Haptics : Research - CPR Mannequin browse

CPR Mannequin

Senior Design Team: Mike Boyle, Nihar Dharamsey, Simon Healey, Nihar Naik, Andrew Stanley
Faculty Advisor: Dr. Katherine J. Kuchenbecker
Sponsor: Matthew R. Maltese, Director of Biomechanics Research, Children's Hospital of Philadelphia


The purpose of this project was to design and build a high fidelity mannequin chest for cardiopulmonary resuscitation (CPR) training. A variety of mannequins already exist to simulate a multitude of human-like characteristics, including breathing, vomiting, sweating, pupil dilation, and proper response to the administration of medicine. Of these mannequins, however, none have chest characteristics that accurately recreate the feel of the human chest. A human chest exhibits non-linear forces and significant hysteresis during CPR, and traditional spring-based mannequins do not exhibit these features.

The projectís primary goal was to recreate this hysteresis in a CPR mannequin. By incorporating data, provided by the Children's Hospital of Philadelphia, from real CPR administrations, the team developed a model that it then used to provide the appropriate amount of damping force so that any CPR per-formed on the high fidelity mannequin would feel more realistic than if it were performed on a traditional CPR mannequin. When combined with a feedback system and graphical user interface that informs users whether or not CPR is being performed correctly, the high fidelity mannequin can serve as a more effective training tool for those learning or improving their CPR.

Modeling the Human Chest

As shown in the data below, the human chest deflects fairly linearly with increasing force during the compression. The force-deflection characteristics of a simple spring-mass system accurately mimics this behavior. During the rebound, however, the human chest rebounds with less force than was required to com-press it, modeled more accurately by adding a damper into the system, with the coefficients calculated from a least-squares fit to all of the cycles in the real data.

Force-Deflection Curve of a single compression of a male 21-year old CPR patient's chest

The plot below compares the real data to both the spring force and the force from our one-way damping model. Force residuals are shown underneath, highlighting the inadequacy of modeling the rebound of the chest with only a spring.

Force versus Time data comparing a real human chest (blue), a spring model fit to the data (red), and a spring with one-way damper model fit to the data (green)

Prototype Components

Spring: provides a spring force throughout the compression cycle. Springs of different stiffness can easily be installed to simulate chests of various stiffness.
Airpots: provide the damping force required in the mannequin by adjusting a valve at the orifice while the pistons move with the compression motion.
Valve: opens and closes based on the voltage supplied to it by the microcontroller. The supplied voltage varies the cross-sectional area of the valve linearly with the voltage applied providing the necessary damping needed at a point in the compression cycle.
Push-Pull Cable: converts the vertical motion of the chest into horizontal motion.
Microcontroller: performs most of the computations required by our CPR mannequin. Using the position of the slide potentiometer and the pressure inside the pneumatic system, the microcontroller will output the necessary voltage to open the valve the correct amount.
Pressure Transducer: determines the damping force exerted by the Airpots by measuring the pressure difference between the Airpots and the atmosphere.
Slide Potentiometer: provides information to the microcontroller about where in the compression cycle the mannequin is so that the proper damping force can be applied at the appropriate point in the CPR cycle.
Force Sensing Resistor (FSR): measures the force applied and sends that measurement to the computer, allowing the computer to provide feedback to the user regarding the quality of CPR being performed. The FSR detects if the CPR performer removes all weight from the chest after each compression cycle to allow for the chest to expand completely as recommended by current CPR guidelines.
Manifold: connects the different aspects of the pneumatic system.

An image of the prototype with components labeled

Control System for Programmable Damping

To achieve the one-way damping characteristics of the human chest, a microcontroller sends varying voltages to adjust the orifice of a programmable valve while the pistons of the dashpots move up and down with the compressions. As the piston moves down, the air compresses, resulting in an upward damping force), so to avoid this damping force the valve is opened during the compression portion of each cycle. During the rebound phase, the valve is partially closed to limit the flow of air into the dashpots as the air inside the cylinders expands, resulting in a downward damping force. The desired damping force at each point during the rebound phase of each cycle is calculated from the model of the human chest described above. The equations in the figure below provide the feedforward term in the control scheme; essentially this is the system's best guess for how far to open the valve at any point in time based on the current chest velocity and desired damping force from the model.

Differentiating the ideal gas law and combining with other equations provides the dynamic model of variable air damping used to get the feedforward term in the control scheme.

A Simulink block diagram, shown below, allows simulation of the entire control scheme implemented on the microcontroller. The model of the human chest and feedforward blocks are described above. The actuator limitations block simulates the fact that, even when the valve is completely open, there will always be some flow constriction and that, even when the valve is completely closed, there will always be some leakage. The plant dynamics are the inverse of the feedforward blocks, rearranged to simulate how the physical system would respond to various control inputs. The feedback compensator closes the loop for force control, measuring the difference between the desired damping force from the model of the human chest and the actual damping force (measured by multiplying the pressure in the dashpots by their combined cross-sectional areas). This measured force error is multiplied by proportional and derivative (PD) constants to improve the tracking of the desired damping force.

Simulink Diagram used for system simulation.

Graphical User Interface

The Graphical User Interface (GUI) provides real time feedback to users about the quality of CPR being performed. The GUI plots pressure, time, compression depth, system force, and valve voltage.

Graphical User Interface shown to user real-time during CPR training session After a training session is completed, the GUI analysis displays information to the user whether the compression depth, compression frequency, and re-lease force have been achieved.

Graphical User Interface shown to user for post-training analysis


The plot on the left shows the actual force-displacement curve from real administration of CPR on a human. The plot on the right shows the force-displacement curve for a traditional CPR mannequin (spring only, in red) and the curve for the teamís high fidelity mannequin (spring and damping, in green). It is evident that the high fidelity mannequin more accurately recreates the feel of the human chest (to the CPR administrator). This can be seen by the fact that the green plot (the high fidelity mannequin) more closely resembles the blue plot (actual data) than the red plot (standard spring mannequin). Though the magnitude of the force varies slightly (as can be seen by the different Force axis scales), this can be accounted for by simply installing a stiffer spring in the high fidelity mannequin.

Quantitative Results for CPR Mannequin