Privacy settings

Utilizziamo cookies tecnici, necessari per il corretto funzionamento del sito, cookies statistici e cookies di marketing (inclusi cookies di terze parti), al fine di mostrarti offerte e pubblicità basate sui tuoi gusti e preferenze. Per maggiori informazioni puoi consultare la nostra Privacy Policy. I cookies statistici e i cookies di marketing (inclusi cookies di terze parti) sono utilizzati esclusivamente con il tuo consenso. Cliccando “Seleziona tutto e acconsenti” consenti il trattamento dei cookies statistici e di marketing (inclusi cookies di terze parti).

Per personalizzare le tue impostazioni privacy, seleziona le tue preferenze tra le caselle qui sotto e clicca “Applica la selezione”. Ti ricordiamo per puoi modificare le tue impostazioni privacy in qualsiasi momento e disattivare i cookies, seguendo le istruzioni riportate nella Privacy Policy.

Show preferences Select all and accept
IMA is loading

Demonstrating a Cloud-based Approach to Freeze Drying Data Analytics: Case Studies on Pharmaceutical Production Data.

Introduction

A complex, energy intensive, product sensitive and expensive process such as freeze drying requires constant real-time monitoring. This includes the use of various sensors and PAT tools to keep track of the process, product and equipment. However, in addition to real-time monitoring of freeze drying, a historical analysis of the data will provide a better insight into the overall system health. The vast amounts of data that is collected during a single production batch makes it is worthwhile to invest in developing and deploying algorithms specifically targeted towards analytics of the process and equipment.
With this premise, IMA Life has developed a health monitoring system that pairs a real-time notification UI with cloud based data analytics. The inputs to drive alerts and notifications on the real-time station are derived from algorithms that analyze the historical data on the cloud platform. Periodic updates are provided to users that improve the algorithms based on system performance & maintenance. In this paper, we will present some case studies on data acquired from three production freeze dryers in the U.S.

Cloud Analytics: Data Acquisition and Mo-del Development

To handle the vast amounts of incoming data and provide rapid computational capabilities with minimum latency, the analytics are hosted on Amazon Web Services. This platform provides a streamlined access to multiple users, while allowing flexibility to monitor multiple freeze dryers at a given site simultaneously. The incoming sensor tags are separated according to their position or the type of ancillary equipment they are on. This allows ease of analytic development through the use of a model tree as shown in Figure 1. In the next few sections, we will present our finding from three different analytics deployed on production freeze dryer data.

Case Study 1: Pressure Fluctuations and Condenser Overload

The refrigeration system in a freeze dryer is responsible for maintaining accurate shelf and condenser temperatures. In addition to providing the heat to the shelves during drying, it is responsible for maintaining the temperature of the condenser below a certain limit. Any fluctuations in the condenser temperature may lead to an increase in the condenser pressure due to the additional vapor pressure of water. Although there are quite a few variations in the construction of freeze dryer refrigeration systems, they may be broadly classified into two categories – Liquid Nitrogen or Compressor based systems. While

Figure 1. Production Data Model Tree for Cloud Data Analytics

they differ significantly in operation and utilities, the objective remains the same.
This case study involves a direct injection liquid nitrogen based system, where a total of 9 condenser coils are connected to a single source of LN2. In normal operation, the amount of liquid nitrogen injected into the condenser coils is carefully controlled through the use of modulating valves to maintain a temperature setpoint. Due to some variations in supply pressure and the volatile nature of liquid nitrogen, some fluctuations in the condenser pressure are expected. However, when enough liquid nitrogen has been injected into the coil, the temperature will stabilize. Figure 2 shows a drying cycle under normal conditions, where the blue line is the chamber pressure which is to be controlled at a setpoint of 0.1mbar and the red line is one of the condenser coils. It can be seen that there is no loss in chamber pressure control for this cycle.
In some cases, the fluctuations in condenser temperature may either never stabilize or take too long to reach equilibrium. As there is an influx of water vapor from chamber to the condenser, a coil temperature variation may lead to an increase in condenser pressure. A systematic increase in condenser pressure may eventually lead to increase in chamber pressure. A deviation in chamber pressure can lead to loss of process control, eventually impacting the product quality. Thus, pressure fluctuations that exceed a certain threshold may even lead to a production batch being rejected. Figure 3 shows one such cycle, where excessive variations of the condenser temperature caused chamber pressure fluctuations.

Figure 2. A freeze drying cycle showing the process under control

Typically, a condenser temperature high alarm is triggered any time the temperature increases beyond – 40°C. However, from figure 3, it can be seen that the condenser temperature never crossed the – 40°C threshold before the first instance of pressure excursion. This means that a fixed setpoint based alarm will not provide any benefit and when the alarm sounds, it will be too late as a pressure excursion has already occurred.
The analytic that has been developed around this phenomena for the freeze dryer under consideration takes these condenser coil variations into account. There is a significant lead time to failure which can be used to alert the user of possible pressure excursions. In this instance, an alarm could have been generated using this model up to 5 hours before failure. Based on this lead time to failure, the freeze dryer can be put in a “product protection mode” to avoid a pressure fluctuation based failure before they actually occur.

Case Study 2: Vacuum Pump Performance

The vacuum group in a freeze dryer is used in a cycle for evacuation to the pressure setpoint and maintaining the pressure during drying. Most production freeze dryers today have two primary vacuum pumps and may have additional booster pumps based on the equipment size. The vacuum pumps are constantly running throughout the freeze drying cycle and the pressure is controlled by injecting nitrogen or sterile air into the product chamber.

Figure 3. Chamber Pressure Fluctuations caused due to Condenser Coil Temperature variations

Hence it is important to maintain all the vacuum pumps in good working condition. However, due to various causes, there may be a decline in vacuum pump performance which may eventually cause a failure. These performance drops will typically manifest in the form of poor vacuum control or an increase in time required to evacuate the chamber to the same pressure setpoint. We have developed an algorithm to automatically perform “stage identification” with time series PLC data from the freeze dryer that allows us to identify not just evacuation, freezing, drying etc but also differentiate in real-time primary drying from
secondary drying. Thus, by trending the evacuation time using the stage identification algorithm, any cycle to cycle deviations can be recorded easily.
Here we compare three production freeze dryers of different sizes, each running more than one product recipe. Since each freeze dryer’s vacuum pumps are sized according to the chamber volume, the evacuation times are expected to be similar.

Figure 4 shows the evacuation time trended for the three freeze dryers. It was found that on average a typical evacuation cycle takes 36 minutes, while the data from Lyo3 between October 2018 and May 2019 shows that the evacuation time has been greater than 70 minutes. Unfortunately, most freeze dryers do not trigger an alarm today for excess time spent in evaluation. This eventually led to a vacuum pump failure in June 2019 on Lyo3. These evacuation time outliers on Lyo3 started 15 cycles prior to the pump failure. Careful monitoring of such a trend, and deviations from expected evacuation times may have alerted the operators of the upcoming failure.

Figure 4. Trending evacuation time for 3 freeze dryers shows anomalies for Lyo3

Case Study 3: Equipment Utilization

Scheduling production and maintenance on a freeze dryer is of critical importance to ensure throughput is consistent. For a sterile operation such as freeze drying, there are a number of operations following a product recipe that need to be performed to bring the filling line back to a stage that is safe for loading. These cycles, collectively referred to as turnaround cycles, consist of freeze dryer defrost, Clean-in-place, Sterilize-in-place, Leak Rate and Filter Integrity Test. Tracking the total time spent in turnaround cycles along with freeze drying will provide pharmaceutical manufacturers and their production managers, the cumulative up/downtime for a particular freeze dryer or a given site. The uptime is the time spent in a freeze drying product recipe while downtime is the idle time on a freeze dryer.
This analytic, Equipment Utilization, is able to read the status of the freeze dryer along with the time series data and breaks down the amount of time spent in different cycles. There are a number of ways to view this data. Figure 5 shows the 14 day rolling average breakdown for uptime and turnaround time. This 14 day rolling average is updated every 12 hours to provide the user with a short term performance picture. Tracking this can alert the operators if any modifications in the schedule need to be made based on unscheduled equipment downtime.
In many manufacturing sites, production batches are run in a campaign, with each campaign having a target throughput. Visualizing the long term, i.e. monthly and annual utilization can help management organize their production schedules. Figure 6a shows the monthly, while Figure 6b shows the annual equipment utilization. A scheduled preventive maintenance (PM) on the entire facility, including the freeze dryer, is usually done every 6-12 months. This can be clearly seen in Figure 6a. Trending this throughout consecutive years will also give the operators an estimate of the number of production cycles that need to be scheduled following a PM. Annual utilization plots can track the growth of a particular drug product through the years, as seen in Figure 6b. This may be beneficial in determining total equipment time required for upcoming years’ production batches.

Figure 5. 14 day Rolling Average Uptime and Turnaround Times for certain time intervals

Figure 6a (top) Monthly Utilization and 6b (bottom) Annual Utilization

The interest in connected Machines, the use of Machine Learning & AI is growing rapidly in the pharmaceutical industry, and for aseptic manufacturing in particular. In the preliminary stages of this project, we found several applications of data analytics and health monitoring in different industries. The use of monitoring stations in brewing, multiple sensors and data analytics for jet engines and highly customized dashboards in modern cars are some of the day-to-day applications of Industry 4.0.
Although it has taken a while for the freeze drying industry to adapt to the current trends, these recent developments that are being driven by an understanding of the process and equipment will certainly benefit today’s pharmaceutical manufacturing industries. The goal of this project from IMA Life is to deploy such models to detect process deviations or equipment failures before they occur and prevent unnecessary freeze dryer downtime from over-maintained machines. The next steps will be geared towards continuous model updates and building additional analytics in conjunction with deployment of these advanced analytics on production freeze dryers around the world.

Download Case study

Contact us

For any further information,

qdebug
Array
(
    [main_title] => 
    [stripe_view] => all
    [type] => thread
    [threads] => Array
        (
            [0] => case
        )

    [count] => 4
    [terms] => 
    [buttons_on] => 1
    [buttons_items] => Array
        (
            [0] => Array
                (
                    [acf_fc_layout] => link
                    [text] => View all updates
                    [icon] => list
                    [internal_link] => 1
                    [link_page] => https://ima.it/pharma/updates/
                    [tab] => 0
                )

        )

    [font_color] => 
    [background_color] => 
)