The study was conducted on a ‘cold’ (non-operating) engine 1Ch 410.5/13. To obtain a vibration signal on an idle engine, a tip was installed instead of a nozzle, which created alternating vacuum and air compression in the over-piston space. The crankshaft of the engine was fixed in a specific position, and as a result, the piston struck the cylinder liner due to a change in the sign of the normal component of the force. The resulting vibrations were transmitted through the block to the sensor installed on it, and then to the electronic equipment for recording and analyzing vibrations.
The parameters of the air transferred to the cylinder (vacuum and pressure) were determined experimentally and chosen to be minimal but sufficient to create a shock pulse from the recorded piston shift. The fixed angle of rotation of the crankshaft was chosen to ensure stable relocation of the piston and that the piston was at the top dead center.
Sensors were installed on the piston to measure its movement, and sets of bushings and pistons with different levels of wear, representing different working lives, were used. Additionally, plugs of various sizes were installed instead of valves to assess the degree of influence of leaks.
The experimental part of the study was conducted using the method of experimental design, as per the theory of experiment planning.
To construct the matrix of screening experiments (
Table 3), we randomly mixed two half-replicas, dividing the factors into two groups. We then created one half-replica from each group. In this work, two half-replicas of the type were mixed to conduct a screening experiment
. One half-replica was assigned to factors
, the other—to factors
. For three factors
:
and
, a PFE plan was drawn up from 8 experiments; the fourth column was obtained using the generating ratio
. For factors
,
, and
, similarly, a plan for a full factorial experiment of 8 experiments was drawn up; the eighth column was obtained using the generating ratio
.
2.3.1. Experiment 1
The screening experiment’s planning matrix, which shows the experiment plan and the coded value of the optimization criterion, is presented in
Table 3. Rows 9 and 10 were selected randomly from both half-replicas to reduce the number of experiments required for the screening experiment.
After the experiments, the results were recorded in the matrix column, where is the average value of the vibration amplitude, and is the number of experiments. The optimization criterion was used to evaluate the results and select the most significant factors for further research.
The screening experiment design matrix, which includes the experiment plan and the value of the optimization criterion in coded form, is an effective tool for evaluating the most significant factors. Randomly selected rows from both half-replicas help to reduce the number of experiments required. The column of the matrix is used to record the results of the experiments, allowing for a quick evaluation of the effectiveness of the selected factors.
The influence of each factor is estimated by calculating the difference between the average values of their levels. This method is used to determine the most significant factors that have a significant impact on the experiment’s outcome. The greater the difference between the means, the more significant the factor is considered to be.
The median value is used as the central tendency measure, which is the value of the variable attribute that falls in the middle of the ordered variation series. This method is useful in cases where the data are not normally distributed or contains outliers. The median is a reliable measure of central tendency that is not affected by extreme values.
If there is an even number of occurrences in a series, the median is the arithmetic mean of the two middle values. The factors and
are distinguished primarily by the difference in medians
. The scatterplot constructed from the initial experimental data (
Figure 2) visually shows that the effects of the first and fourth factors should be evaluated first, as they have the largest differences in medians.
The effects of the factors
are calculated by the following:
where
– are the average values of the optimization criterion in each cell of the table for the factor level (+);
– are the average values of the optimization criterion in the corresponding cell of the table for the factor level (−);
is the number of average values of the optimization criteria
[
26].
To quantify the effects of the factors
and
, let us construct
Table 4 (table with two inputs). In
Table 4, the estimated factors with levels of variation and the calculated average values of the optimization criterion (vibration amplitude) are recorded in each cell of the table.
Thus, at the intersection of row
and column
, the average value of the optimization criterion
is written, obtained by summing the results taken from
Table 3, where factors
and
were at the upper levels (+), and dividing this sum by the number of such experiments. In
Table 3, where the factors
and
were at the upper levels (+), the experiments numbered 4 and 8 corresponds to this. Thus, for the case when the factors
and
were at the upper levels (+), the average values of the optimization criterion are
At the intersection of row
and column
the average value of the optimization criterion
. is written. For this, we took the results of experiments 2, 3, and 10, and as a result we obtained
At the intersection of row
and column
, the average value of the optimization criterion
is written. For this, we took the results of experiments 6, 7, and 9 (the value of the vibration amplitude
), and we obtained
At the intersection of row
and column
, the average value of the optimization criterion
is written. For this, we took the results of experiments 1 and 5, and received the follows:
Using Formula (3), the effects of the factors were calculated. Note that for for the factor level (+), the average values of the optimization criterion will be and , and for the factor level (−) they will be and .
For the average values of the optimization criterion for the factor level (+) will be and , and for the factor level (−) they will be and .
Thus, the effects of factors and .
The significance of the identified effects was checked according to the t-criterion (Student), the numerical values of the t-criterion were
Tabular value of t-criterion with the number of degrees of freedom for a significance level of 0.05, it is 2.447, and for a significance level of 0.1 it is 1.943. Factor Effect significant with 95% confidence probability, and —with 90% confidence probability.
To remove the selected effects, the experimental results were corrected. The correction consists in adding the effects of the selected factors with the opposite sign to the results of screening experiments [
29,
30]. For this, we add 2.2 to all results at the level
and subtract 2.8 from all results at the level
. The data after adjustment
were entered in
Table 5.
2.3.2. Experiment 2
After adjusting the first and fourth factors, a second experiment was set up, consisting of 10 experiments. The results obtained after adjusting the influence of the first and fourth factors are shown in
Figure 3.
From the scatterplot of the results obtained after the first adjustment (
Figure 2), factors are visually distinguished as
(the difference between the medians is 1.1) and
(the difference between the medians is 2.9). An auxiliary table with two inputs was similarly constructed (
Table 6).
To fill in the cell in
Table 6 at the intersection of row
and column
, it is necessary to take the results of experiments (
) numbered 2, 8, and 9 from
Table 5. As a result, the average values of the optimization criterion will be equal to
To fill in the cell in
Table 6 at the intersection of row
and column
, it is necessary to take the results of experiments (
) numbered 5, 6, and 10 from
Table 5, then
To fill in the cell in
Table 6 at the intersection of row
and column
it is necessary to take the results of experiments (
) under numbers 1 and 4 from
Table 5, and for the cells at the intersection of row
and column
, it is necessary to take the results of experiments (
) under numbers 3 and 7 from
Table 5. As a result, we obtain
The magnitudes of the effects of the corresponding factors were calculated:
Calculated values of t-criteria: ; .
Factor Effects and are significant with a 95% confidence level.
To remove the selected effects, the second correction of the experimental results was made. For this, all level results
were increased by 2.5, and all level results
were reduced by 1.1. The data after adjustment
were entered in
Table 7.
2.3.3. Experiment 3
After adjusting the second and seventh factors, a third experiment consisting of 10 trials was conducted. The results obtained after adjusting the influence of the second and seventh factors are shown in
Figure 4. From the scatterplot of the results obtained after the second adjustment (
Figure 4), factors are visually distinguished as
(the difference between the medians is 1.2),
(the difference between medians is 1.0),
(the difference between the medians is 0.5), and
(the difference between the medians is 1.2). However, in the case of
there has already been an attempt to take into account the influence of the factor, so an auxiliary table with three inputs was built.
A supersaturated plan allows a more accurate assessment of the effects, but often does not make it possible to fill in all the cells of the table [
25,
30]. This situation arose when constructing a table with three inputs for factors
,
, and
(
Table 8). Therefore, we had to limit ourselves to building a table with two inputs (
Table 9).
The experiment, at
and
for the case when
was at the upper level (+), was not planned; therefore, in
Table 8, a dash is put in the first cell.
To fill in the cell in
Table 8 at the intersection of row
and column
, for the case when
was at the lower level (
), we take the result (
) of experiment 5 from
Table 7.
In the cell (
Table 8) at the intersection of row
and column
, for the case when
was at the upper level (+), it is necessary to take the result (
) of experiment 3 from
Table 7.
To fill in the cell at the intersection of row
and column
, for the case when
was at the lower level (
), it is necessary to take the results of experiments (
) numbered 7 and 8 from
Table 7. We obtain that
To calculate the average value of the optimization criterion
(the intersection at row
and a column
, for the case when
was at the top level (
), from
Table 7 we take the results of experiments (
), numbered 1, 6, and 9. We obtain that
To fill in the cell at the intersection of row
and column
, for the case when
was at the lower level (
), it is necessary to take the results of experiments (
) with numbers 2 and 10 from
Table 7. We obtain that
To fill in the cell at the intersection of row
and column
, for the case when
was at the top level (
), it is necessary to take the result of the experiment (
) number 4 from
Table 7.
The experiment when
and
for the case when
was at the low level (
), was not planned, therefore, in
Table 8, a dash is put in the last cell.
To further study the effects of factors, we built a table with two inputs (
Table 9).
In accordance with Formula (3), the magnitudes of the effects of the corresponding factors were calculated: and .
Calculated values of t-criteria: ; .
Factor Effects and are not significant either for the 5% or for the 10% significance level.
During the third adjustment of the results of the experiment, all results at the level
were reduced by 0.5, and all results at the level
were increased by 0.1. The data after adjustment (
) were entered in
Table 10.
2.3.4. Experiment 4
After adjusting the third and fifth factors, the fourth experiment was set up, consisting of 10 experiments. The results obtained after adjusting the influence of the second and seventh factors are shown in
Figure 5.
To calculate the magnitude of the effects of the factors
and
, an auxiliary table with two inputs (
Table 11) was built based on the data in
Table 10.
For example, to calculate
(the cell located at the intersection of row
and column
), it is necessary to take the results of experiments (
) from
Table 10 with numbers 1, 5, 9, and 10. We obtain that
The magnitudes of the effects of the corresponding factors were calculated: ; .
Calculated values of t-criteria: ; .
Factor Effects and are not significant for either the 5% or 10% significance level.