Manufacturing COVID-19-related products such as face masks, shields, ventilators, shoe covers, gowns, and so forth, rapidly increased in recent months as the virus pandemic surges across the globe. Governments and Industrialists are keen to formulate quick decisions and are open to decentralize productions within a prescribed time limit using smart techniques. Recommendations on the volume of productions and the producer assignments are remaining a sole concern for policymakers or smart city authorities due to the unforeseen or unpredictable nature of the raging pandemic. This article introduces a RandomForest-assisted Collaborative COVID-19 Product Manufacturing (RFCCPM) framework. It collaboratively decides on producing COVID-19 preventive kits in a cost-eficient manner. The approach was experimented at the IoT cloud research laboratory; it achieved a manufacturing cost eficiency of 66 percent when Threshold Accepting (TA) algorithm was incorporated in the framework.
to the ongoing lockdown and minimal employee situations in production units. HighOne of the great health-related pandemic that quality machines have become non- operahas clouted all growth sectors, including man- tional for months since the eruption of the ufacturing and the world finance sector, and pandemic across the globe. Many manufachas unsolved challenges is the lethality due turing companies have almost closed their opto the COVID-19 public crisis. The virus has erations due to reduced workforce and eruptpredominantly led nations to unwelcoming ing supply chain disruptions. Even the most social distancing practices, inefective com- urgent production of COVID-19-related premunications, sweeping economies, discrimi- ventive products such as masks, face shields, nations in certain locations, distrustful rela- hood caps, shoe covers, and so forth, has wittionships, and so forth, at large. nessed a catastrophe which could adversely
The Manufacturing/Fabrication sector has reiterate until innovative solutions are framed seen a disrupted shift in productions which in the manufacturing/fabrication sectors. A remains as an unpredictable realization by man-single manufacturer of a region could not proufacturers in order to revamp productions due duce/supply all required COVID-19 products with utmost satisfaction in a short period in ISIC 2021: International Semantic Intelligence Conference, the midst of the exponential growth of COVIDFebruary 25–27, 2021, New Delhi, India. 19 cases. " EMAIL: shajulin@iiitkottayam.ac.in (S. Benedict) The manufacturing sector, typically, attracts ~ UORRLC:IhDtt:p0:0//0w0-w0w00.s2b-2e5n4e3d-i2c7tg1l0ob(Sa.l.Bceonmed( Sic. tB) enedict) a major portion of revenue in various counC©re2a0t2i1veCCoopmyrmigohntsfLoricethnissepAatpterirbubytioints4a.0utIhnoterrsn.aUtisoenaple(rCmCitBteYd4u.0n)d.er tries when compared to services or retail secCEUR Workshop Proceedings (CEUR- tors. Accordingly, governments and manuWS.org) facturers are keen to provide solutions and dressed for improving the voluminous proupsurge productions at about 3 to 5 percent- ductions: i) Which company needs to be perage in order to combat the anticipated eco- mitted to develop COVID-19 products? and nomic consequences. However, the most cru- ii) How much quantity of COVID-19 prodcial challenges faced by manufacturers, in- ucts is required in a particular region considcluding COVID-19 product developers, dur- ering the increase of the COVID-19 cases? ing the COVID-19-crisis epoch are listed as This article proposes a Random Forest (RF) follows: algorithmic approach to predict the required number of productions of COVID-19 preven• an abrupt stalemate in material move- tive kits based on i) the demand of a particments, especially while transferring ma- ular location, ii) COVID-19 quarantine cases, terials from international hubs or lock- and iii) the budget availability of a region. It down locations, due to disruption in the introduces a RandomForest-assisted Collabosupply chains; rative COVID-19 Product Manufacturing (RFC• a poor quality in COVID-19 solutions CPM) framework. The proposed RFCCPM apwhile adopting expeditious innovations proach includes a Threshold Acceptance (TA) with limited machinery or experiments scheduling algorithm in the framework to prein a short period from restricted lock- pare a manufacturing schedule that considdown locations; ers the geo-distributed nature of collaborative manufacturing for quick voluminous pro• a difering working environment, es- ductions. The RFCCPM framework has the pecially the concept of “Work from Home“c,apability to initiate production and promotes which urged a limited access to remote economy during COVID-19 or similar healthmanufacturing sites; and so forth. related crisis of the future. It maps production tasks to available or functional manufac
A growing volume of research and prod- turing hubs/units using a Threshold Acceptuct development rapidly emerged globally to ing (TA) algorithm with the objective of imcounterfeit the COVID-19 crisis and associ- proving the cost eficiency of manufacturing ated challenges apart from the core medical units. It can consider the lockdown situation solutions. For instance, solutions relating to of a region while producing products. the digital working environment, adopting a The proposed research was experimented long term planning for cash/resource utiliza- at the IoT Cloud Research laboratory and obtion, delivery management, agile production served the prediction accuracy of around 98.4 of health-related products, and so forth, pro- percent while predicting the COVID-19 quarliferated to improve the near future economic antine cases of Kerala state using the Rancatastrophe. domForest algorithm; and, the manufactur
Health departments and concerned oficials ing cost eficiency of 66 percent while
incorof smart cities are eager to proactively de- porating TA algorithm in the framework for
velop COVID-19 preventive kits and protect preparing the manufacturing task schedule.
the residents as healthcare resources, includ- In short, RFCCPM paves way for a cost-
efiing hospitals and doctors, are scarcely avail- cient production of products considering the
able to nurse the exponentially growing pa- availability of minimal employees during
pantients. However, manufacturing the preven- demic epochs such as COVID-19. The major
tive kits has challenges – i.e., the following contributions of the proposed work, as
disresearch questions need to be clarified / ad- cussed in this paper, include the following:
• RFCCPM, an RF-assisted manufactur- Tens of thousands of innovations and
meching framework, was developed for pro- anisms have been initiated in the recent past
ducing COVID-19 products, including using AI [19], [
The rest of the paper is described as fol- term memory (LSTM) models to analyze the
lows: i) Section 2 explores the state-of-the- risk involved in the spread of the virus;
simiart research in the application of RandomFor- larly, authors of [3] applied LSTM models to
est considering COVID-19 situations; ii) Sec- diferentiate corono-virus from the other
restion 3 discusses the inner details and func- piratory diseases such as pneumonia. In [14],
tionalities of the deep-learning assisted col- authors developed a mobile-enabled contact
laborative manufacturing platform; Section 4 tracing mechanism to avoid COVID-19
coninvestigates into the theoretical aspects of im- tacts. Authors of [1] have proposed a method
proving the cost involved in the production to improve the CNN training model as
sufiof COVID-19 products while incorporating cient COVID-19 images were not available in
RFCCPM framework; Section 5 discloses the the early period of the virus outbreak.
experimental results that were carried out at It is a known fact that the impact of
COVIDthe “work at home“ working environment by 19 highly afected the mobility of humans.
accessing the machines of the IoT cloud re- The speed in spreading the virus and the
seversearch lab; and, Section 6 consolidates the find- ity of occurrence difered from region to
reings and insights of the proposed work along gion. This puzzles almost all solution
archiwith a few future developments. tects [
Many researchers and practitioners agree ufacturing (RFCCPM) approach. In a nutshell,
that online tools, including collaborative tools, the RFCCPM framework allows
manufacturwould become a mandatory point of sale for ers or smart city oficials to quickly produce
overriding the emerging lockdown situations the demanding COVID-19 essentials, for
exin cities [31]. Authors of [
Succinctly, a cloud-based online
productionenabling tool would increase the productiv- 3.1.1. Information Collector
ity of COVID-19 products. A few researchers
[
Besides, Prateek et al. [22] has proposed a computer vision-enabled approach to increase lector entity, a golang based cloud service, provides the information in a csv format to the RandomForest prediction engine after converting the formats of the source repository. the social distancing pattern in the manufacturing location. However, not much research work applies the cloud-based services to develop COVID-19 products – i.e., very few research works have been discussed to improve the manufacturing aspects of COVID-19 preventive products.
Predicting the number of quarantine cases in a particular location is mandatory to decide on manufacturing the number of COVID-19 preventive products. Smart city oficials could utilize the data to fix policies and sketch layouts for a complete/partial lockdown in a location. Manufacturing preventive kits based 3.1
The information is dynamic that the proon the past information of the COVID-19 cases duction depending on the current informamight not satisfy the requirements. Two is- tion would be a futile decision. Instead, RFCsues are possible: CPM utilizes a RandomForest-based predic• A large volume of COVID-19 preven- tion of COVID-19 cases for the future months tive kits may be required in a short pe- so that the framework could proactively recriod – i.e., the productions might get ommend the concerned for deciding on the hampered due to the limited availabil- production of the preventive kits. Besides, ity of employees during the period; or, the RFCCPM enables a collaborated manufacturing platform that attains the manufac• Manufacturing a surplus amount of pre- turing schedule with the limited number of ventive kits leads to under-utilization. employees during the raging pandemic period in a cost-eficient manner. masks using N95 masks which provide higher protection from COVID-19 impacts or using cloths; ii) face shields may be manufactured using low-grade disposable plastics or poly-carbonates; iii) shoe covers could be fabricated using of non-woven materials or polypro pylene fabrics; and so forth. In addition to the COVID-19-related products, manufacturing the packing device that holds the preventive products is also considered as a task. The Task Designer entity of the RFCCPM framework prepares a set of optional combinations for manufacturing diferent COVID-19 products of preventive kits. 2. Packaging Products: Packaging is a task that needs to be accomplished by assembling all manufactured products into the packaging device. 3.1
AI and machine learning [
Tasks such as manufacturing face masks, face shields, gloves, shoe covers, infrared thermometers, sanitizer, kit cover, and packaging them 3.1.4. Scheduler need to be formulated depending on the avail- Scheduler organizes tasks depending on the able manufacturing units within the reach- cost factors of allotting tasks to specific manable vicinity. The Task Designer entity of the ufacturing units – i.e., it identifies the manframework attempts to pool tasks in a cloud ufacturing tasks given the information about database, namely, mongodb database. It con- the available manufacturing units of a region/ siders several combinations of manufactur- location. Although the main objective of the ing options while designing tasks. The tasks Scheduler entity is to reduce the manufacturare described in a machine-readable format ing cost, it has to consider several other pasuch as XML, jade, or JSON. rameters such as the availability of appropri
The tasks included in the RFCCPM frame- ate manufacturing units. For instance, if a
work are classified into two broad categories: face mask needs to be manufactured using
i) Manufacturing tasks and ii) Packaging tasks: N95, an appropriate manufacturing unit that
1. Manufacturing COVID-19 Products: Man- is nearest to the vicinity should be available.
ufacturing COVID-19 preventive kits in- Besides, it has to consider the prediction
reclude manufacturing products such as sults available through the RF prediction
enface masks, face shields, gloves, hood gine. This is crucial as lockdown situations
caps, shoe covers, thermometers, sani- could hamper the transfer of products for the
tizers, and so forth. These products are final preparation of predictive kits.
Threshoften manufactured in diferent cate- old Accepting (TA) [
The following points highlight the crucial processes (see Figure 2) involved in the collaborative manufacturing of COVID-19 preventive kits from geo-distributed manufacturing locations considering the forecasting information of the RF Prediction Engine: enable smart city oficials or enterprises to quickly transport the required amount of preventive kits before the commencement of the “lockdown“. 5. At last, an appropriate schedule of tasks depending on the number of available manufacturing units is prepared based on the TA algorithm. The scheduling aims at reducing the costs involved in the production of the entire COVID-19 preventive kits of a region. 4.1 TA Scheduling Approach old, final threshold, threshold value, number of iterations, and the reduction step size of TA are assigned to the algorithm.
In the meantime, the costs involved ( ) for manufacturing the individual category of products from various manufacturing sites and their corresponding manufacturing time are collected in the cloud-based mongodb database. In addition, the distance between the manufacturing units of various locations is stored in the database.
Next, depending on the task set produced by the TA algorithm, the solutions are obtained for each manufacturing task set – i.e., i) the cost involved in the production of an entire COVID-19 preventive kit of a task set is calculated depending on the available manufacturing costs. Note that the products could be of diferent categories in the COVID-19 preventive kit; ii) similarly, the total manufacturing time of the task set is evaluated; and, the total distance involved for transporting the manufactured products between the manufacturing units in a task set is calculated. Depending on these values, the combined objective of a task set is calculated as shown in the equation 4.1.
where, C represents categories of P products. For instance, a face mask product manufactured using N95 in the manufacturing site 3 is represented as C2{P1}– 3; a face = ∑( , , ) (1) shield manufactured using polycarbonate in 2 is represented as C2{P1}– 2. A com- The of each task set in a population plete list of products and their categories is set is evaluated in an iterative manner. shown in Table 1. In the above task set, eight However, only the best task set is stored in products are included. the database considering the minimal manu
Next, multiple sets of manufacturing tasks facturing costs of the production tasks. Once is formulated in one population of TA – when the entire population set is evalui.e., each population set has collections of task ated, the local best task set is recorded sets . Each manufacturing task set , the along with the obtained manufacturing time preventive kit, consists of one category of the or costs. Next, the populations are perturbed variety of products such as face mask, face with newer combinations and their correspondshield, gloves, head cap, shoe cover, thermome- ing of task sets is recorded. The threshter, sanitizer, and package. Besides, the ini- old value ℎ of TA is reduced while increasing tialization parameters that are required for the populations. Whenever the obtained the scheduling algorithm such as initial thresh- reaches the pre-assigned minimal value or whenever the threshold value ℎ attains the final threshold value, the algorithm stops the further creation of iterations or evaluaamong the local task sets tions. Finally, the global best task set is recorded based on the minimal values
. The pictorial representation of the TA algorithm is shown in Figure 3.
RFCCPM framework. At first, the
experimental setup is explained; next, the accuracy
obtained due to the RF Prediction engine is
revealed; and, at last, the identification of the
manufacturing task schedule considering the
availability of manufacturing units and the
costs involved is disclosed using TA algorithm.
5.1. Experimental Setup
The experiments were carried out at a DELL information of a particular location in the
Kerprecision tower 7810 machine of the IoT cloud
research laboratory. The machine utilizes the
ala state of India. The datasets were collected
from the COVID-19 repository of the Kerala
4.15.0-106-generic kernel Ubuntu version. The government site [
At first, the manifestation of utilizing the RF algorithm while predicting the possible number of quarantine candidates at a location was validated. To do so, fifty percent of the observations were utilized for creating a training model and the other fifty percent of the observations were tested using the RF algorithm.
5.2
30 Date6ID0 90 (a) Quarantine Cases in Kasaragod 120 0
120 30 Date6ID0 90 (b) Quarantine Cases in Wayanad yadnaa W m irvahanpunhuaaT r t 30 Date6ID0 90 (c) Quarantine Cases in Kottayam 120 0 30 Date6ID0 90 (d) Quarantine Cases in Thiruvananthapuram 120 locations, namely, Kasaragod, Wayanad, Kot- 5.2.2. Accuracy Improvements tayam, and Thiruvanthapuram. The y-axis of The prediction accuracy, which is measured the Figures describes the number of quarantine cases for these locations. The x-axis dis- as 2 values, was evaluated to understand if further improvements are possible while tuncusses the increasing date of COVID-19 which ing the algorithm-specific parameters. It was is represented as unique identifiers. observed that there were a few possibilities The training values are represented in points such as including more number of variables Iatncdotuhlde tbeestoebdsevravluedesfraoremsthhoewfignurinesbtlhuaetlitnhees. while dividing at each tree node ( ) to improve the prediction accuracy. Accordingly, training and testing values are almost inclined experiments were held by increasing the numto one another – i.e., the prediction accuracy ber of variables for establishing the training is better for the experiment. models. Table 2 discloses the prediction acHence, it is proven to utilize an RF algocuracy values that were observed while contrqiiunthiermecdafsCoerOspVwrIheDdic-i1hct9irnpeglraettveheentotCivtOheeVkIniDtus-m1in9beaqrucoiatfryarnoer-- ductin=g10e,x perim=e1n5t,sawnidt h =2=02,, f or fou=r5-, teen manufacturing locations of Kerala. a location of diferent geo-spatial monitoring As seen in Table 2, it was inferred that the points. prediction improvement of over 5 percent is noticed in several locations – for instance, note Table 3 the prediction accuracy observed in Thiru- Number of Quarantine Cases in Kerala Districts vananthapuram manufacturing station while improving values from 2 to 20; similar is the case with the other manufacturing locations such as Kollam, Kottayam, and Kasaragod.
SlNo Location 35ℎ Day 1 Kasaragod 433.52 2 Kannur 1196.2 3 Wayanad 394.02 4 Kozhikode 252.35 5 Malappuram 415.76
7 Thissur 291.02 Finalizing the algorithm-specific parameters 8 Kochi 2470.02 to obtain higher accuracy – i.e., =20 and 9 Idukki 612.55 number of trees =100, predictions were 10 Kottayam 1540.22 achieved for the future. The obtained predic- 11 Alappuzha 199.48 tion results were illustrated in Table 3. 12 Pathanamthitta 147.6
From Table 3, smart city oficials or health 13 Kollam 629.7 department or the concerned oficials could 14 Thiruvananthapuram 1084.24 decide to instruct Micro-Small-Medium-Enterprises (MSMEs) for manufacturing COVID-19 preventive kits in a cost-eficient manner. However, the recommendations about which
For example, the smart city oficials of Thiru- manufacturing MSME needs to manufacture vanthapuram shall decide to procure 20 COVID-the products in a cost-eficient manner would 19 preventive kits for the thirty-fifth day based be dependent on the TA algorithm of the frameon the recommendations of the RFCCPM frame-work. Besides, the TA parameters define the work – i.e., a production of 20485 preventive tuning of these recommendations in an elekits has to be manufactured at a lower cost. gant manner. the cost involved in the production of the prod- identified in any one of the task set of a popturer could handle manufacturing all prod- in the consecutive iterations or populations. 5.3. Manufacturing Jobs – Cost
Eficiency
Whether one manufacucts? This subsection identifies the job schedule that considers the cost-eficient manufacturing of products using the TA algorithm.
Threshold = 0.0 Initial_Threshold = 0.099 Final_Threshold = 0 Threshold_Reduction = 0.001 Number of Iterations = 1000 Population Size = 50 The local best value
obtained for the experiments consisting of 50 populations with 1000 iterations of TA algorithm is shown in Figure 5. ufacturing units or MSMEs: collaborative COVID-19 preventive kit man- the minimal The
value is calculated using Eqn. 4.1. uled for the manufacturing site It could be observed that the local best of category 2 is scheduled in 7; face shield value and the manufacturing cost identified for diferent sites vary for diferent populations – i.e., the algorithm searches for the MSMEs or manufacturing units that ofer a minimal manufacturing cost. Even if a lower cost is ulation, the task sets were perturbed in order to search for the best possible solution
among local best task sets
From the available local best values of task sets, the global best value lated by the TA algorithm. It is noticed that is calcu is Rs. 11013.
value is found as 5454; the minimal manufacturing cost is Rs.3734 for manufacturing one preventive COVID-19 kit; and, the manufacturing time is 1620 seconds. The task set that levied the global best result as low: per the TA algorithm is given beC3{P1}, C1{P2}, C2{P3}, C1{P4},
C2{P5}, C1{P6}, C1{P7}, C1{P8} MS7, MS3, MS5, MS13, MS10,
MS4, MS4, MS3
6. Conclusion
The rapid dynamics of the COVID-19 pandemic has manifested the requirement of innovations in various sectors, including the manufacturing/fabrication sector. This article pro- [3] Ali M. Hasan, Mohammed M. ALposed an RFCCPM framework that combines Jawad, Hamid A. Jalab, Hadil Shaiba, the Random Forest and Threshold Accepting Rabha W. Ibrahim, and Ala R. ALalgorithm for enabling collaborative manu- Shamasneh, Classification of Covid-19 facturing of COVID-19 preventive kits in a Coronavirus, Pneumonia and Healthy cost-eficient manner. The framework was Lungs in CT Scans Using Q-Deformed evaluated considering the COVID-19 quaran- Entropy and Deep Learning Features, tine information and MSME enterprises of four- in Entropy Journal, Vol. 22, No. 517, teen manufacturing locations in Kerala, In- doi:10.3390/e22050517, pp. 1 – 15, 2020. dia. The necessity of the RFCCPM framework [4] Alok Baveja, Ajai Kapoor, Benjamin was manifested through experiments that re- Melamed, Stopping Covid-19: A vealed a cost eficiency of 66 percent for the pandemic-management service value identified job schedule. chain approach, Annals of Operations Research, Vol. 289, pp. 173–184, 2020.
oficials, AIC-IIITKottayam, and AIM oficials, for providing constant support through out the research / entrepreneurial career.