Part 1
7 EXHAUST GAS SENSOR MONITORINGGeneral description
The Lambda control loop consists of a linear oxygen sensor upstream catalyst and one or two binary oxygen sensors downstream front catalyst (and post main catalyst). The secondary control loop (downstream of the catalysts) corrects deviations of the upstream oxygen sensor.
Upstream oxygen sensor (wide band sensor)
The wide band oxygen sensor (other commonly used names are LSU or UEGO) measures the oxygen content and the lambda value of the exhaust gas. Because of its continuous characteristic line the wide band sensor is applicable for the upstream catalyst position. This kind of sensor is able to measure a wide range of lambda (e.g. 0.65...lambda... air).
The wide band oxygen sensor has an integrated sensor heater. The heater performance is controlled by the ECM in order to maintain a certain sensor temperature. The sensor temperature is measured indirectly by the internal resistance.
The relevant output signal of a wide band sensor is the pump current (Ip). It is proportional to the lambda value.
Downstream oxygen sensor (binary sensor)
The binary oxygen sensor (other commonly used names are LSF or HEGO) measures also the oxygen content of the exhaust gas. The output voltage of this sensor is reading a level of > 0.45 V for a rich exhaust gas constitution. In case of a lean exhaust gas constitution, the sensor reads voltages < 0.45 V. For stoichiometric mixture (Lambda = 1) the characteristic line of the binary sensor has a steep gradient.
The binary sensor is located downstream the catalyst and is used for catalyst monitoring, monitoring of the upstream oxygen sensor and secondary fuel trim. There is also a sensor heater integrated to ensure a sufficiently high operating temperature.
All sensors are monitored by several single monitoring procedures as follows.
7.1 FO2 Sensor dynamic
Different failure types of wideband oxygen sensors may affect the sensor characteristics:
- Clogged protection tube:
diffusion of exhaust gas into measurement chamber is being slowed down
-->slower sensor signal transition, being detected by front O2 Sensor dynamic
- Crack in pumping cell:
typically results in rich shift of the signal
-->being detected during fuel cut-off
- Crack in reference cell:
typically results in lean shift of the signal
-->being detected by rear fuel trim monitoring
- Poisoning of sensor ceramic surface:
diffusion of exhaust gas into ceramic element is being slowed down
-->slower sensor signal transition, being detected by front O2 Sensor dynamic check
7.1.1 Oxygen sensor, signal too slow, P0133, P0153, bank 1: P0133, bank 2: P0153;
This function monitors the dynamic behavior (responding properties) of a continuous wideband LSU oxygen sensor, which is located upstream of the catalyst (primary oxygen sensor). Slow dynamic sensor behavior may lead to a rise of exhaust gas emissions, and, on the other hand, the results of other monitoring functions that use the signal of the primary sensor (such as the catalyst monitor) may be falsified, or the corresponding monitoring function disabled.
7.1.1.1 Monitoring Strategy
The monitoring function detects symmetrical and asymmetrical faults in the primary oxygen sensor that may be caused by either slow transition or delayed response. The area ratios for each direction and the ratios of the maximum gradients of the actual signal and the setpoint signal for each direction are used as evaluation criteria.
7.1.1.2 Enable Conditions
- Oxygen sensor has reached operating temperature (typically 720 °C)
- A calibratable waiting period is exceeded after the operational readiness was set
- Relative fuel quantity from transition compensation < threshold value
- Volume flow rate of carbon canister purge < threshold value
- Canister purge diagnostic is inactive
- Tank leakage diagnostic is inactive
- Catalyst temperature > threshold value
- No fuel cut-off to any cylinder, all injection valves are activated
- Engine coolant temperature > threshold value
- Engine speed, load, and exhaust gas delay time constant in defined range
- Evaporative system not active, or time of activation > threshold value
- Currently no gear change in process
- This enable condition is optionally calibratable, i. e. it is only active when it is required.
- Gradient of the engine load ≤ threshold value
- "Open circuit in the pump current circuit" monitoring function is completed
- Second lambda control loop not active
- Forced lambda modulation not active
7.1.1.3 Malfunction Criteria
The monitoring function utilizes surges in the amount of fuel injected, and compares the actual oxygen sensor signal with the expected oxygen sensor signal. For this purpose, two corresponding, comparable areas and maximum gradients are calculated. One of the areas and the respective maximum gradient are determined on the basis of the actual sensor signal, the other area and the respective maximum gradient are determined on the basis of the expected sensor signal.
Starting with the response to a fuel surge, the area between the inverted signal and a horizontal line, which begins at the point of the start value, is calculated during a defined time period. Within the same time frame the maximum gradient of the signal is calculated.
Then a quotient from the actual area and the expected area and a quotient from the maximum gradients are calculated. The slower the oxygen sensor responds, the lower the absolute value of the quotient value calculated from these values and/or the two gradients will be. While the area ratio and the gradient ratio are reduced in the case of a slowly rising signal, only the area ratio is reduced in the case of a delayed signal rise. The gradient ratio remains largely unaffected as long as the response of the actual signal occurs within the measurement window. If no response within the measurement window occurs, as in the case of very large faults, both ratios equal zero.
Typically, the monitoring function utilizes an active modification of the setpoint lambda value.
There are two kinds of surges, R2L surges and L2R surges. Single uni-directional surges with a calibratable waiting period between two surges, are applied.
Several options are generally possible when using the fuel surges as a means of activation:
- The monitoring function executes its own lambda modification. This enables double surges with R2L and L2R modifications or single uni-directional surges with a calibratable waiting period between two surges (this strategy is applied starting with MY 2010).
- Measuring the two flanks during the lambda modification of the active catalyst monitoring function (this strategy is planned for future model years)
- Measuring the R2L flank during overrun fuel cut-off and measuring the L2R flank during catalyst purge (rich operation) after overrun fuel cut-off (currently not taken into account)
The options listed above may be applied individually by coding.
The figure shows a fault case where the length of the integration window of the setpoint signal is calculated according to the formula above. The length of the integration window of the actual signal, however, is defined by the end of the common measurement window for both areas as seen in Figure 35: Principle of the response rate integral calculation(brown line).
The Active Modification of the Setpoint Lambda Value and its Marginal Conditions
When the enable conditions are fulfilled, the monitoring function requests a modification of the setpoint lambda value. First, a negative delta lambda value is requested (enrichment). When the areas and the gradients are calculated, a positive delta lambda value is requested (enleanment). The lambda value is conditioned prior to each surge of the setpoint lambda value. Prior to a L2R surge, a lambda value of approximately 1 is applied. Prior to a R2L surge, a rich pre-conditioning is carried out, i. e. the mixture is set to slightly rich (lambda = 0.98 to 0.99). This reduces the NOx exhaust emissions caused by the R2L modification.
Important Marginal Conditions for a Surge of the Setpoint Value
A slow oxygen sensor may cause the lambda controller to oscillate. When this occurs, the L2R modification is done when the slope moves towards "rich", anyway. When an R2L modification is done, the slope of the control action also has to move in the same direction, i. e. towards "lean" (synchronization).
Explanation
Due to the enleanment caused by the control action, the requested enrichment may not achieve the injection of the desired fuel quantity, when the modification of the setpoint lambda value is not synchronized with the control action. As a result, the necessary modification to measure the dynamic response of the sensor cannot be achieved. In case a slow oxygen sensor causes the lambda controller to oscillate, the falling slope of the control action line shows that the delayed sensor signal is responding to a preceding enrichment, i. e. the signal is moving towards "rich". The sensor signal is moving in the right direction, but not as a result of the previously initiated surge. Therefore a dynamic measurement may lead to a false result (e. g. an unjustified pass result).
Further Marginal Conditions for a Surge of the Setpoint Value
- The forced modulation of the continuous lambda control and the second control loop are disabled.
- The current control difference does not exceed a threshold value.
- The intensity of control action has increased several times in succession
or
The intensity of control action has not exceeded a threshold value for longer than a time threshold.
- The total effect of the setpoint lambda value modification and/plus the operational control action must not exceed the maximum permissible enrichment or enleanment for the corresponding control action
- When the L2R modification starts, the difference between the current lambda value and the minimum permissible lambda value must be 1.5 x Delta Lambda of the setpoint value surge L2R. When the R2L modification starts, the difference between the current lambda value and the maximum permissible lambda value must be 1.5 x Delta Lambda of the setpoint value surge R2L.
Calculating the Setpoint Value and the Actual Value of the Inverted Lambda Signal
The area integrals and the maximum gradients are calculated by inverting the setpoint and the actual values of the sensor signal. A low pass filter, which uses a calibratable time constant, is applied to the inverted measured oxygen sensor signal. The filter protects against high-frequency disturbances, which might distort the area integral triggering and the gradient calculation triggering. The setpoint value signal is initially delayed by an exhaust gas delay time element. Subsequently, a low pass filter, which uses the sensor delay time constant, is applied to the setpoint signal. As with the actual signal, a low pass filter, which uses a filter constant, is applied to the setpoint signal again, in order to guarantee its comparability with the actual signal.
Calculating the Area Integrals and the Area Ratio (Dynamic Raw Values)
This describes the calculation of the R2L surge. The L2R surge is calculated in the same manner. For a single measurement, the extent of the surge has to be large enough to measure and thereby to initiate the integration of the setpoint and the actual area.
Detecting a Fuel Surge
To initiate the calculation of the area integrals and the gradients, a relative fuel surge has to be detected. This fuel surge is detected when the difference between the current value and the previous fuel factor value exceeds a calibratable threshold value. Once a fuel surge is detected, both the setpoint and the actual value must respond within a certain time frame.
The time frame for the area integration is calculated when the fuel surge starts. This time frame is calculated from the exhaust gas delay time (gas travel time to the location of the oxygen sensor). The delay time is multiplied with a factor that depends on the sensor delay time constant, which depends on the operating point (engine speed and load / exhaust gas mass flow). The time constant corresponds to the signal increase of an error-free sensor.
The oxygen sensor signal can be expected to react shortly after the end of the exhaust gas delay time. Integration is subsequently started and it continues for the duration of the time constant. This process occurs in both the setpoint and the actual signal. When the actual signal does not respond at the expected time, the integration cannot be executed throughout the entire time frame of the time constant. The area and consequently the dynamic value, decrease.
Due to tolerances in the exhaust gas delay time parameters of the lambda control, the measurement window does not start after the end of the total delay time, but when 80 percent of the current delay time has passed. Tolerances in the sensor delay time constant have to be compensated for by varying the length of the measurement window, using the multiplication factor.
Area Integration of the Expected Inverted Oxygen Sensor Signal (Setpoint Value)
The area integration of the expected inverted signal is initiated when the first response to a fuel surge is detected. For this purpose, the maximum and the minimum values (depending on the direction of the flank change) of the inverted setpoint lambda signal are continuously calculated. These calculations are made after a fuel surge is detected from the lambda controller's response, as described above. An initial response is detected when the signal has changed from the expected signal excursion by a certain factor (typically 0.05 to 0.2). When this condition is satisfied, the area integration and a time counter, which determines the integration period, start. The integration period is calculated from the product of the current time constant and a constant, calibratable factor. When the time counter value is greater than or equal to this integration time, the area calculation is stopped.
Calculating the Maximum Gradient of the Expected Inverted Oxygen Sensor Signal (setpoint value)
The gradient of the expected inverted signal is calculated only after a response to the fuel surge can be detected. For this purpose, the maximum and the minimum values (depending on the direction of the flank change) of the inverted setpoint lambda signal are continuously calculated. These calculations are made after a fuel surge is detected from the lambda controller's response, as described above. An initial response is detected when the signal has changed from the expected signal excursion by a certain factor (typically 0.05 to 0.2). Once this condition is fulfilled, the calculation of the gradient is started, beginning with continuously calculating the difference between the current value and the previous value in 10 ms calculation steps, and calculating the maximum value from the individual gradients of the calculation steps.
Area Integration of the Measured Inverted Oxygen Sensor Signal (Actual Value)
The inverted actual lambda signal is calculated mostly analogously to the integration of the setpoint area. In this case also, the maximum or the minimum value of the inverted setpoint lambda signal is continuously calculated when a fuel surge is detected. Then, the time frame for the integration and the time counter are started when the signal has changed by a calibratable factor, compared to the expected signal excursion. The time frame for the integration is also stopped when the time counter reaches a value calculated from the product of the current time constant and the same calibratable factor, as it is used for calculating the setpoint area.
If there is an excessive delay in the oxygen sensor's response, no response to the fuel surge will be detected in the actual signal. As a result, no area integration can be done, and the area value for the actual signal is set to zero. Thus, the current area ratio also equals zero.
Calculating the Maximum Gradient of the Measured Inverted Oxygen Sensor Signal (actual value)
The maximum gradient of the measured inverted lambda signal is calculated in the same way as the maximum gradient of the setpoint value.
Generating the Diagnostic Values for each Direction
The diagnostic values are calculated for both the falling and the rising flank, respectively. The calculation incorporates the quotient of the actual value integral area (numerator) and the setpoint value integral area (denominator), as well as the quotient of the maximum gradient of the actual value (numerator) and the maximum gradient of the setpoint value (denominator).
Subsequently, the individual area ratios for each direction can be corrected using a factor from a characteristic curve which is a function of the engine temperature (typical factor: 50 °C = 1.5; 90 °C = 1.0).
Explanation: When the engine has not yet completely warmed up, the applied fuel mixture surge reaches the location of the sensor only partially as a result of the fuel being adsorbed to the intake manifold and/or the combustion chamber (surface film). These effects are not considered when modeling the system to be controlled, which is used for generating the target signal. Hence, they have to be taken into account in the diagnostic procedure in order to avoid diagnostic errors with faultless sensors.
From these values, which are calculated as described above, filtered values are generated. The filtered values are the diagnostic values that are used to make fault decisions.
The areas in the examples are determined by successive single R2L and L2R surges.
Calculating the Filtered Dynamic Values
From the values that are calculated for the falling and the rising flanks filtered values are generated. These filtered values are the diagnostic values which are used to make fault decisions.
The event filter constant for the first measurement is 1, which means that the filter is actually initialized with the first measurement value. With each further single result, this filter value decreases by the reciprocal value of the number of measurements. Thus, in each step an average is generated for the individual results that are present up to this point. The reason for this approach is the expectation that the oxygen sensor dynamic does not change substantially between two individual measurement steps (assuming continuous ageing) and therefore all results should be weighted equally.