water/air interface, was observed at 36.39 μs. The time delay difference between the L 1 and Lw echoes was denoted as Δt. Figure 3b shows the frequency spectrum of the L 1 echo in Figure 3a. The center frequency of the L 1 echo was 8.51 MHz, and the 3-dB bandwidth was 8.70 MHz. The SNR for the first round trip echo, L 1 , was 37.6 dB. The SNR value could be calculated according to the following equation: where P and A are the power and amplitude of signals, respectively.

Experimental Setup
In this experiment, a 4 L, air-cooling type shelf freeze dryer machine (TYFD-50005, Tai Yiaeh, New Taipei City, Taiwan), equipped with vacuum chamber, refrigeration and control units, as shown in Figure 4a, was used. The vacuum chamber was for freezing and drying the samples under low temperature (+50~−40 °C) and pressure (760~0.05 torr) conditions. The steel freezing bottle with UT and thermocouple was set on the shelf of vacuum chamber during the freezing process. The refrigeration unit was for the refrigeration and heat exchange processes of the refrigerant and antifreeze. The control unit comprised a programmable logic controller (PLC) and a human-machine interface for controlling the freeze drying process. The measured temperature data from the thermocouple was recorded by the PLC control unit every second during the freezing process. Figure 4b presents the utilized digital oscilloscope (DSO-X2014A, Agilent Technologies, Santa Clara, CA, USA) for recording the ultrasonic signals during the freezing process. The digital oscilloscope has four channels, 100 MHz bandwidth, a maximum sampling rate of 2 GSa/s and a maximum memory depth of 100 kpts/channel. Figure 4c presents the utilized pulser/receiver (5072PR, Olympus, Tokyo, Japan), which is a broadband, negative spike pulser and broadband receiver. It can be applied in reflection or transmission mode. For the pulser, the pulse voltage under no load is −360 V and the pulse repetition rate is 100~5000 Hz. For the receiver, the broad bandwidth is 1 kHz~35 MHz, with high/low pass filters of 1 kHz~1 MHz and 10~35 MHz, respectively.
The freezing bottle was filled with a certain amount of water at room temperature, and then installed on the shelf of the freeze dryer machine. The freezing bottle with water would be cooled down to the shelf temperature. After the temperature of the freezing bottle with ice reached a stable state, the experiment was stopped. The designed water levels were 5, 15 and 25 mm. The temperature settings of the shelf were −20, −30 and −40 °C. The air pressure of the vacuum chamber was 101.3 kPa (1 atm). The liquid utilized to fill the freezing bottle was water. All the experiments presented in this study were conducted in the ultrasonic pulse-echo mode. The ultrasonic signals were acquired every 5 s in this paper.

Freezing Process Diagnosis by Temperature Measurements and Visual Observations
Temperature variation during the freezing process has a close relationship with the rate of freezing, and it would affect the size of the ice crystal nuclei and the quality of the freezing process. In order to investigate the correlation between the water temperature and the water freezing process, the water temperature variation inside the freezing bottle with respect to the process time was determined. The water was filled to a level of 25 mm in the freezing bottle at room temperature and an air pressure of 101.3 kPa. The water and shelf temperatures recorded by the temperature sensors are shown in Figure 5. The shelf temperature was set at −30 °C throughout the entire process. At a process time of 3.2 min, the freezing bottle was put into the freeze dryer machine and the temperature of water started to drop. This point was denoted as point A. From 3.2 to 6.9 min, the water temperature was cooled down from 24.1 to 1.4 °C, with a rate of decrease of 6.14 °C/min. During this period, the sensible heat of water was removed. This period is denoted as the cooling period, ΔPCT, measured by the thermocouple. From 6.9 to 15.2 min, the water temperature was kept within the range of 0.7 to 2.9 °C. During this period, the latent heat of water was removed through the bottle/shelf interface, and the water would change phase from liquid to solid (ice). The points at the process times of 6.9 and 15.2 min would be denoted as points B and C [11], respectively, indicating the freezing point and phase change end of water. This period is also denoted as the freezing period, ΔPFT, measured by the thermocouple. Then, from 15.2 to 40 min, the water temperature was cooled down again from 1.2 to −27.6 °C for further remove the sensible heat of ice. The freezing process stopped at the process time of 40 min when the ice temperature was −27.6 °C.

Freezing Process Diagnosed by Ultrasonic Signatures
Even though the water freezing process is familiar to most people, however, the diagnosis of this process is typically limited to visual observation and temperature methods. According to the authors' knowledge, the use of ultrasound technology to monitor the freezing process may be rare. In order to investigate the correlation between the ultrasonic signals observed and the water freezing process, the amplitude values of the L 4 and Lw echoes in Figure 3a with respect to the process time were obtained. The results are presented in Figure 7. The reason of choosing ultrasonic echo L 4 , instead of L 1 , is that the L 4 echo is more sensitive to the variation of steel/water interface. The amplitude variations of the ultrasonic L 4 and Lw echoes, corresponding to the freezing process in Figure 6, are described as follows: (1) Process time of 2.75 min, temperature of water 24.2 °C: the freezing bottle was filled with water to a level of 25 mm at room temperature (25 °C) and an air pressure of 101.3 kPa. The shelf temperature was set as −30 °C. At this moment, the amplitude of the L 4 echo decreased and the amplitude of the Lw echo increased, due to the fact that a part of the ultrasonic energy was transmitted into the water through the steel bottle/water interface. During the freezing process, the speed of freezing of water would affect the size of ice crystal nuclei and the frozen water quality. Ultrasonic velocity may be one of the candidates to indicate the freezing speed of water, because of its close relationship with the temperature. The ultrasonic velocity in the water could be calculated according to the following equation: where h is the height of the water level in Figure 2 and Δt is the time delay between the ultrasonic echoes L 1 and Lw in Figure 3a. The result is shown in Figure 8.
In Figure 8,

Effect on Freezing of Various Water Amounts
A linear relationship between the cooling/freezing period, indicated by the ultrasonic signature, and the water level would be a fundamental requirement to utilize this technology. To study this relationship, the shelf temperature and air pressure were set at −30 °C and 101.3 kPa, respectively, to freeze various water levels of 5, 15, 25 mm. The experimental results of water temperature and amplitude of the ultrasonic L 4 echo related to water levels of 5, 15, 25 mm ae shown in Figures 9 and 10, respectively.   The corresponding temperature and timing of water-in, freezing point and phase change end indicated by temperature are listed in Table 1. It seemed that there was a linear relationship between the timings of freezing point/phase change end and the water level.   Table 2. It seemed that there was a linear relationship between the timings of freezing point/phase change end and the water level. In order to view clearly the mentioned linear relationship, the cooling/freezing periods were compared with the water level. The results are shown in Figure 11. The cooling/freezing periods were marked by square and circle symbols, respectively. Those indicated by temperature and ultrasound were marked by black and red color, respectively. The estimated errors of cooling/freezing periods for the experimental conditions were less than 3%. In the water level range from 5 to 25 mm, the average cooling/freezing periods indicated by ultrasonic L 4 echo increased from 1.73 to 4.73 min and from 6.28 to 11.68 min linearly, respectively. The straight line for these symbols was obtained by a least squares fitting method, which is explained in the following section. The slopes of the fitting lines were 0.15 and 0.27 min/mm for the cooling/freezing periods, respectively. The cooling/freezing periods can be expressed as: where ΔPCUT and ΔPFUT are the cooling and freezing periods in Figure 10, respectively, and h is the water level in Figure 2. In the water level range from 5 to 25 mm, the average cooling/freezing periods indicated by temperature increased from 3.19 to 4.39 min and from 3.12 to 8.72 min, respectively. The slopes of the fitting lines were 0.06 and 0.28 min/mm for the cooling/freezing periods, respectively. The cooling/frozen periods can be expressed as: where ΔPCT and ΔPFT are the cooling and freezing periods in Figure 9, respectively. This indicated that a higher level of water would result in a longer cooling/freezing periods. The different slopes of the cooling and freezing periods were due to the various specific heat capacity caused by the latent and sensible heat of water. Therefore, the ultrasonic technique can clearly indicate the cooling/freezing completion at each water level to shorten the freezing process and save energy.

Freezing Effect of Various Freezing Speeds
The freezing speed would affect the size of ice crystal nuclei and the frozen ice quality. We were interested in evaluating the effect of freezing speed on cooling/freezing periods indicated by ultrasonic signatures. To study this effect, the water level and air pressure were set at 25 mm and 101.   When the shelf temperature was low, the timing of the phase change end indicated by temperature was earlier. The cooling/freezing periods of various experimental conditions were also indicated. The corresponding temperature and timing of water-in, freezing point and phase change end indicated by temperature are illustrated in Table 3.   Table 4. In order to clearly view their relationships, the cooling/freezing periods were compared with the shelf temperature. The results were shown in Figure 14. The cooling/freezing periods were marked by square and circle symbols, respectively. Those indicated by temperature and ultrasound were marked by black and red color, respectively. The estimated errors of cooling/freezing periods for the experimental conditions were less than 3%, respectively. In the shelf temperature range from −40 to −20 °C, the cooling period indicated by temperature and ultrasonic L 4 echo increased from 4.8 to 3.4 min and from 4.72 to 4.12 min linearly, respectively. The slopes of the fitting lines were 0.07 and 0.03 min/°C for temperature and ultrasonic L 4 echo, respectively. The cooling period can be expressed as: ∆P = 6.20 + 0.07 × T (8) ∆P = 5.32 + 0.03 × where ΔPCT and ΔPCUT are the cooling period in Figures 12 and 13, respectively, and T is the shelf temperature. In the shelf temperature range from −40 to −20 °C, the freezing period indicated by temperature and ultrasonic L 4 echo increased monotonically from 5.39 to 15.63 min and from 10.91 to 18.92 min, respectively. The freezing period seemed reaching a saturated value when the shelf temperature was less than −30 °C. This indicated that the higher shelf temperature would result in the longer cooling/freezing periods. Therefore, the ultrasonic technique can clearly indicate the cooling/freezing completion at each shelf temperature for reducing the freezing process duration and saving energy.

Conclusions
The freeze drying process is a critical and well-established process for long-term storage of food, drugs, biopharmaceutical products, etc. The freezing stage is the first and the shortest step of the freeze drying process, but governs several critical parameters and the final quality of the lyophilized product. The ultrasonic transducer technology is one of the most effective and efficient tools for real-time, non-intrusive and non-destructive monitoring. In this study, an integrated UT was utilized in a stainless bottle for real-time diagnosis of the freezing process of water. The sensor was directly deposited onto the stainless bottle by using a sol-gel spray technique. It could operate at a temperature range from −100 to 400 °C and uses an ultrasonic pulse-echo technique. The progression of the freezing process, including water-in, freezing point and water phase change end, was clearly observed using ultrasound. The ultrasonic signals could indicate the three stages of the freezing process and evaluate the cooling and freezing periods under various processing conditions. The temperature was also adopted for evaluating the cooling and freezing periods. The cooling/freezing periods, indicated by ultrasonic signals and temperature, increased with water volume with the ratio of 0.15 and 0.27 min/mm and 0.06 and 0.28 min/mm, respectively. The cooling period, indicated by temperature and ultrasonic L 4 echo, decreased with shelf temperature with the ratio of 0.07 and 0.03 min/°C. The freezing period seemed to reach a saturation value when the shelf temperature was less than −30 °C. This study demonstrates the effectiveness of the ultrasonic sensors and technology for diagnosing and optimizing the process of water freezing to save energy.