Assessment of long-term sensor performance based on a large THM experiment in the HADES URL

A lot of experience is available with installing and using these types of sensors around the HADES URL. Problems were encountered for some upward boreholes that crossed a water-bearing layer (about 10 m above the level of the HADES URL). Because of the large water inflow and the resulting instability of the borehole, the casing with the instrumentation could not be installed in these boreholes. Apart from this issue, all other sensors could be installed at the planned locations, even for the longest instrumented borehole, which is 46 m long.

Installing instrumentation and determining the position of the sensors, however, becomes more difficult with increasing borehole length. The borehole is measured using a total station and a retro-reflector that is moved inside the borehole. This method requires an optical line of sight between both elements and can therefore only be applied to the borehole section where its deviation is smaller than the borehole diameter. The position of points beyond that depth needs to be determined by extrapolation, which creates larger uncertainties that could reach a few decimetres in the case of curved boreholes (in particular for the vertical position in horizontal boreholes). This can become an issue when assessing the measurements through model predictions as these assume a sensor with an exactly known position.

A few days to a few weeks after installing the piezometers, the filters were connected to the pressure transmitters. It was important to verify that the filter tubes were completely filled with water to avoid that entrapped gas would compromise the pressure readings. Sometimes the filter tubes were manually filled with water until the clay had converged around the filter. In other cases, the pressure transmitters were only connected after water outflow was observed from the capillary tubes.

All pressure sensors installed around the PRACLAY Gallery are still functioning and most sensors provide continuous and reliable measurements. Nevertheless, a few have been replaced with a sensor with an adapted measurement range. One reason was to cope with thermally induced pressure increases, while this range was decreased for some filters near the accessible gallery so that more accurate measurements are obtained (as accuracy typically depends on the measurement range).

All borehole piezometers are located in saturated Boom Clay, which is affected by the plastic deformation around the borehole (no packers or grouting are used for the installation). Due to its mechanical properties, this has no relevant influence on the hydraulic properties. Indeed, the properties can be characterized during the whole lifetime through permeability testing for instance. One sensor set has been installed in precompacted bentonite, where particular attention needs to be paid to some installation details, such as the saturation of the filter prior to installation.

Based on an initial review of these measurements and numerical simulations of the porewater pressures, they are considered to give a representative picture of the actual conditions (i.e. porewater pressure field disturbed by gallery excavation, drainage and heating) in the Boom Clay.

Some, however, show pressure readings that seem unreliable. This could be caused by water leakage into the filter or capillary tubes. These components are in the PRACLAY Gallery and Boom Clay where there are higher porewater pressures due to the Heater Test. During the initial heating phase, the porewater pressures are higher inside the PRACLAY Gallery than in the surrounding clay. It cannot be ruled out that water from the gallery is pushed into the borehole casing and leaks into the filters, for example through glued connections, which were established between the separate casing segments during the installation of the instrumented casing into its borehole. This is a clear example that instrumentation designed with an open and accessible environment in mind, such as the Connecting Gallery, cannot be implemented in conditions imposed by the PRACLAY in situ experiment. The experience gained with PRACLAY has provided some suggestions for improvement, such as sealing the inside of borehole casings. After all, an open (water-filled) volume should be avoided as much as possible as it can cause water circulation, thereby influencing temperature measurements and worsen the consequences of possible leaks in the casing (e.g. due to a non-hermetic assembly between two casing segments).

Another problem occurred for the sensors installed in the seal. These porewater pressure sensors are also equipped with a relative humidity probe. It appeared that the humidity probe was not designed for hermetically sealed environments, as water leakage occurred along the signal cable. Because of this setup, no hermetically closed system was obtained showing irregular porewater pressure measurements.

The complete setup – filter, capillary tubes and pressure transmitter – is sensitive to temperature variations. This can be seen in the porewater pressure readings of sensors whose tubes are close to the heater cables. Every few days these heater cables are alternately turned on and off. This affects the porewater pressure measurements of the sensors whose capillary tubes are next to the heater cables (Fig. 14).

The transmitters are regularly calibrated by applying a pressure cycle (pressure increasing from atmospheric pressure to maximum range, followed by pressure decrease). A pressure calibrator is used to perform this calibration at high accuracy. The transmitter outputs are recorded at 11 points in each direction. This results in a set of (signal output, pressure) measurements, on which a linear regression is applied, which can be considered as the ideal line for each sensor. Each calibration then allows us to determine how the actual sensor output deviates from this ideal line, showing some hysteresis and non-linearity (Fig. 15). The overall deviations are generally well within the accuracy limits (in this case 0.006 MPa). Detailed analysis allows us to identify quality differences between different brands, and even production lots (based on the serial number). The calibration formula that needed to be applied (to convert the transmitter output into a pressure) remained fairly stable over the years, as no clear trend or drift could be observed.

While in general the good performance of previous implementations of the piezometer boreholes was confirmed, its application in a thermal field showed some particular effects on the pressure measurements. The most important issue for application in a DGR context is, however, its invasive nature.

The installation of the total pressure devices went smoothly, as they were integrated mostly in general devices (e.g. flatjack based total pressure devices integrated at the deep end of the piezometer borehole casing). Recesses in the bentonite allowed flatjacks and miniature pressure sensors to measure the swelling pressure inside the seal.

Table 2 shows the number of sensors that failed for each sensor set, demonstrating the differences in performance between the different sensor types. For the sensors installed in the gallery lining, the environment clearly affects their performance too as only the sensors installed in the lining of the heated gallery section have failed. Most probably, the saturated environment and high water pressure (almost 3 MPa) in addition to the elevated temperature have caused these failures through water ingress at the cable connections. A design improvement would therefore be avoiding such connections in a saturated environment.

The impact of the temperatures and pressures also appears from the timing of failure. Most sensor failures occurred during the first three years of the PRACLAY Heater Test (Fig. 16).

Depending on the sensor set, the sensors were subjected to different conditions. While the measuring surface was usually in contact with the saturated clay, some sensors had their cabling submerged in the pressurized and heated environment of the PRACLAY Gallery.

Valuable measurements were obtained in the seal, where the swelling pressure (due to the bentonite hydration) could be monitored rather well.

Despite the high failure rate of the sensors in the lining of the heated section, some important observations could be made. These sensors showed that the orthoradial stresses in the lining segments decreased during the Heater Test. This contradicted what was earlier assumed, namely that these stresses would increase with increasing temperature.

Measurement of total pressure in the host rock through instrumented borehole casings remains an issue, as the drilling operation needed for the installation significantly alters the stress field, which, in contrast with the porewater pressure field, recovers much less with time to the original values.

The intrinsic accuracy of the sensors is mostly better than 1% of the full scale, which is more than sufficient considering the measurement uncertainty due to the installation.

While total pressure monitoring in the host rock still needs to be improved, useful applications are currently mainly situated at interfaces and in engineered barrier system environments.

Installation of both sensor sets went rather smoothly. For the embedded vibrating wire gauges, experience from a previous application allowed for a near-perfect installation, while resistive gauges were installed using established practices.

The vibrating wire strain gauges that had been installed in the lining of the Connecting Gallery had been performing very well (Dizier et al. 2023). Only two out of 270 gauges failed over a period of 20 years. This excellent performance could unfortunately not be reproduced by the strain gauges installed in the lining of the PRACLAY Gallery. By the end of the heating up phase the first strain gauges in the heated gallery section started to fail. Five months later, all gauges had failed (Fig. 17). On the other hand, however, almost all strain gauges in the accessible (non-heated) part of the PRACLAY Gallery are still functioning. Although no exact explanation could be given up to now, the failures are more than likely related to the sustained, elevated temperature.

The resistive gauges could be read using a dedicated setup for this type of sensors. Monitoring indicated the lack of mechanical stress build-up, which could be confirmed by visual observations.

The heated and saturated environment of the PRACLAY Gallery exposed the vibrating wire gauges to harsh conditions, of which the elevated temperature may have caused the sensor failures. In addition, cable connections in this environment could not be avoided and might have contributed to the failures. This (as for many other sensors located in this environment) is to be confirmed during the dismantling of the setup.

Mixed results were obtained for the strain gauges. For the vibrating wire gauges, the harsh environment in the heated part of the PRACLAY Gallery caused some deteriorating signals prior to failure, while clear signals and, hence, measurements are obtained in the accessible part of the gallery. Resistive gauges evolved from clear monitoring results to deteriorating values after several years.

Strain gauges are typically characterized by a bulk calibration factor instead of an individual calibration. Comparing the measurement amongst each other (in particular where redundant installations exist) and with model predictions, it appears that the accuracy is sufficient to obtain a reliable picture of the strain evolution.

Vibrating wire gauges can deliver reliable results for several decades. For elevated temperatures (in this case around 80°C), special designs are needed to guarantee longer-term performance. Resistive gauges on the other hand, are, even according to their manufacturers, not suited for measurement campaigns exceeding a few years.

Novel techniques for strain monitoring have popped up, most notably techniques based on fibre optics. These have not been applied in PRACLAY, but should definitively be considered for long-term monitoring in DGR conditions.

Due to the small size of most sensors (of which many were also integrated in other sensors), the installation proceeded overall in a smooth way without relevant issues.

A total of 379 thermocouples were installed. These can be grouped into five sets whose performance is summarized in Table 3. Similar to the total pressure cells, the failure rate strongly depends on the location. The thermocouples with wiring inside the heated section of the PRACLAY Gallery were most affected, with elevated failure rates of up to 77% for the thermocouples installed inside PG piezometer casing. The failure rate started to increase after the temperature reached the stationary level of 80°C (Fig. 18).

The performance of the thermistors is closely linked to that of the strain gauge sensors and flatjack sensors in which they are integrated. As a result, most sensors with wiring running inside the heated section of the PRACLAY Gallery eventually failed. This confirms that cabling and cable connections are the most vulnerable part in saturated and pressurized conditions. Furthermore, the RH sensors cannot cope with saturated conditions and failed when their seal became saturated. When these sensors failed, also the temperature sensors integrated in them stopped working. The inclinometer temperature sensors are still all functional.

The device used to read out the optical fibres (Brillouin optical time domain reflectometer) did not produce reliable temperature data. It required a more advanced interpretation of the reflected optical signal. Other applications of this measurement technique abroad (Vogt et al. 2012) indicate that an on-line calibration of the fibre (i.e. a section of the fibre is kept in two temperature-controlled baths) is needed to obtain meaningful and reliable temperature data. This complex set-up was not anticipated, nor specified by the instrument supplier, and hence was not further developed in PRACLAY.

The different sensor sets are subjected to the variety of environmental conditions present in PRACLAY. Thermocouples inside the piezometer borehole casings installed from the main access gallery (Connecting Gallery) are subjected to dry conditions, while thermocouples integrated in the lining segments have their signal cable running through the heated and saturated PRACLAY Gallery.

To illustrate the complexity of defining the measurement quality and the exact failure date, the signals of the thermocouples installed in the piezometer casing PG50S (which is the horizontal piezometer installed from the PG in the middle of this gallery) are illustrated in Figure 19. Signals can start drifting from the expected value, noise can appear, or sudden changes to a new (more or less stable) value can be detected. Sensor readings that do not give representative measurements could, however, still be valuable, as they could give an indication of the failure cause, such a failing sensor cable connection. Such a failing connection could result in readings that do not give the temperature at the location of the thermocouple, but at the location of the failing connection. In this case, the thermocouples are installed inside the piezometer casing, but with extension wiring that runs along the inside of the heated part of the PRACLAY Gallery. When this thermocouple does not show values that can be expected inside the clay, but rather resembles the temperature inside the gallery of around 85°C (in Fig. 19, three thermocouples show this behaviour), this can indicate that a connection inside the gallery has failed. Such a failure could have created a shortcut at this connection (e.g. due to water ingress) and a new hot (or measurement) junction replacing the original (deeper) junction. It indicates the usefulness of maintaining the recording of the raw signals, even if they are at first sight not representative anymore.

Within PRACLAY, different resolutions were also obtained for the thermocouple measurements, depending on the data acquisition system. For the thermocouples in PRACLAY, type T was selected. This type has a rather limited temperature range (safe upper limit of 300°C, but usually limited to lower temperatures depending on the cable type) compared to other types, but which is still largely sufficient for the experimental conditions of the PRACLAY setup. On the other hand, its accuracy of 0.5°C is the best of all thermocouple types. With this accuracy, it was considered sufficient to have temperature values read with a resolution of 0.1°C (according to the specifications of the selected data-acquisition system, the MX100 from Yokogawa). However, it has been observed that for slowly evolving temperatures, this can result in a stepped graphical representation. When data acquisition equipment with a better resolution is used (in this case, a Campbell Scientific CR1000X with a 24-bit analogue-to-digital conversion), smoother graphs are obtained. It indicates that the notion of ‘accuracy’ encompasses several factors, such as a systematic component (which typically remains constant over a limited range) and repeatability, which is much smaller and is not or hardly noticeable in the measurements. For type T, the repeatability has been shown to be less than 0.1°C (Wang 1989), and a resolution better than 0.1°C is therefore recommended.

Point measurements based on sensors such as thermocouples, platinum resistance sensors or thermistors offer reliable measurements if their installation (including extension cabling and cable connections) is compatible with the environment. Novel techniques based on fibre optic technology (in particular distributed sensing) offer promising perspectives as they offer a large number of measurement points with a minimal invasive effect. Their measurement performance needs, however, to be improved to allow hassle-free measurements over a long term.

Based on experience from the CLIPEX project, the inclinometer was installed using an additional casing to avoid the inclinometer casing from collapsing. This helped to install the inclinometer without any problems. For the SOFO gauges, an installation protocol was developed because no such protocol for installing these gauges in boreholes was defined by the manufacturer. This protocol consisted in embedding the gauges inside a flexible hose filled with a modified silicone polymer kit. This not only provided sufficient rigidity to install the setup, but also additional protection to the gauges. Nevertheless, one gauge failed during the installation.

The SOFO sensors suffered from a fragile construction. The connectors and the read-out devices that were located near construction activity, such as seal installation or gallery backfilling, had to be repaired or serviced frequently. Repairs consisted of fibre optic splicing, e.g. due to damaged connectors, while the read-out device (SOFO V from SmarTec) was sent back several times for repair. This resulted in an incomplete data series, which is more difficult to interpret. Some periods with more frequent readings are available (up to a few readings per day, when the read-out device functioned autonomously), while manual readings were taken at more irregular times, depending on the availability of the instrument or operators. Finally, these sensors failed when the water pressures exceeded 1.5 MPa.

Both the inclinometer and the SOFO sensors were installed inside the saturated clay. The inclinometer casing was, however, protected additionally by a steel casing for both mechanical and water-proofing reasons.

The SOFO sensors provided measurements with a high resolution (1 µm over the complete gauge length of 5–10 m) and a high repeatability (differences between repeated measurements were limited to 2–3 µm).

The SOFO sensors provided measurements with a high resolution (1 µm over the complete gauge length of 5–10 m) and a high repeatability (differences between repeated measurements were limited to 2–3 µm).

When interpreting the inclinometer sensor readings, it was noticed that the calibration of some sensors showed a low accuracy. As the inclinometer interpretation is based on an integration of individual level sensors, errors in previous sensors will accumulate when determining the vertical position or displacement at a certain location. This makes it difficult to obtain an accurate picture of the movement of the clay due to the excavation and heating.

Displacement measurements in a poorly indurated clay such as the Boom Clay require specific attention, in particular when moving parts are involved. The fibre optic interferometric measurement principle proved to show readings with a high resolution and precision, showing its promising potential; its implementation, however, needs further improvement to make it a reliable long-term measurement technique for DGR conditions. Current inclinometer versions are now based on micro-electro-mechanical systems sensors, which have better measurement characteristics than the now outdated electro-levels. Traditional inclinometer designs require, however, invasive boreholes, limiting their use in a DGR rather to the edges of the monitored zones.