Data Acquisition and Recording

The data in a GMP monitoring system follows a defined lifecycle. In accordance with GMP requirements, the integrity of data and records must be maintained throughout their entire lifecycle. Moreover, the data must be representative of the process to which it relates.

The first stages of the data lifecycle in a GMP monitoring system include:

Reading
Conversion (if necessary)
Recording

During these stages, a set of raw GMP-critical data is formed, which can be used in subsequent data processing and decision-making processes.

Data Reading

One of the most critical stages in ensuring data representativeness is the reading process. Key factors that influence representativeness at this stage include:

Sensor Placement
Reading Intervals/Times

For example, if a temperature sensor is placed in a "dead zone" in the corner of a cleanroom, where there is insufficient air mixing, the temperature readings may not represent the overall air temperature, especially in the working zone where the air environment might directly affect the product. Therefore, sensors should be installed at a certain distance from corners and ceiling joints – typically at least 15 cm away from cleanroom walls (30 cm is recommended). The decision on sensor placement should consider airflow characteristics and be based on a risk assessment.

Reading intervals/times are also crucial for ensuring representativeness, especially when monitoring parameters that can fluctuate rapidly, such as static differential pressure or pressure cascades in rooms. If the interval between differential pressure readings is too long (e.g., more than 10 seconds), the monitoring system might fail to detect sudden pressure spikes that occur between sensor readings. This could have serious consequences, such as contaminated air entering the working area of an aseptic process due to a short-term pressure cascade failure that goes unnoticed. On the other hand, reducing the reading interval increases system load and the size of the collected data set, which is limited by the capabilities of the GMP monitoring system in use.

Some manufacturers integrate an averaging system in sensor transmitters or input modules, where the transmitter polls the sensor's sensitive element at short intervals, calculates an average value, and transmits this to the monitoring system at longer intervals, such as every minute. This data collection principle may be acceptable for inert parameters like temperature and humidity, but it is entirely unsuitable for dynamically changing parameters like differential pressure.

Sensor Accuracy and Data Transmission

Another key aspect of data reading is the accuracy of the sensor and protection against data distortion during transmission to the monitoring system. According to GMP requirements, both the sensor and the entire data transmission chain to the monitoring system must be calibrated.

Typically, sensors measure parameters using a sensitive element that changes its properties based on the measured physical property of the external environment or object. These properties could include electrical conductivity/resistance, capacitance, inductance, and other electrical characteristics of the sensitive element, which can be recorded by an electrical circuit. Thus, an analog electrical signal is initially generated, which the sensor can transmit for further processing by the system. Common analog signals generated by sensors include resistance (e.g., in thermocouples), voltage (e.g., 0–10V), and current (e.g., a 4–20mA current loop).

Unlike analog signals, digital signals with built-in data integrity protocols are protected against distortion during transmission. Therefore, the analog signal from the sensor should be converted into a digital signal as close as possible to the sensor's installation point.

Data Conversion

In a GMP monitoring system, the analog signal (voltage, current) must be converted into a digital signal for processing. Typically, the first digital signal received from transmitters or input modules is a discrete integer value, often represented as a 16-bit word (two bytes). This corresponds to a range of 0000 to FFFF in hexadecimal or 0 to 65535 in decimal. The extreme values of this range usually represent the measurement range of the corresponding sensor/transmitter (e.g., -40°C to +80°C).

To obtain the parameter value expressed in physical units, such as °C, the initial discrete signal (0 to 65535) needs to be converted using a formula of the general form:

y = ax + b

Where: a and b are conversion coefficients, x is the initial discrete value, y is the converted value expressed in physical units.

This process is commonly referred to as "scaling".

In cases where the sensor/transmitter is designed so that, in the event of a malfunction (e.g., a broken cable), the monitoring system receives one of the extreme discrete values (0 or 65535) – a typical scenario for most commercially available products – there is a risk that the system will not detect the malfunction. Instead, it may record extreme values within the sensor's measurement range. To mitigate this risk, the GMP monitoring system software should implement a slight "narrowing" of the sensor's valid measurement range. Values falling outside this narrowed range should trigger an error message to indicate a sensor malfunction.

Therefore, the data conversion process must yield not only the parameters expressed in physical units but also the measurement channel status, indicating whether the channel is "operational" or has a "sensor error".

Data Recording

The converted data and measurement channel statuses must be reliably recorded and stored. In compliance with GMP requirements, data integrity and records must be ensured. This includes adherence to the guidelines outlined in Annex 11 of the EU GMP Guide (and corresponding national guides) and 21 CFR Part 11 of the FDA regulations.

These regulations outline the following principles:

Data Integrity:
All data must be attributable, legible, contemporaneous, original, and accurate (ALCOA principles).
Electronic Records:
The system must ensure that data is securely recorded in a manner that prevents unauthorized alterations or deletions.
Audit Trail:
All modifications, including creation, modification, or deletion of records, must be traceable, with a complete audit trail that tracks all user actions.
Data Backup and Retention:
Systems must implement secure, redundant data storage solutions to ensure that records are available for the entire required retention period.
Access Control:
Only authorized personnel should have access to the system, and access levels should be clearly defined to prevent unauthorized changes to records.

By complying with these principles, the system ensures that the data it records remains valid, secure, and available for regulatory review throughout its lifecycle.

Tarqvara GMP Monitoring System

The Tarqvara system utilizes innovative principles for data acquisition, statistical processing, and recording. Data collection from all sensors (transmitters) is performed cyclically at a frequency of 1 second, during which the following operations are carried out:

Data collection modules (software) send requests to all local input modules (hardware) connected to the sensors.
The data received from the input modules is recorded in a raw data array.
A scaling/conversion module processes the data, and the resulting physical parameter values are recorded in the measured parameters array.

The data collection protocols, such as RS-485, Modbus/IP, and TCP/IP, include checksum verification to ensure data integrity. In the event of data collection failures, such as a communication breakdown between the server and the gateways, a system alarm is triggered.

The data collected every second is processed and, every minute, the following values are calculated and recorded from the 60 second-based parameters:

The first value
The average value
The minimum value
The time (in seconds) when the minimum value occurred
The maximum value
The time (in seconds) when the maximum value occurred

These six values allow the original set of 60-second readings to be accurately modeled while preserving all critical GMP parameters: the average, minimum, and maximum values for each 1-minute period, with the exact second for the minimum and maximum occurrences, as well as the actual values at the boundaries of the 1-minute periods. This ensures the registration of any spikes or deviations in the parameters, representing critical GMP events.

The six values for each channel are recorded every minute in two mirrored databases (for redundancy). Each minute's entry is accompanied by a coded checksum to verify its authenticity. Additionally, the data files are equipped with overall coded checksums. If any records are found to be compromised, the system automatically restores them using data from the duplicate database.

The data files are highly compact: the annual storage requirement for a 120-channel system, with full redundancy, is approximately 4-5 gigabytes, depending on the number of warnings or alarms generated. Therefore, copying the entire database for 10 years of system operation would only require about 25 gigabytes of storage capacity.

A utility program included in the system's package allows data files to be converted into an open format compatible with MS Excel.