1 Introduction

Concrete placement and vibration are important in achieving desired construction quality measures, because inappropriate placement and vibration of concrete can significantly decrease concrete quality by causing concrete defects, such as honeycombs and air voids. These defects can result in delays, cost overruns, and disputes among project stakeholders, as well as frequent maintenance, elevated safety concerns and shortened service life of facilities (Aljassmi & Han, 2012; Barber et al., 2000; Mills et al., 2009). For these reasons, project managers have made considerable efforts to reduce the defects by ensuring proper performance of concrete placement and vibration.

However, many projects still face problems with concrete quality, because faulty workmanship cannot be easily detected on the spot and immediate corrective actions are rarely made during the concrete construction operations (Lee & Skibniewski, 2019). The current method by human supervision and inspection is not a practical solution for continuous monitoring of compliance, because (1) managers cannot perform all-time monitoring by always hovering around the concrete workers; (2) their sights of view on construction workers and equipment are often disturbed and occluded by objects; and (3) they have different levels of knowledge and field experiences, possibly leading to different interpretations of observed workmanship. Due to these limitations of the visual monitoring method, concrete defects are typically inspected and repaired after the concrete is hardened. Although many of such defects can be repaired, this reactive method is inefficient because inspection and repair would incur additional time and cost (e.g., related costs of material and equipment use, inspectors’ time for examining, evaluating, and reporting the defects). Moreover, slipshod inspection and repair may lead to failures in detecting defects or unsatisfactory quality of the repaired concrete. These reasons explain why proactive and preventive quality control measures should be achieved.

This study introduces a monitoring and warning solution of concrete workmanship inspection to help managers ensure the quality of concrete placement and vibration operations, thus to prevent the concrete defects in a proactive manner. Since various studies have explored and developed many tracking technologies for concrete vibrators, this study will focus on achieving monitoring and management support by utilizing the data collected from the tracking technologies.

2 Literature review

2.1 Current monitoring and supervision of workmanship

The quality of concrete is susceptible to the operators’ workmanship as well as the degree of skills in which the construction activities are performed. Even concrete made of good materials in a properly proportioned and batched manner can be defective if performed with poor workmanship (ACI Committee 311, 2007). When freshly mixed concrete is deposited, it will normally not be able to completely fill the spaces in the formworks and around embedded items (e.g., steel reinforcements) with its own weight. Rather, it may become “honeycombed”; that is, it may contain entrapped air at a level of approaching 20% of its volume. Improper placement and vibration may account for non-uniform density and excessive air voids. Such defects would weaken the structural integrity and mechanical properties of concrete members, such as compressive strength and bond stress with reinforcement (ACI Committee 309, 2017; Cement Concrete & Aggregates Australia, 2006). Additionally, such defects may cause long-term durability problems by allowing water seepage into the concrete members and exposing the reinforcements to corrosion (Roy et al., 1999; Siegfried, 1992). By monitoring and supervising how concrete is placed and vibrated, project managers can assure the quality and minimize defects. Generally, they walk around the site on a regular basis to visually inspect and supervise the concrete operations. They assess the observed behaviors of the workers based on their knowledge, experience, and relevant rules and guidelines. If managers detect improper workmanship, they will typically ask for corrections from the foremen and/or workers.

However, the vision-based monitoring of concrete work entails several problems that affect its reliability. First, the workmanship is often monitored temporarily and intermittently. In fact, visual monitoring is too time-consuming and labor-intensive to be performed continuously, because it requires the project managers’ entire focus on the concrete work. The managers need to stay at the placement sites all the time, while keeping their eyes on the workers and equipment. In fact, managers are also responsible for taking care of cost, schedules, and safety, in addition to the quality of concurrent construction activities performed by various crew and equipment. For this reason, nonconformity of rules and guides may not be detected when managers are tied up with other tasks.

Moreover, visual monitoring and inspection is error-prone and subjective (Kim et al., 2015; Lundkvist et al., 2014; Park et al., 2013). Human visual feedback can be erroneous when their views are unclear or occluded by obstacles. Cast-in concrete is often poured into a formwork that is tightly sealed to prevent leakage, and it is hard to recognize objects inside the formwork. Moreover, concrete is placed and vibrated by a crew consisting of workers who play different roles (e.g., adjusting the direction of the pump hose, handling concrete finishing tools, or operating concrete vibrators). In these dynamic circumstances, managers’ view is frequently blocked, and it is challenging to track and monitor all workers simultaneously. Furthermore, since the information acquired by the visual monitoring methods may be inaccurate, supervisors often need to rely on their own subjective interpretation and evaluation, which may vary depending on the levels of their knowledge and field experiences (Gong et al., 2015; Tattersall, 1991).

Last, the process of learning and improving workmanship is an inefficient and long-time one. Inexperienced workers without the requisite expertise and know-how often have to go through several trials and errors by performing the work and evaluating the resulting quality. The next visit to a certain site by a concrete worker can be days and weeks later at which they may not remember the behaviors during the placement. This time gap reduces the learning rate of workers, meaning that more practice is needed to improve their workmanship. The problem of low-learning rates also applies to project managers: the knowledge of concrete quality and workmanship only remains in their memories and is often forgotten and distorted. Thus, not having tangible workmanship data becomes an obstacle to extracting and sharing knowledge with other project managers and workers for effective improvement.

Due to the problems of monitoring workmanship as described above, faulty site workmanship is common, and concrete defects in construction are frequently found commonly in buildings and infrastructures (Ahzahar et al., 2011; Lin & Fan, 2018; Love et al., 2017; Neeley, 1993). The resulting quality problems are managed by a diagnostic and corrective approach, which is to inspect and repair the defects (Love, 2002; Lundkvist et al., 2014). Although inspection and repair are important for meeting the desired quality level of the final products, they are inherently reactive, as they merely correct non-conformance in an "after-the-fact" manner. Repairing defects can be technically difficult and costly, requiring more efforts than taking preventative measures, because the concrete has already hardened and gained strength (ACI Committee 309, 2005; Kennedy, 2005; Rwelamila & Wiseman, 1995). Performing inspection and repair cannot guarantee good quality if the defects are not detected during data entry or when newly applied material is not bonded with the existing concrete (ACI Committee 309, 2005; Kim et al., 2015; Lundkvist et al., 2014; Park et al., 2013). For these reasons, concrete defects should be minimized and prevented by following proper placing and consolidating procedures (Neeley, 1993).

2.2 Previous studies on monitoring on-site concrete workmanship

In this section, the existing research on advanced monitoring of concrete work is reviewed (Table 1). Studies that did not particularly focus on concrete work (e.g., excavators and general worker location tracking) are not included. Lee and Skibniewski (2019) introduced a conceptual framework of a concrete placement and vibration work monitoring system. They proposed to utilize ultrasound (US) positioning and computer vision technologies, to track locations of objects and perform monitoring based on the key operational factors that can represent the critical workmanship related to concrete quality. Gong et al. (2015) attached a ultra-wideband positioning sensor to a vibrator head, in order to acquire its positioning data. They tested the positioning accuracy from simulated work and visualized the locations of the vibrator head. However, since the purpose of their research was to apply and test the tracking technology, there was no consideration on how to support managerial decisions with the collected data. Tian and Bian (2014) tracked vibrators’ locations by using the Global Positioning System (GPS) and the Global Navigation Satellite System (GLONASS). In the study, they fixed a carbon fiber bar that has a mobile antenna with the vibrator and estimated the vibrator head’s location from the antenna’s location and the length of the fiber bar. They also displayed the vibrator status (insufficient, normal, and excessive vibration) in the experiment. Wang et al. (2021) developed a computer-vision-based vibration quality monitoring method. The system, which used an infrared sensor to measure the distance between the sensor and the concrete surface and a camera to capture concrete surface images. The system was applied to a heavy vibrator machine which is equipped with a head of multiple vibrating bars. Although the system could monitor undervibration and overvibration from the concrete surface, the vibrator’s location was not recorded. Tian et al. (2019) developed a real-time visual monitoring system for vibration effects by using the Global Navigation Satellite Systems (GNSS), including GPS, Beidou, and GLONASS. This monitoring system could track the vibrator tip’s location based on the location of the mobile antenna on the worker’s wrists. They assumed that the locations of the tip were alone on the linear line between the two hands. The system had a warning function to instruct the construction workers to act against observed undervibration. Liu et al. (2015) developed a quality monitoring system for storehouse surfaces of roller compacted concrete (RCC) dams. The system utilized GLONASS and GPS to acquire the positioning of roller compactors. It monitored rolling speed and the number of passes of compaction and sent warning messages to supervision and management personnel.

Table 1 Studies about monitoring on-site concrete workmanship
Full size table

Various technologies have been used to track vibration effort, and many of them adopt GNSS which requires a clear view of the sky. The major focus of these studies was on the technological development and visualization by testing and improving the applicability of the technologies to concrete work. Some studies modified the design of internal vibrators to improving the tracking performance, while others used different vibrating machines to install and implement the tracking sensors. Regarding the decision support features, Lee and Skibniewski (2019) identified the important factors of concrete placement and vibration work, but failed to develop the proposed concept of the monitoring system. Many of the existing studies either focused only on undervibration of concrete or did not have a warning system that supports real-time decision making.

Moreover, they monitored the output of the product-focused concrete work processes (e.g., concrete surface after vibration), rather than workers’ behaviors during concrete placement. For example, the existing monitoring systems can display undervibrated areas of concrete and send an alert about these areas, but they did not monitor the movement of the workers (e.g., short insertion duration, or insufficient insertion depth) which triggered these warnings. Monitoring how the work is performed would foster workers’ awareness of their relative compliance to a set of guidelines and clarify certain behaviors that cause concrete quality issues and expedite the learning process by changing their workmanship immediately.

Lastly, the studies of monitoring concrete placement are scarce. The existing studies were generally focused on tracking concrete vibrators but lacked efforts in tracking concrete-pouring activities. Proper placement of concrete is vital, as excessive free fall of concrete and pouring concrete directly on vertical forms or reinforcing bars may cause segregation. Furthermore, the vibration status of concrete can be understood only when it is tracked with the poured concrete, because without knowing the locations where the concrete is poured, it is impossible to determine whether the concrete is vibrated or not.

The research question addressed in this study focuses on (1) achieving real-time quality assurance for concrete placement and vibration operations, and (2) overcoming the limitations of the existing monitoring methods, such as the lack of monitoring and warning features for workers’ behaviors and the issues with the placement workmanship quality.

3 Methodology

To help managers assure the quality of concrete workmanship in real time, a monitoring and warning solution is designed and developed in this study. Specifically, the solution will perform data collection on concrete placement and vibration with ultrasound (US) and electromagnetic (EM) sensors attached to concrete equipment (e.g., vibrators or pump hoses), continuously compare the observed work-related parameters with the threshold values, and alerts managers when the workmanship does not comply with the recommended practices. The US positioning technology was selected based on a comparison by Lee and Skibniewski (2019), and the EM positioning technology was chosen because its accuracy is not affected by occluded line-of-sight.

Fig. 1 shows the procedures for developing the concrete work monitoring solution. First, both appropriate and inappropriate practices of concrete construction are determined. This study focuses on the common techniques of concrete placement and vibration, which use concrete pump hoses and internal vibrators (also as known as spud or poker vibrators). Manuals, reports and guidebooks about concrete work and inspection are reviewed to determine the KOFs. Then, these KOFs are used to figure out their parameters for measuring and determining the compliance or violation of the KOFs. For example, the parameter of the KOF “Compaction must be done while concrete is still plastic” can be the time gap between placing and vibrating concrete. By measuring this time gap, the violation of the KOF can be monitored and determined.

Fig. 1
figure 1

Procedure for developing the concrete work monitoring solution

Full size image

In the development step, the major tasks of the solution were identified and classified into four groups, and the modules dedicated to performing each task group were developed. These modules were used to extract design information, collect and merge sensor data, update status of placement and vibration, and perform real-time warning. The performance of the solution was tested and evaluated at a construction site. The reliability of the collected data was assessed by comparing them with a video recorded during the work. If data were not collected for a moment or the accuracy was too low, the possible causes and remedies were discussed. In addition, the sensor data and the video record were used to analyze the warning instances and their impacts on the concrete quality. The resulting concrete quality was visually inspected to decide whether there was any quality problem not prevented by the warning module.

3.1 Determining key operational factors (KOFs)

Fifteen KOFs are summarized in this study, and they are what managers shall focus on to make timely feedback to the workers. The first six factors are about concrete placement, while the remaining nine are about vibration. Table 2 shows the KOFs and the consequences of violation. After determining the KOFs, their parameters and thresholds for measuring and determining the compliance or violation were investigated (Table 3).

Table 2 Key operational factors and the consequences of violation
Full size table
Table 3 Parameters and recommended values of the key operational factors for general concrete work
Full size table

3.2 Development of the solution and its modules

3.2.1 Module 1: Extract the design information

Module 1 extracts the geometric information of concrete members and forms. From the corner points of a concrete member that a user has entered, the module creates the concrete points and placement grids used for tracking the status of concrete placing and vibrating. Extracting design data is initiated by calculating a coordinate conversion matrix, which will convert the coordinates of concrete members and forms in CAD drawings into the US sensor coordinates. A coordinate conversion matrix (4 × 3) is calculated by using the following Eq. (1) with a set of CAD points and their corresponding US points. By multiplying with the conversion matrix, the CAD points are converted and expressed in the US coordinate system. There should be at least three corresponding non-collinear points in order to calculate the conversion matrix.

$$\left[ {\overbrace {{\begin{array}{*{20}c} {a_{1x} } & {a_{1y} } & {a_{1z} } \\ {a_{2x} } & {a_{2y} } & {a_{2z} } \\ {a_{3x} } & {a_{3y} } & {a_{3z} } \\ \vdots & \vdots & \vdots \\ {a_{nx} } & {a_{ny} } & {a_{nz} } \\ \end{array} }}^{{{\varvec{P}}_{{{\text{cad}}}} }}\left| {\begin{array}{*{20}c} 1 \\ 1 \\ 1 \\ \vdots \\ 1 \\ \end{array} } \right.} \right] \times {\varvec{C}} = \left[ {\overbrace {{\begin{array}{*{20}c} {b_{1x} } & {b_{1y} } & {b_{1z} } \\ {b_{2x} } & {b_{2y} } & {b_{2z} } \\ {b_{3x} } & {b_{3y} } & {b_{3z} } \\ \vdots & \vdots & \vdots \\ {b_{nx} } & {b_{ny} } & {b_{nz} } \\ \end{array} }}^{{{\varvec{P}}_{{{\text{us}}}} }}} \right],$$
(1)

where Pcad indicates a matrix of CAD points (n × 3), C indicates a coordinate conversion matrix (4 × 3), Pus indicates a matrix of US points (n × 3), and n indicates the number of points used for the conversion.

For example, if the four points of a CAD drawing are (2.4257, − 3.8481, 2.0000), (− 0.5743, − 3.8481, 2.0000), (− 0.5713, 1.2446, 2.0000) and (2.4257, 1.2446, 2.0000), and their corresponding points are (3.0000, 0.0000, 0.0000), (0.0000, 0.0000, 0.0000), (0.0030, 5.0920, 0.0000) and (3.0000, 5.0930, 0.0000) in the US coordinate system, then its conversion matrix is calculated in Eq. (2).

$$\left[ {\overbrace {{\begin{array}{*{20}c} {2.4257} & { - 3.8481} & {2.0000} \\ { - 0.5743} & { - 3.8481} & {2.0000} \\ { - 0.5713} & {1.2446} & {2.0000} \\ {2.4257} & {1.2446} & {2.0000} \\ \end{array} }}^{{{\varvec{P}}_{{{\text{cad}}}} }}\left| {\begin{array}{*{20}c} 1 \\ 1 \\ 1 \\ 1 \\ \end{array} } \right.} \right] \times {\varvec{C}} = \left[ {\overbrace {{\begin{array}{*{20}c} {3.0000} & {0.0000} & {0.0000} \\ {0.0000} & {0.0000} & {0.0000} \\ {0.0030} & {5.0920} & {0.0000} \\ {3.0000} & {5.0930} & {0.0000} \\ \end{array} }}^{{{\varvec{P}}_{{{\text{us}}}} }}} \right],$$
(2)
$${\varvec{C}}= \left[ \begin{array}{ccc}1.0000& 0.0000& 0.0000\\ 0.0000& 0.9999& 0.0000\\ 0.0000& 0.0000& 1.0000\\ 0.5743& 3.8476& -2.0000\end{array}\right].$$
(3)

After converting the coordinate systems of the points, the forms’ plane parameters are calculated, and the placement grids and internal points of concrete members are created. The form’s plane parameters represent the shapes of the forms in three dimensions, and they are used to compute the distance between a point and the form as well as an intersection between a line and the form. The placement grids are created by dividing an area of a concrete member along with x and y axes, and similarly, the internal points are created by distributing points along with x, y, and z axes. The placement grids enable tracking of the height of fresh concrete deposited in the forms, while the internal points are used to monitor the vibration status at all critical points of the concrete members. Figure 2 shows an example of a concrete member in the CAD coordinates and the US coordinates.

Fig. 2
figure 2

A concrete member in (a) the CAD coordinates and (b) the US coordinates

Full size image

3.2.2 Module 2: Collect and merge sensor data

The solution collects positioning data from two US sensors for tracking a pump hose, two US sensors for tracking a vibrator operator and three EM sensors for tracking a vibrator shaft. Figure 3 shows the procedures for processing and merging collected data from the sensors. There are four groups of processes: (1) data collection and retrieval; (2) US data processing for a pump hose; (3) US and EM data processing for a vibrator operator; and (4) integration of US and EM coordinates. The solution first checks whether the data are available before initiating data processing. If it fails to retrieve any of the data required, it will move on to the next loop, trying to retrieve the data again.

Fig. 3
figure 3

Procedure for processing and merging sensor data

Full size image

After that, the data for the pump hose and the vibrator are merged, respectively. When the data are merged, a linear interpolation method is applied to match their timestamps. The US and EM sensors have different coordinate systems. Therefore, the coordinates are aligned before estimating the location of the vibrator head. Whereas the reference point of the US sensor coordinate system is fixed during data collection, the EM sensors’ reference point is based on the EM source. Thus, the points in the EM sensor system will change depending on the location and the orientation of the source. For this reason, the degrees of tilt and orientation of the source are measured by attaching the US sensors with inertial measurement units to the EM source, and the EM points are converted to the US coordinate system.

3.2.3 Module 3: Update the status of placement and vibration

The height of concrete in each grid is estimated based on the pumping rate and the duration that the pump hose stays toward the grid surface. This simplified approach is significantly faster and less computationally expensive than scanning the actual shapes of deposited concrete. From the location information of the pump hose, the change of concrete heights over time can be tracked. The procedures for updating the heights of concrete based on the location of the US sensors mounted on the pump hose are presented in Fig. 4.

Fig. 4
figure 4

Procedure for updating the concrete grid’s height with the US data of a pump hose

Full size image

First, the positioning data of the two US sensors mounted on the pump hose are retrieved. One sensor is attached to the upper part of a pump hose, so that it can collect the location data reliably after the hose tip is immersed in the forms. Another sensor is attached to the bottom of the pump hose. If the pump hose is higher than a certain point at which both sensors secure the line-of-sight with the stationary sensors, the location data of both sensors are used to estimate the deposited point. When the hose is down and may not secure the line of sight of the bottom sensor, it is assumed that the bottom sensor is located at a point vertically down from the top sensor. The horizontal distance between the top and the bottom sensors can be generally ignorable when a pump hose is lowered deeply to pour deep concrete members, such as columns and girders, because reinforcing steel would restrict the pump hose’s horizontal movement. After calculating the deposit point, the height of concrete in each grid is updated based on the pumping rate and the duration that the pump hose stays in the grid. The internal points that are lower than the current concrete height are marked as “placed”, and the time of placement is saved.

Figure 5 shows the procedures for updating the vibration status of the concrete internal points. The vibrator head’s location is first estimated based on the locations of the EM sensors mounted on the vibrator shaft. This solution fits the locations of the three EM sensors with a linear line in a top view and with a quadratic curve in a side view. Then, the actual distances between the vibrator head and the sensors are used to acquire the head’s location. With the location of the vibrator head and the height of concrete, the solution updates the vibration status of the concrete members’ internal points. When the height of the vibrator’s head is lower than the height of concrete, the insertion duration begins to be counted. After a certain period of duration, the nearby points of the vibrator’s head are marked as “vibrated”, and the time of vibration is saved. Then, all the updated data of the placement grid and internal points are passed onto the warning module in order to determine whether there is any violation.

Fig. 5
figure 5

Procedure for updating vibration status of concrete internal points

Full size image

3.2.4 Module 4: Real-time warning

The warning module compares the acquired data with the predetermined threshold values, and it will send warning messages to the managers if a violation is detected. The conditions for violation of KOFs are determined as shown in Table 4. For example, KOF #12, “vibrators should not be allowed to touch the forms”, can be considered violated when the head of the vibrator bar stays within 0.l m of a form face for t consecutive seconds. This factor needs threshold values for its two variables (l and t) to be determined. The threshold values can be referenced and calibrated from Table 3.

Table 4 Conditions for violation of KOFs
Full size table

The KOF #7 through #11 are monitored while making sure that the vibrator head stays near the concrete points for a desired period of time. Then, for example, the solution may inform the managers of undervibrated internal points, which encourages an immediate remedial action such as re-vibrating the undervibrated points. In this way, the monitoring is focused on “products” to help achieve quality assurance for all internal points of concrete. It is also important to monitor and manage the “behavior” of the vibrator operator. The vibrator operator needs to know about the poor workmanship, in order to learn about the faulty behavior and prevent its recurrence. In a word, the warning module constantly checks the distance, duration, and depth of insertions before alerting the users if they are not within the set range. When a violation is detected, its details are delivered to the managers, with the warning information displayed on the screen of a local computer and uploaded to a cloud server.

The data of the malpractice are presented in readable formats, such as tables and figures, facilitating the managers to understand the details at a glance. The warning module compresses the sensor data together with the violation time points and saves them in an Excel datasheet. This sheet allows the managers to revisit what happened around the time of violations. It also creates and stores.jpg images to help managers interpret the coordinates and pinpoint the locations. Figure 6 shows an example of KOFs #1 and #5. The images show the information about the violations, such as their time points and locations. The left figure below highlights the placement grids that exceed the acceptable height difference, while the right figure shows the locations where concrete is placed directly on the vertical forms.

Fig. 6
figure 6

Warning images for KOFs #1 and #5

Full size image

4 Site application

A site application was conducted to confirm whether malpractices were caught as intended and to investigate whether there was any false alarm. The project is located at the University of Maryland, and its scope includes designing and constructing a 4-story building that has labs, conference rooms, and student lounges. The application area was 7.3 m by 3.8 m with a thickness of 0.46 m. The project used a concrete bucket instead of a pump hose for its placement method; and the concrete had a slump of 15.2 cm with the characteristic strength of 24.13 MPa. Figure 7 shows the installed sensors on the concrete bucket and the vibrator shaft. Given that workers would be more careful with their behaviors when they were aware of the experiment, the threshold values were intentionally set tighter when acquiring the warning data.

Fig. 7
figure 7

Installed ultrasound and electromagnetic sensors

Full size image

Prior to the concrete work, the researchers visited the site to attach four stationary US sensors to the columns where the sight is secured and four mobile US sensors to two concrete buckets, while acquiring the CAD drawings of the application area and calculating the conversion matrix. 8 placement grids and 191 internal concrete points were generated. The areas of placement grids ranged from 1.291 to 3.386 m2 after subtracting the areas of columns in the girds. The internal points had equal distances of 0.389 m, 0.381 m, and 0.25 m along with x, y, and z axes, respectively. The placing and vibrating operation was performed for approximately thirty minutes. During the work, one researcher stayed near the vibrator, taking a laptop to receive the sensor data remotely. At the same time, another researcher recorded a video of the work to confirm the quality of the sensor data and the warning instances. The timestamps of opening and closing the bucket gate were recorded to estimate the concrete pouring speed and location.

Figure 8 shows the compared locations of insertions of the collected sensor data and video records. About 95% of the insertions (72 out of 76) were accurately located in the construction environment, with 4 out of 76 showing a reduced accuracy. The 4 red points are supposed to be observed at the blue points close to the tower crane. A comparison with the video records revealed that the wood panels enclosing the tower crane have blocked the line of sight between the stationary sensor and the vibrator operator, thus reducing the accuracy of the US sensors. The error of the red points ranged from 0.85 m to 1.72 m. These results demonstrate that securing the line of sight of the US sensors is critical to collect reliable insertion data, and this can be achieved by separating the path of the vibrator operator from that of bulky equipment and materials, installing stationary sensors at a high position to minimize occlusions, and zoning the tracking areas to reduce blind spots.

Fig. 8
figure 8

(a) Observed and actual vibrator insertion locations and (b) occlusions at the site

Full size image

Table 5 summarizes the descriptions of the warning data generated during the site application. The warning module detected the following: a large concrete pour causing a height difference between nearby placement grids; two undervibrated concrete internal points; one withdrawal action of the vibrator that was too quick; and two occurrences of large tilt angles of the vibrator head. In addition, the vibrator head moved horizontally while being inserted into concrete. The warnings were investigated and confirmed with the recorded video. However, one warning of KOF #15, which is moving horizontally while being submerged into concrete, turned out to be a false alarm. This case occurred when the operator turned off the power of the vibrator machine and let the head become submerged while waiting for the next pour from the bucket. Due to the long insertion duration, the solution cumulated the small errors of the vibrator’s location, and it created a false warning. The remaining violations were confirmed by the video. The authors have finally confirmed with the manager on site that these violations had negligible impact on the quality, considering the slump value, the amount of reinforcement, and the location, frequency, and severity of the violations.

Table 5 Warning instances during the application, results of data validation, and their impacts on the quality
Full size table

5 Discussion

From the site experiment, it was observed that the locations were generally accurate, except for the case where the line of sight between the US sensors was not secured. The monitoring solution was able to capture the violations of the key operational factors reliably. Six warning instances were generated, and one of them, which detected excessive horizontal movement of the vibrator head occurred due to the abnormal insertion duration. The remaining warning instances were generated properly as the operators’ behavior exceeded the set thresholds of KOFs.

This system promises multiple benefits both for project managers and concrete workers. First, the monitoring and warning solution proposed in this study helps achieve proactive and data-driven management of concrete work and enables real-time sensor-based feedbacks, so as to prevent concrete quality problems caused by poor workmanship. The modules of the solution translate the raw positioning data into the information of workmanship and then deliver warning messages to the managers for prompt managerial decisions. With these warnings, both managers and concrete workers can take immediate actions to prevent defects, rather than waiting until the formworks are stripped for inspection and repair.

In addition, the solution can expedite the learning process of project managers and concrete workers by enabling collection, analyses, and sharing of data. Since the collected data of concrete placement remain in a database, managers can perform in-depth data analysis to understand how a certain malpractice contributes to an occurrence of a defect. Sharing the results of such analyses will reinforce the learning and feedback loop, leading to a solid and proactive defect prevention plan. Based on the workmanship analysis, project managers can give clear and unequivocal instructions on the desired workmanship. For example, a specific range of distances and durations of insertions can be recommended, with particular site conditions taken into account. In addition, the warning data can be used to coordinate training sessions for new workers. Informing new workers of the circumstances and locations where the poor workmanship is more likely to occur will help them avoid the malpractices, without learning through experiencing defects.

To maximize the benefits of the monitoring and warning solution, this study highlights the importance of project managers considering the following points before they adopt this solution. First, the managers should clarify and share the purpose of adopting this solution with the stakeholders. Since this solution can work as a data collection and monitoring tool, its application may prompt conflicts among the stakeholders. If it is used to clarify the responsibilities of the stakeholders concerning certain concrete defects in case of a lawsuit, its deployment may create a distrustful atmosphere within the projects, therefore inhibiting its full adoption by the stakeholders. For these reasons, it is suggested that this solution be used as a tool to improve the internal quality assurance and enhance collaborative monitoring for concrete work, rather than for assigning blames to any of the stakeholders.

Second, it is recommended to prepare a long-term application plan, so as to maximize the benefits from the automated monitoring of concrete work. One of the limitations of the proposed solution in the beginning stage is the lack of knowledge on determining threshold values. Project managers may not know the best performing thresholds and may simply follow the values suggested in Table 3. However, since projects have different site conditions, concrete equipment, and properties of concrete mixes, the same thresholds may or may not be able to indicate certain poor workmanship. A long-term use of this solution could allow project managers to assess whether the set values are appropriate for delivering the desired quality. This experience will ultimately enable the data-driven calibration of the threshold values.

Lastly, creating a communication channel between the solution and concrete workers will help them respond promptly to warning messages. The solution designed in this study can deliver warning messages to project managers and concrete foremen for real-time assessment, so that they can require workers to take corrective actions promptly. Project-level decisions on how to respond to the warning messages are provided. For example, if too much concrete is poured at one location, the managers may instruct the pump hose workers to reduce the pumping rate and ask the vibrator workers to insert the vibrator’s head deep enough, while share the warning information with an inspector for a thorough quality investigation on all the locations where poor workmanship is defected. If adopted by a concrete subcontractor, this solution could be more efficient to send the warning information directly to the operators, without passing through the project managers or concrete foreman.

6 Conclusion

Problems with concrete quality are commonly detected at construction sites, and they can lead construction projects to delays and cost overruns. Thus, project managers have long tried to prevent such quality problems by monitoring workmanship. However, since visual monitoring of concrete work is time-consuming and labor-intensive, managers can only perform the monitoring tasks temporarily and intermittently. Current monitoring methods rely much on the visual feedbacks from managers, a process which is error-prone and subjective. Consequently, poor workmanship may not be detected on the spot, and immediate and corrective actions are often not made during the concrete operations.

The current method, which depends heavily on inspecting and repairing defects, is costly, may fail to observe defects, and may be too late to initiate reparative actions when defects are detected. Moreover, due to the absence of instant feedback, inexperienced concrete workers may go through frequent trials and errors in learning to improve their workmanship. The workers may forget the details of their movements when they are able to see the result of their work, which is after the fresh concrete has hardened and the formwork is removed. Observations and experiences of managers about concrete work are often forgotten and distorted, and therefore, opportunities to share knowledge and analyze workmanship may be lost.

There have been numerous attempts to advance the monitoring of concrete work. Some studies have shown the possibility of collecting and displaying data from sensors in a lab environment. However, they are not suitable for the current operational practices of concrete work and construction environments. In fact, their developed tools and devices were not tested with actual concrete, or considerable changes are required for workers to operate the machines to vibrate concrete. By contrast, this study focuses on developing a solution that can be used at the concrete workplace, while not changing the current operation procedures. As a result, a positioning-sensor-based solution has been developed to help monitor concrete placement and vibration, and the solution has been implemented on a construction site. This study has determined certain key operational factors that represent good/bad workmanship, developed algorithms to translate positioning data into the workmanship records, and realized delivery of real-time warning for quick responses by managers. In sum, this study contributes to broadening the applicability of positioning sensors by demonstrating that the positioning data can be used for real-time management of concrete operation.